This model was built on nearly 1000 races from 2020 to 2024, which provides a fantastic resource not only to model the opinion of the betting market, but also to measure what factors influence a rider’s performance relative to the betting market’s expectations. In short, the market gives us a list of riders expected to perform well and not well in a stage. While their prices to win the race are certainly not 1 to 1 correlation between the order they expect riders to finish in, it is a decent proxy which incorporates the opinions of bettors, recent performance, past performance, estimates of team tactics, and many other factors.

Subscribe now

Modelling finish position by betting market Win price

I set up a model to predict the finish position of a rider in a race just given one input: the implied probability from the betting market (which I log transformed such that 10% is double the impact of 1%). I also log transformed the finish position such that finish 1st = 0, finish 3rd = 1.1, finish 7th = 1.9, finish 20th = 3.0. This model asks “given the Win price for a rider, what is the expected finish position for a rider?”.

The results are significant at about 0.24 R^2, meaning about 24% of variance in finish position are explained by pre-race Win odds. Not bad. The best fit equation is 1.67 - 0.35 * log(implied_probability).

This indicates a large favorite like Tadej Pogacar tomorrow (2.0) should finish about 8th-9th position on average tomorrow, while the 15th favorite - Paul Lapeira - has an expected finish of about 25th. Fabio Jakobsen is the worst lined rider for tomorrow and he is expected to finish 126th on average.

Here are the implied finish positions for the top riders in the betting market tomorrow.

Modelling the impact of other variables

Now that we have baseline expectation for the expected finishing position of a rider based on their betting market price pre-race, we can examine a rider’s performance relative to that expected finishing position based on other factors.

One very interesting variable in the impact of temperature. Various forecasts of tomorrow’s race between Florence and Rimini place the expected temperature in the high 20s or 30s celsius (eg, San Luca racebook site shows 31 celsius as an average for the day). That is significantly hotter than the average race and 31 celsius would be about 89th percentile for heat across all races. We can analyze a rider’s ability to cope with the heat and deliver results relative to expectation by incorporating temperature into the basic model above.

I centered the temperature relative to 21 degrees celsius (a median value across all races) so we’re looking for the impact of relatively hotter or colder temperature on results. The model call looks like lmer(log(finish_position) ~ log(implied_probability) + (0 + temperature | rider))

The results have essentially the same intercept and coefficient for implied probability. The coefficient for the impact of temperature are small; they range from about +0.04 to -0.03. Giulio Ciccone - a climber for the Lidl Trek team - has the coefficient which suffers the most from the heat at 0.038. That means if he is 5% to win a race he is expected to finish 14th place in average temperature (21 C), 21st place in hot temperatures (31 C) and 10th in cool temperature (11 C).

Applying the model output to tomorrow’s prices, using 31 C as expected temperature and comparing to what finishing positions would be with average 21 C expected temperature we get the following for the top riders in market.

Ciccone suffers a drop of about 11% relative to average temperature expectation, while Philipsen gets a 7% boost. Healy (+5% improvement) and Pidcock (+3% improvement) are two others who get better results in hot weather. Pogacar relatively struggles (-6% worse). In fact, among the four major GC contenders only Vingegaard is not expected to get worse results in hotter weather. 3rd favorite Bettiol also has struggled relative to expectations in the heat.

Share

All of that data combined together in an intelligent way can give you strong insight into who is likely to win on the day. This way of analyzing a race can be looked at as the bottom-up approach; take a bunch of factors which could impact the result, combine intelligently, and see what sifts out. Today I want to discuss another way of approaching the problem which can be considered a top-down approach: using past betting market prices to guide you.

Over the past five years, I’ve gathered pre-race odds for almost 1,000 stages/races from the top level of men’s cycling from providers like Bet365, Unibet, Betway, and others. These pre-race market prices are influenced and sharpened by savvy bettors (though certainly not to the extent of eg the NFL or Premier League given the huge amounts bet on those competitions relative to cycling). As such, the prices before a race kicks off are a strong indicator of who oddsmakers and sharp cycling punters think will succeed considering all of the factors I discussed above (performance, parcours, weather, tactics, etc) and others.

We can leverage this data to make inferences about how the market views certain riders, where the market sees them as likely to perform well, and who is trending up or down in form. The market will also likely respond quicker to jumps or drops in form than a machine learning model as it is being shaped day-by-day by the opinions of hundreds of punters who watch and analyze the races.

Subscribe now

How to model this data

First we need to convert data which is in decimal odds format with margin applied to data which is in probabilities without margin (100% price). Converting decimal odds to implied prices is as easy as dividing 1 by the price (eg, 6.5 = 1/6.5 = 15.4%). There are several useful approaches to removing margin from betting market prices; check the implied package for some of them. I use the power method which adjusts the implied price using an exponent (eg, 15.4% ^ (1/k)) and then optimizes that exponent over all prices to find the correct value for k.

This is an example of the top of the market for Tour de Suisse stage 7 last week. Almeida and Adam Yates were significant favorites at 6.0 and 6.5 followed by Rubio, Paret-Peintre, Martinez, and others. When we account for margin, Almeida and Yates come out to around 12% and around 10% to win. Repeat for all 1,000 races and now we have the market’s best estimate of a rider’s likelihood to win the race.

We now need to consider how to model this data. In the past, I’ve modelled data like finishing position using mixed effects linear regression. This approach essentially finds the individual rider-level impacts of various factors on their finishing position. For example, a model which considers the climbing difficulty of a race and whether it ended in a bunch sprint to model a rider’s finishing position will likely have very negative intercepts for riders like Vingegaard and Philipsen (as both finish highly in races), but have a negative slope for climbing difficulty for Vingegaard and positive for Philipsen (as Vingegaard finishes better on races with a lot of climbing and Philipsen finishes worse. Vice versa, the bunch sprint variable will likely be very negative for Philipsen as he cleans up bunch sprints, while being positive for Vingegaard as he does not.1

Such an approach can also be leveraged to model these betting market prices. We can use the same information (climbing difficulty, bunch sprint or not) to find rider-level impacts on likelihood of winning the race.

However, modelling probabilities is very tricky using linear regression. Linear regression can often produce coefficients that lead to negative predicted probabilities or probabilities predicted over 100%. Indeed, a linear version of the model I’ve laid out above sets the intercept term (basically, how good is this rider in general without controlling for climbing difficulty or bunch sprint finish) as a negative value for three quarters of all riders. Sprinter Sam Bennett is predicted to be -16% to win a high mountain stage while climber Tao Geoghegan Hart is -0.5% to win a bunch sprint stage. Obviously neither of those is possible.

Instead of modelling this using linear regression, I settled on a beta regression. Beta regression is perfect for modelling any continuous data which is bounded by 0 to 1 which is exactly what probability data is. The glmmTMB package is specifically designed to fit mixed effects models using a range of non-normal distributions.

After deciding on a framework for the model, I also chose to include two other variables:

• First, the strength of peloton to penalize high probabilities in weak races and boost low probabilities in strong races

• Second, a variable meant to identify the impact on rider’s probabilities of races that are very likely to be contested by the top riders (essentially first days of Grand Tours and one day races) versus potentially not controlled (Grand Tour breakaway days). If the market suspects that a certain race will be punted by top riders to a breakaway and adjusts prices accordingly, we don’t want to penalize those top riders. On the flip-side, we want to sharpen our estimate of a rider’s probability to win using races where the top riders will certainly be contesting the win.

Results

The coefficients outputs are harder to interpret as they must be converted, but we can see who the model considers as the top 10 riders based on the betting market’s opinion. Perhaps only Matej Mohoric is a surprise, but he’s closely followed by Pidcock and Adam Yates - and Mohoric does win often. Jonas, Pogacar, and Carapaz are the riders most positively influenced by tough climbing day, while Bennett, Philipsen, and Mads Pedersen are hurt the most. Groenewegen, Bauhaus, and Jakobsen are the most helped by a bunch sprint finish.

Pogacar, Van Der Poel, and Christophe Laporte are most positively impacted by the one day race/stage 1 of Grand Tour variable. We can immediately see the limitation this variable has; Laporte is indeed a strong rider in one day race, but he will certainly not be the Visma-LAB leader for stage 1 (even if the parcours suited him better). Pello Bilbao, Mohoric, and Michael Matthews are seen as riders hurt by this variable; generally we’d expect riders who gain their success in week long stage races or breakaway days to be hurt by this variable.

Applying to Tour de France Stage 1

Stage 1 of Tour de France is a long (206 km) medium mountain day (3800 vertical meters in 7 categorized climbs) with a ~25km downhill/flat run-in to the finish. In short, it’s definitely going to be a selective day for climbers/puncheurs, but probably not something that will end in a solo unless someone absolutely flies up the final climb and group 2 syndrome bites on the run-in.

Applying that context to the model, we can output predictions for stage 1.

Pogacar is the massive favorite at about 4.3 (or about 7-2 or +350).

Mathieu Van Der Poel leads a host of the second favorites between 17.0 and 23.0 including Roglic and Remco - two of the main GC favorites. The final GC favorite Vinegaard is 6th favorite but only at 33.0.

Van Der Poel is given credit by the model for his typical very strong odds in races where the top riders will be competing for the win, but he’s the most impacted by the climbing difficulty of these top riders and if we were to ratchet up the climbing difficulty slightly, his price would cut in half quickly.

Limitations

This approach is just a tool in the tool box of pricing up who is likely to win a race. Among other things it ignores:

• actual results; this model doesn’t know Alberto Bettiol just won the Italian National Championships last week which is a main reason he’s 3rd favorite in actual market

• races which weren’t priced by the market; many cycling races below World Tour level are simply not priced by any of the bookmakers I track. 6th favorite Maxim Van Gils was surely highly rated to win the Grosser Preis des Kantons Aargau a few weeks ago, but I didn’t manage to find a market for that race

Past Grand Tour Stage 1s

As a bonus, this is how the six road race Grand Tour Stage 1s have played out in the last six years.

• 2024 Giro d’Italia: Pogacar was 1.75 favorite; Narvaez won at 23.0

• 2023 Tour de France: Pogacar was 3.0 favorite with Van Der Poel (5.0) and Van Aert (6.5) next favorites; Adam Yates won at 81.0

• 2022 Giro d’Italia: Van Der Poel was 2.5 favorite over Girmay (5.0) and Ewan (7.0). Van Der Poel won

• 2021 Tour de France: Van Der Poel (2.75), Van Aert (6.5), and Alaphillipe (7.0) were favorites. Alaphillipe won

• 2020 Vuelta: Roglic was 2.75 favorite with Valverde (5.5) second favorite; Roglic won

• 2020 Tour de France: Ewan (3.15) and Bennett (3.25) were favorites trailed by Nizzolo (8.5) and Van Aert (10.0); Kristoff (41.0) won in a crash marred stage

A nice mix of a significant surprise in Adam Yates, outsiders in Narvaez and Kristoff, and massive favorites winning.

The other main storyline was the speed the peloton completed the race in, 46.11 km/h, which is a new record time. That breaks the mark of 2023 (45.77 km/h) which broke 2022 (45.33 km/h). Now five of the eight fastest finishes in history have been set since 2019, with only the 2020 edition raced in August not making the top eight.

Below I’ve plotted kilometer by kilometer over the closing 55 km of the race how fast the leading available finisher raced relative to the other seven finishers since 2017. Eg, Jasper Stuyven in 2021 raced the kilometer between 11 and 10 kms to go (right before the Poggio) in 64.1 km/h which was 9% faster than the average speed of that segment. Blue marks indicate slower than average speeds on that segment and red marks indicate faster than average speeds on that segment.

The stark takeaway is the sheer speed of the closing kilometers of the race in the last four years. The three years starting in 2017 yielded average speeds of:

2017: 50.4 km/h

2018: 45.5 km/h

2019: 47.9 km/h

2020 had a different route due to Covid restrictions and also occurred in scorching weather in August so that can be ignored in the aggregate.

Starting in 2021, the speeds rocketed up with:

2021: 50.7 km/h

2022: 50.9 km/h

2023: 51.5 km/h

2024: 51.5 km/h

Overall, the 2021 to 2024 finales of the race were about 6% faster than the average speed from 2017 to 2019.

Looking more specifically at different segments of the race, the overall speed increase is consistent over each sector of the race, but larger in some areas. For example, the ~30 kilometers along the coast where the peloton hits three small climbs has been raced almost 7% faster in the recent four editions. The flats between the Cipressa and Poggio have also seen speed increases of 8% (though interestingly the 2024 race was bang on average as the pace slowed significantly). The major increase in speed between 2017-19 and 2021-24 has actually been driven disproportionately by faster pace on the small climbs and flat coastal road.

Segment                           2017-19     2021-24   % increase
Coast road/three small climbs     49.5        52.9      +7%
Cipressa climb                    32.7        34.7      +6%
Cipressa downhill                 55.0        58.0      +5%
between climbs                    50.8        54.9      +8%
Poggio climb                      38.5        39.9      +4%
Poggio downhill                   49.9        52.7      +6%
Final 2km                         54.7        58.0      +6%

The sector which has increased the least has been the Poggio - the 3.7 km 3.7% final climb - which has only sped-up by a bit less than 4% in recent years.

Looking specifically at 2024, the fastest two sectors were the Cipressa climb - paced aggressively enough to significantly reduce the peloton at about 8% faster than average - and the final 2 kms on the Via Roma (a special case as multiple attacks forced the pace very hard as the peloton chased first Mohoric, then Sobrero and Pidcock). The slowest sector in a relative sense was the stretch between the Cipressa and Poggio where the pace slowed significantly, attacks went off the front, and the peloton doubled in size. That was only 1% faster than average year.

Below is a clearer look at the final 10km of the race which features the Poggio climb, downhill, and final run in to the finish. These datapoints are consistent with the timed climbing segments by Mihai Simion on his Climbing Records site. Eg, Nibali’s 2018 win came with a climbing time of just 6:13 into a headwind, about 10% slower than the 2024 rider by Pogacar.

Perhaps more interesting is the relative changes in speed. The first part of the Poggio in 2024 was not paced hard as UAE Team Emirates struggled to get organized, but the pace was high enough afterwards to break climbing record.

First, lets see how the projections for 2023 performed.

These were generally much better than the 2022 projections - which is unsurprising given the improved methodology (discussed here). The only two extremely egregious misses were on Lotto-Dstny and Arkea Samsic. Lotto-Dstny performed poorly in 2022 and were relegated out of the World Tour, but responded in 2023 by achieving the 8th best PCS Points total - 70% better than my projection. Arkea received the World Tour spot vacated by Lotto-Dstny’s relegation, but did not perform like a World Tour team and fell 29% short of my projection.

Beyond those two, I picked the top 4 and five of the top six which is reasonable. It doesn’t take a genius to design a system which says UAE and Jumbo will be the two best teams, but the system identified Quick Step bouncing back from 7th to 3rd.

2024 Rider Projections

I’m not going out on a limb with any of the top 20 projected riders. These are the stars of the sport who have performed the best over the last three years. In fact, the top 3 match perfectly the top 3 from the 2023 season and the top 6 are the same top 6 from 2023.

The biggest projected changes versus 2023 are riders like Richard Carapaz (101st to 46th), Enric Mas (52nd to 29th), and Juan Ayuso (45th to 22nd). The simple story-line here is riders who performed well when they raced, but because of injury or DNFs they missed accumulating points. Ayuso returned to racing only in the end of April while Carapaz/Mas both crashed out of the Tour de France on stage 1.

Other riders simply underperformed their established level; Biniam Girmay is projected to jump from 114th to 58th after a down 2023. Ethan Hayter (125th to 59th) and Daniel Martinez (273rd to 115th) are two others in this mold.

Moving in the other direction, Rui Costa is projected to decline the most from 28th in 2023 to 63rd in 2024. Sepp Kuss, Felix Gall, and Einer Rubio are three others who might take a step back in terms of points simply because the model doesn’t see multiple years of top-level point accumulations.

2024 Team Projections

The immediate takeaways for me here are

1. UAE and renamed Visma-LAB (formerly Jumbo Visma) will again be the two best teams. The model is naturally conservative and projects some regression to the mean here, but even significant regression will keep those two on top. Visma-LAB is projected to be a Tadej Pogacar-level season ahead of 3rd placed Bora!

2. INEOS and Quick Step are projected to be eclipsed by Bora and Trek. Both Bora and Trek have invested heavily in the transfer market over the off-season and enter as two of the four most improved projections vs 2023 performance. Inversely, Quick Step and INEOS project to decline by the most and 5th most vs 2023.

3. Jayco AlUla projects to improve from 14th in 2023 to 9th in 2024. They’ve brought in some very useful transfers (headlined by Caleb Ewan, Mauro Schmid, and Luke Plapp) with only a single significant departure.

4. Tudor Pro Cycling has done enough transfer business to put themselves in the tier of Pro Teams worth discussing in these projections. Four of their top six projected riders for 2024 (Trentin #1, Dainese #3, Storer #5, and Mayrhofer #6) are in-transfers.

5. Astana projects as having improved their squad the most for 2024 thanks to 14 in-transfers, headlined by Morkov and Ballerini from Quick Step to fill-out Mark Cavendish’s sprint train.

2024 Transfers

This table shows PCS Points for 2023 for both in-transfers and out-transfers2. As discussed, both Quick Step and INEOS are shipping out the most talent based on 2023 results and Trek, Bora, and Tudor are bringing in the most reinforcements.

Bora is bringing in #5 projected Primoz Roglic in the year’s most significant transfer. They’re also adding Sam Welsford to sprint (#87 projection), Daniel Martinez to support in mountains (#115 projection), and two other top 200 projected riders in Roger Adria and Matteo Sobrero.

Trek is adding #50 projected Andrea Bagioli, #53 projected Tao Geoghegan Hart, and #89 projected Jonathan Milan as well as solid domestiques Patrick Konrad and Carlos Verona. The projection for Geoghegan Hart is likely underestimating his 2024 prospects; his 2023 points total was significantly impacted by his Giro d’Italia crash and his missing the remainder of the season. He was sitting in 3rd in GC after a blistering spring campaign and he averaged the 19th most PCS points per raceday in 2023 among all riders. Bagioli is another rider who might have better 2024 prospects than his projection given he ended the season 3rd-1st-2nd in three major Italian classics.

Top Transfers

We’ve discussed Bora and Trek’s additions, as well as Jayco’s strong business to reunite with Caleb Ewan and add 24 year old Mauro Schmid. They also add Luke Plapp from INEOS; he’s flashed significant potential in recent years and just won the Australian national championship weekend in both the road race and time trial.

Visma-LAB is adding the third best transfer in Matteo Jorgenson - a rider who contended for a Tour de France mountain stage win as well as finished top 10 in Flanders and E3. Visma-LAB is also adding two other U23s with World Tour experience in Cian Uijtdebroeks (#11 best transfer) and Ben Tulett (#18 best transfer).

Visma-LAB is projected to receive the most points from riders who spend at least half the season at 23 or younger - with double that of UAE Team Emirates.

Groupama - FDJ is another team which will heavily rely on younger riders; Romain Gregoire projects as 62nd best in 2024, Paul Penhoet as 98th best, and Lenny Martinez as 107th best. Those are three of the nine best riders racing at least half the season under 23.

Lotto-Dstny has the lowest quality adjusted age which is just a weighted average of predicted points and age. Their top four riders will be 22.2 (Arnaud De Lie), 25.2 (Florian Vermeersch), 26.0 (Andreas Kron), and 24.5 (Maxim Van Gils) on July 1st.

On the flip-side, Bahrain Victorious is relying heavily on riders approaching or over 30.

The final column in this chart shows the percentage of racedays where a rider on that team was the top rider on their team. Eg, EF Education’s riders for 2024 were the team leader in 26% of their racedays in 2023 - largely because of adding a bunch of top level neo pros like Archie Ryan, Darren Rafferty, and Lukas Nerurkar. This can indicate riders being forced to accept smaller roles.

On the flip-side, no roster had fewer racedays as the leader than INEOS at 15.6%. They will be looking for riders like Tom Pidcock and Carlos Rodriguez to step up into leadership roles.

Subscribe now

Overall, the projections performed as mediocrely as you would expect, failing to identify Intermarche as a potential breakout team and unable to point to Quick Step struggling (though my model narrowly had them finishing out of 1st place - but they ended up 7th far off the pace!).

Picking EF to finish 8th when they finished 16th was a sizeable miss, as was missing the emergence of Arkea and Alpecin where I underrated both. On the bright side, I identified BORA - hansgrohe as the team with the 2nd best transfer business and their new riders absolutely crushed with Vlasov finishing 5th, Higuita 18th, and Hindley 19th in PCS points in 2022.

Going into 2023

For 2023 I rebuilt everything. This time I generated individual rider projections based on two other projections: 1) how many points a rider will earn per raceday and 2) how many racedays they will compete in. Feeding into those projections are data points from the last three seasons (2020-22) along with their age, a dummy variable for the covid year in 2020, and what level a rider is competing at (World Tour, Pro Tour, etc).

When predicting points earned per raceday, the model weights the last three years as:

Last Three Years Average = 0.20
Last Two Years Average = 0.26
Last Year Average = 0.54

When predicting racedays, the model regresses much more heavily to the mean and looks more strongly at the last two years of data - while projecting most World Tour riders for at least 30 racedays.

The model is still quite conservative - relying heavily on previous seasons of data with small affects for age and level being raced at. This isn't the type of model which is going to identifying a massive breakout by a neo-pro like Magnus Sheffield had in 2022; the goal is instead to set very reasonable benchmarks for all riders by smoothing out unrealistically strong 2022 seasons, while giving historically strong riders who were weaker/injured in 2022 a little extra credit.

I also don't do anything special with projecting riders for a certain schedule. Riders transferring to teams who heavily race French / Belgian one day circuits are probably more likely to rack up points (vice-versa for those transferring out of those squads). The opposite is true for teams who focus strongly on just World Tour races like EF. I also haven't factored in changes in team schedules; Uno-X in particular should be racing a stronger level of race which could lead to more points.

The Top Riders

Given the conservatism of the model, it's no surprise that the top four riders from 2022 (Pogacar, Van Aert, Evenepoel, and Vingegaard) are the top four riders projected for 2023. These riders could account for multiple grand tours and multiple monuments in 2023. Each has a strong multi-year track record (even Evenepoel still ranked 17th and 18th in PCS points in 2021 and 2020). None of them did anything in 2022 that was unsustainable either.

Below that top four, Roglic, Van Der Poel, Vlasov, and Philipsen are the next tier of riders. While Vlasov and Philipsen are both top 10 riders from 2022, Roglic and Van Der Poel had down seasons due to injuries/crashes/DNFs. Van Der Poel seems to be focusing less on stage racing in 2022 which should help him earn additional points (though the model does not know this!).

Ninth place is where the model stumbles for the first time in my opinion. Adam Yates has a lot of strong racing in his pedigree for 2020-22, but the model doesn't explicitly know he's moving to a team with multiple established GC leaders. I've tried to incorporate forward looking projections of whether a team has too many leaders, but couldn't get anything significant to pop out. The devils advocate in me would highlight Yates is coming from a team where he already had loads of competition for GC leadership and perhaps moving to UAE will allow him to target more one day races and non-Pogacar weeklong races.

Rounding out the top 12 riders are Richard Carapaz (moving to EF and certainly will have more obvious opportunities to ride for himself), Mads Pedersen, and Arnaud De Lie.

Largest Improvements and Declines

Among Pro Tour/World Tour riders who were ranked in the top 500 riders in 2022, the largest projected declines come from older riders who over-achieved in 2022. These are riders like Julien Simon on Total Energies - who ranked in the top 75 riders in 2022 at 37 years old - and Alexander Kristoff - who ranked 11th at age 35. Others like Louis Meintjes, Bob Jungels, and Benjamin Thomas had seasons out-of-line with their recent performance and in many cases are on the wrong side of 30. The largest expected declines are something like losing 33% of points so a rider who earned 600 PCS points in 2022 would be projected for about 400 points in 2023.

On the positive side, riders who were impacted by injury in 2022 are projected to return to strength. These are guys like Maximilian Schachmann, Kasper Asgreen, Julian Alaphilippe, Nacer Bouhanni, and Jack Haig. Alaphilippe scored about 28 points per raceday in 2020-21 which is similar to what the non-Van Aert/Pogacar top riders scored. I have him projected for just 14 PCS points per raceday in 2023 which ranks 28th best among all riders - still an improvement over 2022, but below his top seasons.

Team Level Projections

With a decent baseline established for each rider, we can aggregate their expected points into an overall team projection.

Below I have aggregated projections for the World Tour teams plus the four most significant Pro teams in the peloton. These projections assume Astana completes the transfer for Cavendish/Bol; if they don't Lutsenko projects as their best rider and Astana would fall very close to Uno-X in overall team projections.

I've included the delta vs 2022, the team's projected top rider, and a calculation of the team's Gini Index (basically a larger number mean fewer riders are projected to gain a larger share of that team's points).

The theme of these projections are:

1. The rich only get richer - UAE and Jumbo project to improve by the most and 3rd most due to leveraging strong transfers and additional young talent.

2. Quick Step is projected to return from 7th best to 3rd best in the peloton.

3. Intermarche is projected to decline significantly. Most interesting for them is that they have by far the lowest Gini index among these top teams meaning they have great depth to still be able rank in the middle of the pack, but lack a clear massively productive rider. Perhaps Girmay takes another step forward this year.

4. Lotto-Dstny is incredibly reliant on De Lie and to a lesser extent Ewan to generate their points. Those two are responsible for approximately 35% of their projected points. They've shed four of the top 10 riders from 2022 from a team which was already in the bottom third.

5. A different model loves EF-Education! To be fair, adding Carapaz is a major addition of the projected 10th best rider. Honore is also projected to improve vs 2022 and they have a full season of Piccolo.

6. INEOS might have the most interesting projection; they look like they are losing a lot of talent given Yates, Carapaz, and Van Baarle are out and only Arensman is in, but there is a load of young talent which could be given more responsibility. If I had to pick a team to overperform their projection they would be it.

I also ran the numbers specifically looking at team's aggregates for top 20 riders vs all their riders with the new UCI rule changes to count UCI points for the top 20 riders instead of top 10 in the relegation battle. The results were uninteresting with no team ranking more than two spots different (Team DSM 18th on all riders and 20th on the top 20 riders).

Finally, I wanted to understand which teams had age on their side. I could take a straight average of rider age, but that counts neo pros who aren't going to race a ton or make a huge impact the same as a prime age superstar rider. Instead, I just weighted a rider's age by their predicted points. This produced a very tight distribution of quality adjusted age, but the high and low teams are interesting.

Lotto-Dstny, Team DSM, and Trek have the four lowest quality adjusted ages among these top teams at between 26.7 and 27.0. INEOS ranks fourth best at 27.2 - below even Uno-X. The clear oldest is Israel Premier Tech at 32.6. Astana, TotalEnergies, and Bahrain are the next three oldest at over 30.

Typically, transfer quality has been evaluated immediately prior to a new season by judging the quality of riders brought in (measured perhaps by UCI or PCS points) minus the riders lost. This method works as a rough accounting of riders in and out, but such a method fails to adequately judge transfers by 1) focusing only on the just-finished season performance (ie, a rider is considers to be their most recent body of work) and 2) not accounting for actual performance of the rider (which is subject to impact of age, new training methods, health, race schedule, and ultimately randomness, but is still an important factor to judging transfers).

For example, in the above link Quick Step going into 2022 ranked 18th of 18 World Tour teams in PCS points surplus for riders in vs riders out. They lost Almeida and Bennett, while signing a handful of younger riders. In the end, their five new young riders scored nearly 1000 PCS points (almost +800 vs 2020-21 average), while the riders they lost scored only 1600 PCS points (a drop of -600 vs 2020-21 average). Ultimately, based on improvements by riders coming in and declines by riders going out, Quick Step probably had a top three transfer performance among World Tour teams.

In-transfers

I introduce a new evaluation system which measures the change in PCS points won by riders in year N + 1 as compared to years N - 1 and N, while also looking at the source of a transfer (Continental team, junior/national/club team, or Pro/World Tour level team).

By this method, Quick Step stands out as the superior World Tour team in the period of N = 2012 to N = 2021 on average improving the output of the riders acquired by 81% (eg, an average of 1000 points in years N - 1 and N = 1810 points in year N + 1) compared to a range of -11% to +38% among other World Tour teams.

Just looking at transfers brought in from established professional teams Quick Step also leads with an average improvement of 49% year-over-year compared to a range of -22% to +22% among other teams. Quick Step has signed riders who produced an average of 346 PCS points vs 514 PCS points in their first season with Quick Step. The trailer in this regard is Cofidis who have signed riders with an average of 246 PCS points who have produced 192 PCS points in the first season with the team.

Quick Step isn't necessarily shopping at the top of the market either given that their average PCS points per signing rank just 9th highest (about average) among World Tour teams at 346 per signing (just looking at riders signed from other pro teams). UAE Team Emirates ranks by far the highest with 446 points per signing while Intermarche - Wanty shops at the bottom of the market with 174 points per signing.

Teams don't just sign riders from other professional teams; they have to identify and convince top juniors and U23 riders to join them directly or join the development team with a path to the World Tour. Looking at teams who sign riders from all non-professional sources (Conti level, juniors, national teams, club teams), Quick Step again stands out with acquired riders producing about six times (+521%) the average points they produced in the two previous years. This compares to a range of +19% to +490% for World Tour level teams. These massive gains are because PCS points scales aren't designed to fully reward races below professional levels (.2, U23, Junior, and Nations Cup races) and riders race more often in World Tour.

Since 2012, Quick Step has three of the eight best neo-pro transfers with Evenepoel in 2019, Gaviria in 2016, and Jakobsen in 2018.

The simplest way to illustrate Quick Step's transfer success is to just look at the percentage of signings who score more PCS points in the next year for signings from other professional teams; across the World Tour that value is 40% for all signings. Quick Step manages to improve 60% of their signings - only BORA - hansgrohe crosses the 50% success mark.

Again, we can simply explain Quick Step's success at signing non-professional riders by calculating the percentage of transfers who score 100+ PCS points in the next season (this is roughly the level of the 500th best rider in the pro peloton). Quick Step manages to get 65% of these transfers to this level compared to just 32% of riders on other World Tour teams. Only three World Tour teams manage even 50% of riders scoring 100+ points.

Out-transfers

Quick Step is also well-known as a team which knows when to allow a rider to leave (after a career year will make them too expensive or when age will cause their performance to decline). I repeated the same analysis for teams looking at how riders who left the team performed in the year after (N + 1) compared to an average of the two years prior (N & N - 1).

Again, Quick Step has distinguished themselves by ranking third best over this time period among current World Tour teams with their riders declining by -24%, behind Intermarche at -29% and BORA - hansgrohe at -27%. The trailers in this category (meaning riders leave and win more points) are Alpecin - Deceuninck (+39%) and Cofidis (+14%). Alpecin has had comparatively few and typically unproductive riders leave which might skew the numbers.

Quick Step has four of the seven riders who transferred out and lost the most PCS points dating back to 2012 - Elia Viviani (Cofidis in 2020), Marcel Kittel (Katusha in 2018), Philippe Gilbert (Lotto in 2020), and Michal Kwiatkowski (Sky in 2016).

Limitations

Riders who transfer in after injury or another less typical situation might artificially inflate certain team's transfer successes. For example, it's not immediately clear how much of Mark Cavendish's relative success in 2021 vs 2019-20 was due to 1) Quick Step's unique team situation, 2) improved health/ability to train, 3) dumb luck, or some other factor. In this analysis, Quick Step gets the credit for identifying his talent, convincing him to join them, and executing on returning him to greatness.

This approach also only considers the immediate N + 1 season where many riders are signed for 2+ seasons. Young riders particularly may be signed with an eye towards development by certain teams with the idea that they will pay-off in N + 2 or further seasons. Not considering contract length and performance over a contract is a limitation.

None of this analysis considers salary. A rider who produces less value than in previous years, but who was signed to a cheaper than expected contract could be a successful transfer. A rider who improves year-over-year, but was signed to a more speculative high-price contract may be an unsuccessful transfer. Without knowledge of salary it's always going to be tough to evaluate transfer success. However, a very rough proxy for salary is a riders recent performance. A rider who scores 1000 PCS points and then signs a new contract will likely be paid more than a rider who scores 250 PCS points and then signs a new contract. Thus in the aggregate, we can probably judge a team who consistently signs a rider scoring 1000 points who then scores 800 points the next year as getting poor value.

Also, none of this analysis considers need or role. For example, a team which is desperate for a sprinter to fit in their team might be justified paying for a sprinter who is likely to underperform their prior seasons of performance. Also, a team with a GC captain, but lesser support riders, might be justified signing a few support riders who had bigger roles elsewhere knowing their output will decline, but on aggregate the team will achieve more.

Pogacar vs Roglic?

Pogacar is the bookmakers favorite - significantly - with a price of about 1.77 for/1.90 against at Pinnacle (implied about 52% to win the Tour). Both Roglic and Vingegaard are implied around 12-14% based on odds of about 5.0. Vingegaard has shortened significantly since a strong performance in the warmup Dauphine race as Roglic used to have roughly two thirds of the win probability among the two in March, but that has shortened to about 50/50 or even advantage Vingegaard in the last two weeks. So while this race will certainly be billed as Roglic vs Pogacar, Vingegaard is coming in very strong to have equal odds - especially given the team will certainly defer to Roglic a bit.

Of course the question must be asked whether Pogacar deserves to be the massive favorite? He is likely the biggest Tour de France pre-race favorite since Chris Froome in 2013 (I've seen about equal odds for Froome 2013 as Pogacar this year). In fact, that's still true looking at all grand tours back to that 2013 TDF, so Pogacar's as big a grand tour favorite as we've seen in 26 straight races.

If you average each rider's top 7 GC performances in the last three years - using Pro Cycling Stats points system - Roglic and Pogacar come out well ahead of the competition, but close to one another. The 300+ points accrued by this scale is approximately what Roglic entered last year's race at and similar to where Chris Froome entered the 2017 edition. Froome entered the 2018 Tour at over 400 points on this scale.

In fact, whatever way you slice and dice it, Roglic and Pogacar have gained points in GC races at nearly identical rates. So why is Pogacar such a massive favorite?

My hypothesis is that Pogacar has shown himself more capable of putting out truly dominant performances. I've generated a quick method to find the most dominant stage race performances in recent years. What I've done is strip out riders in the breakaway, and then take the average seconds gained over other top riders in the stage. We're trying to identify performances like Chris Froome's multi-mountain raid in Stage 19 of the 2018 Giro where he won by 180 seconds over 2nd place.

Indeed, among grand tour stages since 2018, Froome's victory in that stage rates #1 with a weighted average of 273 seconds over the chasers. Pogacar's Stage 8 victory in last year's Tour ranks #2, and Richard Carapaz on Stage 14 of 2019 Giro ranks #3. That's a pretty good short-list of dominating efforts - basically multi-climb mountain raids.

Expanding out from grand tours, Pogacar also has >60 seconds gained dominant performances on Stage 6 of 2022 Tirreno Adriatico, Stage 20 of 2019 Vuelta, the Stage 20 Time Trial in 2020 Tour de France, Stage 5 of 2021 Tirreno Adriatico, and Stage 9 of 2019 Vuelta. That's a total of six massive efforts in three years which won him four races and produced his shock podium at 2019 Vuelta.

In the same time period, Roglic has just three similar efforts - all in the final week of the 2021 Vuelta on Stages 17, 20, and 21 (and that Stage 20 effort was a group effort with other riders). Roglic just has not shown the ability to produce massive race-winning efforts nearly as often as Pogacar relying more on a very strong time trial and late attacks on climbs.

Of course, looking physiologically it's also possible Pogacar just has more watts available for longer than Roglic. CronosWatts.com produced a phenomenal article comparing the two riders over their careers back in March, looking at their best climbing performances and the times and watts per KG they produced. I've overlayed their two graphs showing the two riders with Roglic represented by the green trend line and Pogacar in blue.

Frederic Portoleau's conclusion:

The 2 Slovenians have a very similar level in the mountains for durations of effort of less than 25 minutes. For the long climbs, a small advantage for Pogacar. On a climb like Alpe d'Huez Pogacar must be able to achieve a time of 38min30sec or a little less in the event of maximum effort. Roglic for his part, has the potential to climb Alpe d'Huez in 39 min.

Frederic Portoleau from https://www.chronoswatts.com/news/203/

Perhaps you can argue Roglic is faster on the sub 10 minute climbs which might allow him to steal some time on the finishes of stages 6, 8, 9, and 14, but overall they are even on efforts like those faced on Planche de Belles Filles on Stage 7 and Peyragudes on Stage 17. Pogacar has the advantage on climbs like Col du Granon (Stage 11), Alpe d'Huez (Stage 12), and Hautacum (Stage 18) - at least using these historical values.

The same site has published summaries of 2020 and 2021 Tours with major climbs. Unfortunately they don't include the Planche de Belles Filles time trial in 2020 or the full Ventoux ascent where Pogacar was dropped by Vingegaard and lost 40 seconds in 2021. They also include non-competitive climbs like Stages 6/16 in 2020 where GC riders were not riding full gas. Looking at the summary Average Standard Watts Pogacar beat out Roglic by maybe 0.5% in 2020 and Vingegaard by about 2% in 2021. The missing data works against us here, but just using their times in lieu of power estimates, Pogacar rode 8% faster than Roglic on Planche de Belles Filles and Vingegaard rode full Ventoux ascent 1% faster than Pogacar. Combining those values with other climbs says Pogacar has been about 1.5% better over the last two Tours. That seems like enough to call him a clear favorite.

Teammates

The Pogacar vs Jumbo-Visma battle won't just be confined to those three riders; depending on tactics we could see one of Jumbo-Visma, UAE Team Emirates, or even INEOS Grenadiers try to control the race by leveraging their teams. In fact, we could see this early on potentially windy stages like 2, 3, and 4 or the cobbled stage 5.

I've ranked team quality on x-axis of who has the better time trial riders and y-axis of who has the better classics/one day riders. This might give us an indication of who is best setup to support their riders on the flat or hilly days where there is wind or cobbles in play.

Jumbo-Visma is the clear leader here as they have very strong time trial riders supporting Roglic like Van Aert and Vingegaard while also having strong classics riders like Laporte and Benoot. UAE is one of five strong teams behind Jumbo-Visma along with INEOS, Quick Step, and Bahrain, and BORA. Advantage Jumbo-Visma, but this isn't a chasm like between Jumbo and Movistar.

Moving to the mountains, during the last two Tours there's been a lot of talk about Pogacar's team not being strong enough to support him, while Jumbo-Visma has been seen as a super team with multiple GC contenders lining up to support Roglic. Measuring a climbing domestique's ability to support GC riders is still definitely not a solved problem, but I've tried leveraging my rider ratings which identify how good riders are at racing certain parcours based on their finishing position. Lower values below indicate better expected finish positions across the top 4 support riders on each squad.

In 2020, Jumbo-Visma had by far the best climbing domestiques to back up Roglic and they rode a defensive race which delivered Roglic to the final time trial with a minute advantage. Roglic's teammates could only watch as Pogacar made up the difference and won the Tour.

Last year, Jumbo-Visma again had a wide advantage over UAE, though INEOS was strongest, but a strong team was less important after Pogacar's incredible first week and the team had an easy job to protect a five minute lead after nine stages.

This year, Jumbo-Visma will again have an advantage over UAE, but only because of how much stronger their lineup is this year. Both squads have improved vs 2020 and 2021. Sepp Kuss is likely the best pure climbing domestique in the race - ranking 14th in my climber rating - which will allow Jumbo-Visma to have something like three of the final 15 riders in the lead group. UAE added veterans Marc Soler and George Bennett over the offseason which should give Pogacar's team something like 5 riders in the last 40 riders in the lead group compared to just two in 2021.

Other Contenders

Based on betting odds and making reasonable assumptions about where the vig is on the GC winner market, books are pricing Pogacar, Roglic, and Vingegaard at something like 80% for one of them to win. That leaves about a 20% chance of a big surprise whether from a former winner like Geraint Thomas, a perennial contender like Yates or Quintana, or one of the younger crowd of podium contenders like Vlasov or Enric Mas.

This gives INEOS perhaps a 6-7% chance of winning their first Tour in three years. Thomas has just won the Tour de Suisse - one of two big warm-up races, but only after the favorite Vlasov left with a positive Covid test. Martinez had an incredible spring with a win in the Tour of the Basque Country and podiums after two other big races, but looked undercooked at the Suisse warmup and has a best grand tour GC result of just 5th. The final INEOS rider Adam Yates ranks as the third best performing rider on climbing stages in the race, but has just two 4th places in his GC career largely due to a poor time trial and big drop-off in performance in later stages of races.

Of the remaining riders, Mas, O'Connor, and Haig will likely be done in by the 40km time trial on stage 20 where they could easily lose two minutes plus to the Slovenians/Vingegaard. O'Connor was the final rider dropped by Roglic/Vingegaard on the final stage of the Dauphine tune-up race and while he finished 4th last year, he benefitted from gaining 6.5 minutes on other GC riders in a breakaway and likely wasn't the 4th best rider in the race.

Vlasov has a string of strong week-long GC performances in the spring including a massive win in the Tour de Romandie mountain time trial, but he left Tour de Suisse with Covid. If he's back on form he has a decent podium chance as his team support ranks 3rd best in the mountains with a strong squad of Austrian/German climbers.

Who are the best sprinters?

I've written this year about my Bunch Sprint Model which evaluates sprinting success based on finishing position solely in sprints a rider contests while also considering the strength of opposition a riders sprints against again considering only those opposing sprinters who contested the sprint. You can read more about methodology and results at this link. Think of this model as looking to identify the best sprinters if they all had a chance to sprint against one another.

Back in February this model loved Fabio Jakobsen due to a very high hit rate in sprints he actually contested. Since then, Jakobsen has continued to sprint at a high level with six wins on a variety of parcours, while Jasper Philipsen performed very well at UAE Tour and then hasn't done much since. This model evaluates the two of them as neck-and-neck on top of the sprinting world.

Behind them, the model rates Cavendish 3rd and Olav Kooij 4th. Cavendish failed to be selected, while Kooij also couldn't rate selection on Jumbo-Visma's GC focused squad. Caleb Ewan (5th) and Wout Van Aert (6th) are the other two elite sprinters at this year's Tour. Van Aert has only participated in a bunch sprint seven times in 2022, but six have been podiums.

Further down the list are veterans like Alexander Kristoff, Dylan Groenewegen, Peter Sagan, and Mads Pedersen. Kristoff landed Stage 1 and the yellow jersey in 2020, but hasn't won a World Tour sprint since. Sagan has had multiple covid bouts, but finally landed his first win in Tour de Suisse a few weeks ago. Pedersen might be more of factor on the more classics-like finishes as three of his five 2022 victories have come on either uphill finishes or finishes with a small climb right before the finish.

Groenewegen has been a bit in the wilderness due to his suspension, being eclipsed by younger riders at Jumbo Visma, and his subsequent transfer to Bike Exchange. He's won five times this year, but has only a single podium finish in World Tour sprints. He's actually won his last three contested sprints across three sub-World Tour races, but was dropped on several climbs in Dauphine and left that race without contesting a sprint.

The only other sprinters it makes sense to mention are Team DSM's Alberto Dainese and Bike Exchange's second sprinter Michael Matthews. Dainese won a shock victory in Stage 11 of the Giro, but doesn't have another finish better than 5th in a sprint all year. Matthews is really more of a tough parcours sprinter at this point in his career as his only wins since 2020 have come in one day classic Bretagne Classic and on a tough stage of the Volta a Catalunya this spring.

Ewan's Lotto Soudal Team Changes Strategy

Caleb Ewan's Lotto Soudal team has a well established approach to grand tours since landing Ewan in 2019. They've brought the 6th, 2nd, and 2nd heaviest lineups to the last three Tours de France and 4th, 5th, and 4th heaviest to the 2019, 2021, and 2022 Giros - driven by big engines like Roger Kluge, Jasper De Buyst, and Thomas De Gendt. Beyond the size of Ewan's teammates, they relied on experienced riders to back Ewan, regularly trotting out lineups where 3-4 riders had ridden 25+ bunch sprints with Ewan in recent seasons. That element will be different in 2022 as while they will again have one of the heaviest starting squads, Ewan's teammates have very little experience supporting him in sprints. Riders like Kluge, De Buyst, and recent additions Michael Schwarzmann and Rudiger Selig were left out in favor of more classics focused engines like Frederik Frison and Florian Vermeersch. It will be interesting to see if their tactics shift more towards Ewan surfing wheels rather than utilizing a big sprint train.

Philipsen's Rise

Jasper Philipsen was always a highly touted sprinter, landing three World Tour stage wins before his 23rd birthday, but he was squeezed out of UAE Team Emirates by veteran sprinters and the team's GC focus around Tadej Pogacar and transferred to Alpecin for the 2021 season. There he has blossomed into potentially the best sprinter in the pro peloton thanks to a massive 2021 season.

His 2021 story was inextricably tied to Mark Cavendish as he was Cav's main opponent in his breakout Tour of Turkey in 2021 (Cav landed four wins to Philipsen's two) and again he kept coming up short to Cavendish in the Tour de France where Philipsen reeled off six stage podiums, but couldn't score a win. Philipsen followed up the Tour with two Vuelta stage wins and four one day race wins, including on cobbled terrain. If the cobbled stage 5 turns out to be less vicious than expected it wouldn't be a surprise to see Philipsen sprinting for the win as he's handled similar terrain in the past.

Bike Exchange All in For Groenewegen

Before Bike Exchange's team announcement there was all possibility of them sending a balanced team to chase GC or stages with Simon Yates, but instead they've gone all in on Groenewegen and Matthews as of their announced team only Nick Schultz is anything of a climber. They will likely have the heaviest lineup at the Tour at 73.75 kg; that would also be the heaviest of any team at the Tour since Lotto Soudal's 2016 team built around Andre Greipel and his sprint train. Bike Exchange will hope that extra power will allow them to keep Groenewegen up front during potentially windy stages 2, 3, and 4.

Fabio Jakobsen + Michael Morkov

Michael Morkov's dominance as a leadout man has been well established in recent years as he's guided Sam Bennett and Mark Cavendish to back-to-back green jersey wins and six Tour de France stage wins. Of course combining him with Fabio Jakobsen should produce good results. However their success in 2022 has been massive with Jakobsen winning five of seven sprints he has contested with Morkov in the lineup.

It's difficult to know which are the best leadout riders on Quick Step as the team is just phenomenally well drilled overall, but it seems like Jakobsen has ridden more with the 'B' team than the elites. Since last July when Jakobsen started sprinting again he's ridden most often with Florian Senechal and Bert Van Lerberghe in bunch sprints. On the other hand, Cavendish has been most often deployed with Morkov and Davide Ballerini.

I've written in the past about how tough it is to evaluate sprint helpers and that the best guide may just be to look at how teams deploy riders in different races. With a start in 2022, Morkov will now have raced four straight Tours for Quick Step (2019-22) - as well as the 2022 Giro with Cavendish. Compare that record to Senechal and Van Lerberghe. Senechal has ridden just two grand tours (neither the TDF) in that time period, while Van Lerberghe has similarly raced just two grand tours and won't feature in this TDF either. While Quick Step's full sprint train might be a bit lighter than past years, combining Jakobsen with Morkov could still produce tons of success.

Van Aert's Green Jersey Bid

Van Aert appears to be a massive favorite for the green jersey points competition as his odds - even after a minor injury last week - sit at 1.65 (implied at around 56%). He benefits from a race bereft of many true sprint stages (only four are evaluated in the Tour regulations as flat stages: 2, 3, 19, 21) where true sprinters could challenge him, while there are also a lot of classics-esque and medium mountain stages where he should find great success from a reduced peloton.

This graph above shows the result of a model which considers whether a rider was able to survive in the group and sprint for the win (finish in top 25 in a bunch sprint race) depending on the climbing difficulty of the parcours. Something like stage 6 into Longwy or stage 8 in Lausanne would rate a 5.5 on this scale, while the flat stages 2 and 3 would rate <1. Stage 10 which features a long, shallow drag to the finish in Megeve would rate just off this scale ~8.5.

Van Aert, Matthews, and Philipsen show fairly consistent ability to stick with the front group as climbing intensifies, but the other sprinters show degraded abilities on tougher terrain - particularly Jakobsen and Groenewegen.

It will be a huge advantage for Van Aert that he has been >85% to survive to the sprint finish regardless of the difficulty of the climbing that day. His ability to survive on those tougher sprint days like stages 4, 6, 8, 13, and 15 and to even get into the break on harder days will make him tough to defeat.

Looking at the last 10 points competitions, we can split up the source of points between finish-line sprints and intermediate sprints. Finish-line sprints can be accrued by being a great bunch sprinter, while intermediate sprints can be accrued by getting breakaways or by tactically out-sprinting opponents on the road. In the last 10 competitions, the green jersey winner ranked 1st in points from finish-line sprints 8 times and 2nd twice. The record on intermediate sprints was more mixed with Kittel, Cavendish, Bennett, and Sagan twice taking green without gathering the most intermediate sprint points. We haven't recently seen someone take green by dominating intermediate sprints and not being one of the two best on finish-line sprints. Will Van Aert be one of the two best sprinters on finish-line sprints?

If not, the market shows Jakobsen and Philipsen as the best odds among pure sprinters. Jakobsen won the points jersey at 2021 Vuelta and might look to follow Cavendish and Bennett and Quick Step green jersey winners. Philipsen was fairly close to Jakobsen in points competition when he abandoned Vuelta, but he managed only 4th in green jersey race in 2021 Tour de France due to hardly contesting intermediate sprints.

The issues with that Log Rank model are: 1) it considers all different parcours of races in building the overall impact data point, not just races ending in bunch sprints, 2) it considers all bunch sprints for a rider, even those where a heavier sprinter was jettisoned a climb and failed to participate in the sprint finish, 3) it considers bunch sprints where a rider was present in the bunch, but was actually helping a teammate (eg, Davide Ballerini often sprints for himself in smaller races, but is in the sprint train for bigger ones), and 4) it doesn't consider the quality of the sprinters participating alongside each rider in the sprint (ie, the competition on that day may be much reduced by tougher parcours, mechanicals, crashes, or splits in the bunch).

So how to account for these issues. First, we want to just evaluate sprinters based on bunch sprint finishes. Anything which doesn't end in a bunch sprint is ignored by this new model. Second, we want to ignore any race for a rider where they didn't finish with the first group in the sprint AND in the top 25 positions; this indicates they were capable of sprinting. Third, we want to ignore any race where a rider wasn't the top finisher on the team. Many riders participating in as a lead-out man can rack up 10th place finishes which can pollute our understanding of them as sprinters in races where they compete as team leader. And fourth, we consider the cumulative strength of the sprinting field which meets these first three criteria based on the simple Log Rank model outputs.

Determing strength of sprinting field

How does point #4 above work in practice? Seventeen sprinters in UAE Tour stage 1 on Sunday qualified for these criteria including the top 13 finishers. My basic Log Rank model predicts following finishing positions in a generic strong race for those seventeen riders.

A lot of very talented sprinters were in this race - including seven with an expected finishing rank of 7.0. Compare to stage 1 of Tour of Oman where the top sprinters were Fernando Gaviria (5.0), Mark Cavendish (7.0) and no one else with a predicted rank better than 10.0.

To determine the cumulative strength of sprinting field, I just take the reciprocal of each rider's predicted log rank (1 / predicted log rank) and add them together. A top sprinter like Bennett or Philipsen will contribute 1/3 or 0.33 points while someone with a very low prediction like Canaveral or Schwarzmann will contibute 1/40 or 0.03 points.

The top races for sprinters tend to be the Tour de France, Milano-Sanremo, Paris-Nice, and UAE Tour with cumulative strength of sprinting fields around 3.0 to 4.0 depending on the specific day. World Tour races in general average just under 2.0, with a wide range, while .Pro races average just above 1.0, again with a wide range. The lowest pro races at .1 level tend to average just below 1.0 with hardly any rating better than 1.5.

With that data calculated, it is simple to specify a model using this strength of sprinting field and rider to predict both finishing rank and whether a rider won the sprint. Both of these models find 1) the impact of individual rider on success metric and 2) a potentially non-linear impact of the cumulative strength of sprint field.

To Predict Finishing Rank:
gam(log(finish_rnk) ~ rider + s(strength_sprint_field))

To Predict Win:
gam(win ~ rider + s(strength_sprint_field))

I ran both models for this example on data since the start of 2020, only considering riders who participated in at least 16 sprints meeting the criteria laid out above. This ranged from Wout Van Aert with 19 sprints to Philipsen/Ackermann with 45.

Who is the top sprinter in early 2022?

Both models produce similar results given the data. Fabio Jakobsen is seen as the most likely sprinter to win a given race and the sprinter who will finish with the best finishing position overall. For example, in a typical World Tour level sprint the models predicts Jakobsen to win 53% of the time and finish an average of 1.9. Wout Van Aert is predicted 2nd in win probability at 44% and 4th in finishing rank at 2.7. Sam Bennett is tied with Caleb Ewan for 3rd in win probability at 38%, but slightly ahead of him for 2nd place in finishing rank at 2.2. Ewan is predicted at 2.6 in finishing rank.

Those four comprise a fairly clear top group with Jakobsen fairly clearly the #1 sprinter in the world. Behind those four are guys like Philipsen, Groenewegen, Cavendish, and Demare. As a sign of his diminished form in recent years, Peter Sagan ranks outside the top 25 in predicted win probability and 15th in predicted finishing rank.

Fabio Jakobsen

Looking at the data in this way it's obvious why Jakobsen is the top predicted sprinter while ranking only fifth in the PCS Sprinter Ranking and 14th (!) in my own basic Log Rank model. Jakobsen had three week long stage races in his comeback from serious injury last year where he didn't compete as a sprinter. Basically, the basic Log Rank model sees a guy who was "awful" at sprinting for a dozen sprints. But, when we restrict just to races where he was the team leader and he was in the sprint pack, the graph below shows he has been dominant.

Jakobsen is winning nearly 70% of his sprints where he is the leader and is contesting the sprint since the start of 2020. That blows everyone else away, with Van Aert and Ewan managing only a mid 40% win rate in that time. Jakobsen has raced lesser competition than guys like Van Aert, Ewan, and Bennett, but he's dominated that competition.

One of the big stories of this and last cycling season is Mark Cavendish's return to massive success with Quick Step, including tying the record for career Tour de France stage wins. He has twelve wins since the start of 2021 - including two this season - and easily rates as a top 10 sprinter in the world right now. Because he and Jakobsen race for the same team, only one of them is likely to make the Tour de France team where Quick Step sprinters have been steered to 14 sprint stages in the last five races. Unfortunately for Cavendish, Jakobsen isn't simply just another top 10 sprinter - he's the best in the world right now.

As we move into a new season, teams have hired new riders and let others go. I had a go at making some basic projections on how teams strengthened or weakened their squad with transfers, age based regression, and natural regression/progression in points earned. One of my caveats in that article was that the projections did not account for team strategies or rider schedules based on transfers. There are only so many leadership positions to go around and teams who hire more leaders are at risk of needing to demote some leaders to support roles in certain races.

In this piece I define "leader" as the top finisher for a team in a race. Of course the top finisher is not always the rider(s) who were designated as the leader at the beginning of the race. However, the top six riders in % of races as leader in 2021 were Nairo Quintana, David Gaudu, Giacomo Nizzolo, Guillaume Martin, Aleksandr Vlasov, and Tadej Pogacar so I think it's a reasonable proxy.

How This Plays Out For Teams

A quick example, UAE Team Emirates transferred in five major signings who spent at least some races in 2022 as their team's leader - sprinters Pascal Ackermann and Alvaro Hodeg and climbers Joao Almeida, George Bennett, and Marc Soler. They transferred out four major riders who spent some races as leaders - sprinter Alexander Kristoff, climbers Joe Dombrowski, and David De La Cruz, and puncher Sven Erik Bystrom. Five in, four out. The riders leaving were UAE's #1 rider on a race day 52 times. The riders coming in were their team's #1 rider on a race day 76 times. UAE also hired wunderkind climber Juan Ayuso who was the leader in a race - primarily at U23 level - 17 times in 2021. In total, they raced 233 times in 2021 and the riders on their team for 2022 were the #1 rider on their team 288 times - a surplus of 55 races.

We can repeat that same calculation for the other 17 World Tour teams and actually most teams have a surplus; 11 have at least 7% more leaders in their team than 2021 race days, another 5 are within +/- 3%, and only Lotto Soudal (6% fewer) and DSM (13% fewer) aren't equal or with a surplus. Overall, the surplus is 12% at World Tour and 8% at Pro Tour level. This makes total sense. Teams tend to discard riders who don't have the capacity to be leaders anymore and hire those that do as a natural progression of the sport. However, some teams are legitimately going to be squeezed for leadership opportunities in 2022 - even if we don't see Covid related cancellations like the prior two years.

EF Education is probably the most over-subscribed in terms of leaders. They hired riders who were team leaders 102 times in 2021, but got rid of riders who were team leaders just 36 times in 2022 - a surplus of 66 races. Their issues might not be as extreme as represented here as many of their additions come from non-World Tour level teams and/or are developing riders who might need a year before becoming full-fledged leaders. In fact, only 77% of the 2021 leaders on EF came while racing for a World Tour level team (86% is the average for the full World Tour).

All Leaders Aren't Equal

We need to account for the difference acquiring a leader like Giacomo Nizzolo (who finished 1st on his team 60% of races at World Tour level) and one like Marijn Van Den Berg (moving to aforementioned EF Education) who led his team in 45% of races at U23 level. If we arbitrarily assign a weight of 1x for leaders while riding for World Tour teams, 0.67x for leaders while riding for Pro Tour teams, and 0.33x for any other leaders, we can get a better idea of how much competition there will be for leadership roles. At EF, they now rank third with a surplus of about 18%. The World Tour in general averages a 3% surplus by this method.

Doing that weighting shows BORA and Jumbo Visma as the two with the most competitive leadership competitions. BORA ranked 4th best in adding talent through transfers per ProCyclingStats and 2nd best at adding talent by my projections. They added climbers Aleksandr Vlasov, Jai Hindley, and Sergio Higuita who combined to lead their team 62 times in 2021, and sprinters Danny Van Poppel and Sam Bennett who combined to lead 41 times. Sprints-wise, they should be fine as they're also losing Peter Sagan and Pascal Ackermann (47 races as leaders) and Van Poppel has also said he's switching to support Bennett.

Where BORA will see the squeeze is in general classification and hilly/mountain stage leadership. Just filtering to leadership in hilly/mountainous races, BORA rode 94 races in 2021, while their currently employed riders were leaders of their team in such races 146 times! That's a greater than 50% surplus - far beyond any other World Tour team.

Flip that around to flatter/classics races and Jumbo Visma looks to be the team with the most issues with too many leaders. Despite moving star sprinter Dylan Groenewegen onwards, they've still a tight squeeze. They have a surplus of 36% due to adding Christophe Laporte (punchy sprinter), Tosh Van Der Sande (leadout man), and Tiesj Benoot (classics rider). What looks most likely is that those three will simply sacrifice more of their own ambitions to support Wout Van Aert in classics and young sprinters like David Dekker and Olav Kooij in flatter races.

I wrote in my 2022 team projections piece about DSM's losses in the transfer market. They were especially hard hit in the climbers/GC riders department where they lost Michael Storer, Jai Hindley, Tiesj Benoot, and Ilan Van Wilder. Those four combined to lead in 34 of DSM's 102 hilly/mountainous races in 2021 and the other transferred out riders combined for 9 more for a total of 42% of DSM's races being led by riders leaving the team. They only added a sprinter - John Degenkolb - from a World Tour team, with the rest of their additions coming from lower level squads. Still on the team is Romain Bardet (leader in 63% of hilly/mountainous races he entered), but no one else who led in more than 20% of their hilly/mountainous. In races without Bardet, they'll be handing out leadership opportunities to their wide array of young climbing talent and hoping for quick development.

Competition for Leadership vs Depth/Optionality

The flip-side of framing this as an issue of too many leaders is that talented riders who were leaders in smaller teams can now move up and support superstars like Van Aert. The team also has cover in case of injury; for Jumbo Visma, if Van Aert suffers an injury their spring classics season isn't completely ruined as they can plug in competent classics riders like Benoot or Laporte.

BORA just released their preliminary plans for the three grand tours, but they also have the option within those plans to either leave off a rider who is struggling with form or choose to fully back a rider in strong form for GC. Between Buchmann, Vlasov, Kelderman, Hindley, and Schachmann they have riders who have finished 4th in Tour de France, 4th in the Giro, 2nd/3rd in the same Giro, and won a World Tour stage race in back-to-back years. And that ignores Higuita, Konrad, and Kamna who have won grand tour stages in the last three years. There's definitely option value there in knowing that you can select the best of that bunch for your main focus in major races.

A few points on methodology:

1. I've used the PCS Points from ProCyclingStats.com at the rider-level to build these.

2. I've built a very basic model for projecting points which only knows what a rider did the previous season, how old they are (age matters!), and whether their team is at World Tour or not. Only riders competing in the following year on a World Tour or Pro Tour (and ProConti for past years) level were modelled. Obviously what happened in 2019 and 2020 is relevant, but I will leave a more advanced model to next season.

3. All riders with a Pro Tour or World Tour contract as of start of this January were predicted for 2022, with their projected points aggregated to determine the collective points projections for each team.

4. That's it. I did nothing to account for 2021 injuries, changes in how riders would be deployed across races, and any #gainz which may have occurred over the off-season. This is certainly wrong as many riders who missed large chunks of 2021 will race full schedules in 2022 (Caleb Ewan, Remco Evenepoel, etc.) and we're already seeing injuries to riders like Mathieu Van Der Poel which will affect points earned in 2022.

Rider Level Projections

A model which just considers the previous year's performance + age and level of team will tend to produce projections which closely match the rankings from the previous year. The same top five from 2021 is projected to be the top five in 2022, while young stars like Remco Evenepoel (17th in 2021 to 11th in 2022) and Ethan Hayter (26th in 2021 to 17th in 2022) are projected to improve their ranking. Older riders are projected to decline with Alejandro Valverde (12th to 37th) and Mark Cavendish (22nd to 32nd) being the sharpest expected declines.

Above graph describes how riders tend to retain points from year 1 to year 2. Peak age riders tend to regress about 20% or said another way they retain about 80% of their points the next year. Riders who score highest number of points tend to regress more in year 2, while those scoring closer to zero point in year 1 regress less. Younger riders tend to hold onto their points the most (though even highest point riders here tend to regress more). Older riders fall off significantly with a rider scoring 1000 points in year 1 at age 35 retaining more like 70% of their points in year 2.

However, these rider projections are fairly dumb; a projection system which ignores Mark Cavendish doing nothing for four seasons before resurrecting his career is probably not going to make great specific projections for riders. Where I hope the projections do well is at the aggregate team-level where the errors of predicting 25-30 individual riders can cancel each other out.

Team Level Projections

Based on individual rider projections/performances, I created three different team totals: 1) 2021 points earned by the team, 2) the 2021 points earned by the riders employed for 2022, and 3) the projected 2022 points earned by the riders employed for 2022. This way I can calculate who hired the best new riders vs who lost the best riders vs who has riders most primed to improve or decline. Delta due to rider development shows how riders are expected to earn points differently in 2022 vs 2021 due to age or regression. Delta due to Transfers shows how teams added either better or worse riders based on 2021 points. Eg, EF Education hired better riders based on 2021, while Lotto Soudal hired worse riders. However, Lotto is expected to improve due to age in 2022.

Of the 18 World Tour teams, I see EF Education improving the most versus 2021 - primarily due to transfers. They signed four top 200 riders in my PCS Points projections (equivalent to a ~top 10 rider on the average team) including Esteban Chaves (projected as their 2nd best rider in 2022). They have only lost three riders who signed with a Pro Tour or World Tour team - headlined by Sergio Higuita (103rd best rider in 2021).

BORA-Hansgrohe is another who looks set to improve significantly due to incoming transfers. They signed the 26th, 29th, 86th, and 88th best riders in my 2022 projections with Sam Bennett hoping to return to his 'best sprinter in the world' form. They also added the aforementioned Higuita and Aleksandr Vlasov. BORA loses two strong riders in Peter Sagan and Pascal Ackermann, but they should come out ahead on aggregate.

Lotto-Soudal and Team DSM should improve primarily from internal development of younger riders. DSM is by far the youngest team in the World Tour but has a lot of the early 20s riders who tend to increase significantly. DSM do have to deal with significant losses due to transfers as they were the hardest hit team in percentage terms. Lotto also has a lot of younger riders and do not have any significant regression candidates as their top scoring rider in 2021 was Tim Wellens at only 65th in PCS Points. Caleb Ewan will also presumably have a healthier season (21st and 11th in 2019-20 PCS Points).

My model also projects Quick Step to not lead the World Tour in PCS Points in 2022 (falling just short of UAE Team Emirates by 200 points). They have led in total points accumulated every year since 2013, but the model sees significant riders lost (Joao Almeida in particular ranked 5th in 2021) and significant decline from its current crop of riders (Cavendish in particular). However, the model doesn't know Quick Step basically got half a season each out of two very promising young riders in Remco Evenepoel and Fabio Jakobsen. My bet is the Belgians manage to pull off their tenth straight #1 ranking by the end of the year.

Among Pro Tour teams, three teams stood out in 2021: Alpecin Fenix out-earned twelve World Tour teams, while Arkea Samsic and Team TotalEnergies earned points like the weakest World Tour squads. The projections see modest regression for both Alpecin and Arkea driven by regression for their top performers and not particularly strong transfers. Team TotalEnergies added Peter Sagan - once the best rider in the world - and should be improved by 30% due to their quality of transfers, but they also are a quite old team which means their gains will probably be more modest in the end.

Among others, Kern Pharma is a very young team which should improve due to aging of their riders. They also signed Hector Carretero from Movistar World Tour team who would've ranked third on their team in points in 2021. Along with that, they lose only a single rider from 2021.

Uno-X is a team which the projections aren't particularly high on, but which may be able to improve in ways the models are ignorant of. They are adding Tobias and Anders Halland Johannsen - two elite U23 riders who finished 1st/2nd (Tobias) and 7th/8th (Anders) in the two major U23 races in France and Italy. The U23 points scales on PCS are probably underweighted relative to the difficulty of those races so the Johannsen's are better positioned to earn points. Not accounting for new opportunities / lesser opportunities for transferred riders is another blind-spot of my model.

Research on many team sports indicate varied "peak" ages for players between early 20s and 30s for different sports. For example, this Baseball Prospectus piece reviews three different approaches and finds somewhere between 26-28 as peak age for hitters. This CJ Turturo piece examines the impact of aging in NHL hockey and finds age 22 as peak for forwards, age 24 for defenders, and age 27 for goalies (part II of that document). Others in studies quoted by Turturo have found 25 for forwards, 22 for defenders, and 24 for all skaters. In 2013, I found golfers peak in their early 30s, which makes sense as golfer is less of an physically demanding sport compared to baseball or hockey. In a later study, I found different aging curves for different skillsets within golf.

I applied similar methodology to these studies above to identify the aging curve in cycling, from which we can derive a peak age and determine how much we should expect young cyclists to improve and old cyclists to decline. Using the delta method where a rider season is compared to the following rider season identified a peak around 26-27 with riders improving before that age and declining after that age. Using the GAM method where a curve is fit to all rider careers identified 26-28 as the peak with riders improving before those ages and declining after 28. The two methods differ in the steepness of the aging curves; delta method shows a steeper curve of improvement < age 25, while GAM method shows a less steep curve of improvement at those ages and a much sharper decline from age 35 onwards.

Methodology and Data

I gathered PCS points per season (raw total) for each rider between 2010 and 2021. PCS points are awarded for race finishes, GC finishes, and points/mountains jersey finishes. The top points scorers tend to reflect who is considered the top riders, but in my opinion they overweight success in one day races and underweight success in stage races (in the individual stages). Nevertheless, they are a well-accepted and discussed data point which is available consistently going back over a decade.

One thing to consider is accumulating PCS points is part performance and part opportunity. A rider who at age 22 races for a Continental level team as the leader in U23 races and at age 23 races for a World Tour team as a domestique will have fewer opportunities to earn points (though improved performance may cancel that out and there are always freaks like Pogacar and Evenepoel).

I also adjusted points earned in 2020 and 2021 to account for the impact of Coronavirus on races being held. 2020 had 14% fewer points earned and 2021 had 3% fewer points earned than an average season.

Important to note: I am using age on June 30th of that season as the age for that season when binning, but otherwise am using continuous ages relative to that June 30th date. Eg, Peter Sagan (January 1990 DOB) is considered as a discrete age 32 in 2022 (as he will be 32 on June 30th) and a continuous age of 32.4 in 2022 (as he will be 32 and 5 months on June 30th). Some other websites report current age and/or use discrete ages which will make ages look lower.

DELTA METHOD

For the delta method, I simply compared points accumulated by a rider in year 1 to those accumulated by that rider in year 2. I used the rider's age on June 30 of year to determine the rider's age for that season. The delta method just measures the change between year 1 and year 2, averages across all riders at that age, and ascribes the total average change to aging. My yearly age samples for seasons in the mid-20s were over 1500 and were over 500 for all seasons between 20-33 and over 100 for all seasons between 19-38.

Riders improved their PCS points from 19-20, 20-21, and 21-22 by an average of 88% (eg, 100 points to 188 points). Age 22-23 and 23-24 earned improvements of an average of 38%, followed by 15% from age 24-25. At that point, performance was fairly steady from 25-26 to 28-29 at between up 5% and down 4%. The peak age seems to be from 26 into 27.

Performance starts declining more significantly as a rider moves into their 30s (an average of -12% down for 29-30, 30-31, 31-32, 32-33, and 33-34). The sharper declines follow that, averaging 25% down from age 34 onwards.

GAM METHOD

For the GAM method, I built a non-linear model which aims to approximate the average aging curve for the full population of riders across their career. The model is in the form of PCS_PTS ~ s(age) + rider so that the overall model finds the average curve over a career; the rider term allows for the height of the curve to vary between massively successful riders like Froome and Cancellara and lower level riders who have scored few points. I included all seasons where riders were between 19 and 38 years old (the ages for which I had > 100 rider samples) and all riders with 4+ seasons in my 12 year sample (using 4+ or 6+ seasons did not impact results).

The aging curve produced was very similar to the delta method. What differed was that the growth curve for riders at 23-24 and under was also much shallower (average of 30% from 19-20, 20-21, 21-22 instead of the 88% from delta method and average of 15% from 22-23 and 23-24 instead of 38% from delta method). The decline curve was sharper after age 35 with 35-36, 36-37, and 37-38 meaning an average decline of 48% instead of 25% shown by delta method.

This GAM method graph should be interpreted slightly differently as the average progression for a rider throughout their career. The Delta method graph just shows the average change in points season to season at each age. Notably, older ages feature better riders (eg, age 34 is actually the peak for average PCS Points per season because you've filtered down to riders who have aged better than the average rider).

What does this mean for 2022?

Summarizing the results of these two approaches, we can see 1) riders tend to improve in earning PCS Points thru age 24 into age 25, 2) riders tend to earn similar PCS Points from age 25 through age 29, and 3) riders start declining in PCS Points earned from 30 onwards, accelerating from age 32-33 onwards.

Among top riders who are in that age 32-34 range we have sprinters like Elia Viviani, Giacomo Nizzolo, and Matteo Trentin, punchier riders like Diego Ulissi and Ion Izagirre, and climbers like Mikel Landa, Primoz Roglic, and Rafal Majka. The most prominent rider who switched teams over the winter was Peter Sagan who will be 32 for the entire season. Some of these riders will decline - some precipitously - while others will fend off age and produce just as strong as season as 2021.

In the aggregate though, these aging curves suggest teams which are more comprised of 30+ year old riders will fall-off more than those with younger riders. Among the World Tour teams, Israel Start-up Nation had the oldest roster in 2021 and now again in 2022 with their 30.8 year old average. Based on this aging curve, their riders are set to decline by 5% on average from their 2021 point totals. Since last year, their major additions were Nizzolo (age 33), Jakob Fuglsang (age 37), and Hugo Houle (age 31).

Team TotalEnergies races at the Pro Tour level and is the team which signed Peter Sagan (along with several of his mid to late 30s support riders). Their average team age ballooned from 28.9 in 2021 to 30.6 in 2022. They are also in-line for a 5% overall decline in their performance versus 2021. Those aren't huge declines, but considering the salaries being paid to stars like Nizzolo, Fuglsang, and Sagan and the performances expected, they will be fighting against that current to produce.

The younger teams most likely to improve collectively in 2022 mostly race at that Pro Tour level. Equipo Kern Pharma, Sport Vlaanderen, UNO-X, and Bardiani will all average under 25 years old in 2022. Those four are projected to improve just by aging by 7-9% in 2022 versus their 2021 performance.

However the most interesting team is Team DSM in the World Tour. DSM has added eight riders in 2022 - six of them under 25 - while they lost their two oldest riders from 2021. They are the only World Tour team with an average age under 27 in 2022 (25.7 years old). They are expected to improve collectively by around 5% versus 2021 performance by the aging curve. Riders like Kevin Vermaerke, Thymen Arensmen, Mark Donovan, and Andreas Leknessund all fit the bill of having previous World Tour experience + being aged 23 and under.

A lot of the work I did with professional golfers was related to scheduling: where they would be able to play their best golf and where that best golf would be most rewarded by the arcane point system in professional golf. I've applied that type of approach below to identify: 1) how difficult it is to achieve different results in stage races and 2) where those results are disproportionately rewarded by cycling's own arcane points system.

First, it's not much more difficult to achieve results in the three three-week long grand tours than it is in the week-long stage races in the World Tour. Generally, it's the same set of riders competing for those results whether it's the Tour of the Basque Country or the Vuelta a Espana.

Second, grand tour success is heavily rewarded relative to other races. GC positions which are equally difficult to achieve can be rewarded 2x more in grand tours relative to those other World Tour stage races and sometimes 3x more in grand tours relative to other lower level stage races.

These two findings explain why teams and riders compete so much for minor top 10 placings in grand tours even when those minor placings are ~10 minutes back of the GC leader.

Difficulty of achieving GC results

The easiest way to compare the difficulty of achieving a GC result in one race vs another is to simply compare results within the same rider/season. Eg, Tadej Pogacar raced five stage races in 2021 coming in 1st in UAE Tour, 1st in Tirreno Adriatico, 3rd in Basque Country, 1st in Slovenia, and 1st in Tour de France. Based on those five finishes and completely ignoring any context around them, we might judge Basque Country race as the toughest as Pogacar failed to win there.

However, we have hundreds of similar comparisons between these races just from the last decade of results. 227 riders have ridden Basque Country and Tour de France in the same season in the last eight years. 123 have ridden Basque Country and Tirreno Adriatico, 36 have ridden Basque Country and UAE Tour, and 30 have ridden Basque Country and Slovenia. We can leverage those comparisons to judge the relative difficulty between each pair of two races.

Above I've shown these aggregate difficulty comparisons for Tour of Basque Country and the ~40 races with at least 30 comparisons in 2014-21. They're ordered by difficulty where the last column value is the expected finishing position in Race A (Basque Country) given a 5th place GC finish on Race B. Eg, if a rider finishes 5th in the Tour de France they would be expected to achieve an equivalent of 4.3 in Basque Country.

Pogacar's 2021 races are highlighted in red where Tour de France is the toughest, UAE and Tirreno are similar difficulty to Basque Country (expected finishes of 5.2 and 5.5), and Slovenia is viewed as much easier with expected finish of 18th in Basque Country for someone finishing 5th in Slovenia.

This method confirms the primacy in difficulty of the Tour de France as every comparison race is easier to achieve results in than the Tour. However, it also shows the two other grand tours are not any more difficult to achieve results in than the bigger week-long World Tour races like Basque Country, Tour of Catalonia, Tirreno Adriatico, Paris-Nice, and the Dauphine. A 5th in the Giro d'Italia is worth about 5.6 in those five races on average. A 5th in the Vuelta a Espana is similarly worth about a 4.6 in those five races on average. A 5th in the Tour de France is worth a 3.8.

Scaling all races versus those five week-long stage races shows the following hierarchy:

Below I've included my top 20 GC riders entering the Tour of the Basque Country in April and whether they raced Basque Country and the three grand tours in 2021. 12 of the top 20 raced Basque Country - including the two clear best - while 14 of the top 20 raced the Tour de France, only 5 of the top 20 raced the Giro, and 12 of the top 20 raced the Vuelta.

UCI Points

There are two popular point system in professional cycling: the unofficial ProCyclingStats points - which I've referenced before - and the official UCI points - which determine team eligibility for the World Tour and other qualifications. The point system is explained well here by INRNG, but basically the UCI decides how to group races together and assigns them different points (eg, the Tour de France is its own category awarding between 18-25% more points for a given placings than the Giro or Vuelta and more than 3x more points than the minor World Tour stage races).

We can use those scales to make equivalencies between what the UCI thinks are similar finishes. Eg, a 12th place finish in the Giro or Vuelta is worth 8th place in Catalonia or Basque Country, 13th place in the Tour de France, and 7th in UAE Tour. Moving outside the World Tour races, that 12th place is worth 2nd place in a 2.1 stage race (Tour of Sicily, Route Occitanie) and 4th in a 2.Pro stage race (Arctic Race Norway, Tour of Denmark).

The UCI is saying 5th place in the Tour de France is worth 1st place in those big week-long stage races, 4th place in the Giro/Vuelta, and more than 1st place in every other stage race.

Rewards vs Difficulty

We can combine these two difficulty measures from my research and UCI point scales to find which races over and under-reward finishing highly. I use my research from above to find equivalent performances and then look to see how those are rewarded between races. Eg, I found it roughly equal in difficulty to finish 5th at Vuelta a Espana and Basque Country. However, Vuelta rewards finishing 5th with 2x the UCI points as Basque Country.

Basque Country is rewarded between 50-75% as much for eleven World Tour stage races, but is rewarded 2x as much for the minor 2.1 level stage races like Valencia, Besseges, and Alpes Maritimes. These races attract difficult fields, but are shorter/lower level so they receive fewer points for equivalently difficult results.

The Sweet Spot

So earlier I said the sweet spot in terms of rewards were grand tours with the inverse being some of these week-long stage races at World Tour level. That is without factoring in the time the race takes (grand tours require 24 days of racing with rest leading in and recovery time leading out while the Basque Country is only 6 days of racing with many riders taking just a week before and after between races). Obviously if you're optimizing at the team level with this data, you'll factor greater time commitment for grand tours and the position of races on the schedule into planning.

Many of these effects are driven by different strength of fields in different races. I've shown GC ratings for riders before (others like PCS have similar rankings). Aggregating those ratings by race yields this plot where the x axis shows the strength of the riders in that race. Eg, Tour de France has the strongest field of GC riders, followed by the Vuelta. Part of the reason results are difficult to achieve in Catalonia and Basque Country are because they rank 3rd and 4th in strength of their riders. The Giro shows up as having a relatively weaker field (more comparable to the week-long stage races) which means the rewards in terms of UCI points are higher for equivalent positions.

Evaluating sprinters within this eco-system is difficult. Javi Angulo has a recent piece using the Glicko method (using head to head results) to rate sprinters where he rates top sprinter at end of 2021. Teun van Erp and Rob Lamberts (with an assist from multi-TDF winner Marcel Kittel) use video and power analysis to analyze the determinants of sprint performance in a 2021 study. Besides these detailed analyses, we can use stats like win rate/podium rate/average rank/PCS points won to judge sprint performance at a more superficial level.

But what about measuring the impact on sprint performance of teammates?

I've designed a handful of methods which could illuminate our knowledge on this subject. These methods all have clear flaws - most notably that we do not know team strategy/roles within team.

Simple win or loss calculations

At the most basic, we can assume teammates share equal credit for victories in bunch sprints (and equal penalty for not winning). Therefore we can calculate each rider's team win rate in bunch sprints (I've included all 2.1/1.1 or higher races where 20+ riders finished within 3 seconds of the winner).

For 2021, the top results are above. If you follow World Tour rosters closely you'll realize that all ten of these riders are Quick Step riders - the team which year-on-year dominates the sport. The results are not so Quick Step biased in every year, but they show impact of multi-collinearity (a fancy statistics way of saying that it's difficult to tease out the unique impact of multiple different factors when you rarely observe them apart from each other).

Eg, in 2020, the top five riders are Arnaud Demare and his sprint train. At least three of them + Demare raced together in bunch sprints 24 times in 2020 (16 times with all five riders together). Demare raced in a bunch sprint just one other time and the other riders were in just five bunch sprints apart from Demare. So was Demare's extreme success in 2020 (11 wins in 25 races) mostly his dominance, mostly his teammates dominance, or a mix?

We can look at larger samples of seasons to try to look at changes in team personnel. However, looking at Demare's team going back to 2018 still sees 112 bunch sprints where 66 came with at least three of his sprint lieutenants in the lineup and another 35 with two of the four riders present. FDJ has raced 234 bunch sprints without Demare and won 4% of the bunch sprints vs 26% in 112 races with Demare.

With all four of his helpers, Demare has won 37%. With three of the four, he's won 23%, with just two he's won 17%, but with 0 or 1 he's won 27%. So while it looks like more sprint helpers/more familiar helpers means better results, we're not much closer to saying how important Demare is vs his helpers vs his team.

We can expand this analysis and just credit riders for team wins where they were present in the bunch at the end of the race. For example, Tim Declercq was only in the bunch for 5 of 24 bunch sprints in 2021 season vs 72% for Michael Morkov. Perhaps if we're analysing who has the most impact on bunch sprints, we should ignore riders who weren't present? Of course, it's likely that without Declercq some of those bunch sprints would have turned into either breakaway victories or a late attack would've been launched or Quick Step would've burnt other riders essential to the sprint train.

Who Does the Team Trust?

Too much of sports analytics is results oriented, but there is loads of information that is not captured by the result. Teams are privy to training data, injury data, interpersonal relationships, and much else that we can access by looking at how team directors choose their lineups. Eg, analytical models based on results would probably tell you Cavendish shouldn't have been in the start list for the Tour de France, but Quick Step had seen enough from Cav to think he was the best option to sprint.

We can look at which start lists to see who teams trust to race alongside their best sprinters. To do this analysis, I built a simple rating system for sprinters based on their finishing positions which easily discriminates between the best sprinters (eg, Bennett, Ewan, Van Aert in 2020-21) and also-rans. I then looked at how riders in a team were deployed alongside sprinters. Eg, if Quick Step sends Morkov to a race with Sam Bennett, but Pieter Serry to race with a lesser sprinter like Alvaro Hodeg that might say something in the aggregate.

For 2020-21, Sam Bennett raced with the highest quality best sprinter (obviously as he is one of the best in the world and was the best sprinter on his team in every race he participated in). More interestingly, Morkov was the clear 2nd place rider as he raced with Bennett in 42 of 62 bunch sprints (and 42 of 46 bunch sprints that Bennett participated in). At the bottom of the list are riders who were mainly deployed alongside weaker sprinters or in lineups without a clear sprinter like Honore, Serry, and Vansevenant.

For the full peloton, the top 10 of riders who were not the best sprinter on their team a majority of races is below:

We see Morkov appear along with three of Caleb Ewan's dedicated sprint train from Lotto Soudal and four of FDJ's previously discussed sprint train for Demare. Theuns typically rides with Stuyven and/or Pedersen and Consonni was the main support for Elia Viviani.

At the other end of the list are riders typically not deployed with top sprinters at least on teams that have one.

We can also simply look at which riders raced alongside the top sprinters in the highest percentage of bunch sprint races from 2020-21. Eg, for Sam Bennett the importance of Morkov is obvious as Morkov featured with him in 91% of bunch sprints vs 48% for the next highest Quick Step rider. Caleb Ewan relied on both Roger Kluge and Jasper de Buyst for 84%+ of his bunch sprints. The aforementioned FDJ sprint train represents four the eleven most common pairings between top sprinters and a teammate. The other standout was the pairing of Gaviria-Richeze for Team UAE. Richeze missed just one of Gaviria's bunch sprints in 2020-21.

Advanced Statistical Models

Statistical models for teasing out multi-collinearity do exist. One promising approach used in sports like basketball or hockey is regularization (where coefficients are penalized using Lasso or Ridge regression). Running lasso regression on this type of data essentially produces coefficient estimates which are often zero if the model can't determine that the term is significantly impacting the results.

To set up the data, I've filtered first for riders with 60+ bunch sprints in the last four seasons (2017-20) and then set-up a matrix with a 1 in the rider's column if they were in that race or a 0 if not. This produces a matrix with over 400 riders. The regression is just run on a binary outcome of a win for the team or not (could run a similar regression of finishing position or podium as well). I've also controlled for quality of the best sprinter on the team for each race and the level (.1, Pro, or World Tour) of the race.

Training the model on 2017-20 and testing on 2021 yields interesting results with honestly some nonsensical results at the individual rider level (the top impact rider is Damien Howson - a climber who apparently improves his teams chances of winning a sprint from 6% to 17%). Many of the other top impacts seem believable; eg, I could believe Davide Ballerini increases his team's chances of winning from 6% to 12% because someone has to be responsible for Quick Step's incredible ability to win sprints year-after-year.

Turning to the predictions at race level in 2021, Quick Step's Volta ao Algarve squad in stage 1 was seen as best positioned to win a race of the season. That is driven by Sam Bennett being really good, but also the model viewing riders like Jakobsen, Morkov, Asgreen, Archbold, and Ballerini as all having strong impact on winning. They are viewed with 47% probability of winning stage 1. With Bennett, but with six neutral teammates, they would have only 15% probability.

To evaluate the model I compared it to the predictions with a neutral model (just considering ability of best sprinter). If this model says anything valuable, we'll see a difference between the quality of prediction for model and the neutral model. We can also compare to a completely naive model which just assigns every team a probability of 1 / N teams in race.

The mean square error of these three different predictions show the neutral model improves on the naive model by 0.031 or about 7% improvement towards perfect. So knowing how good the best sprinter on the team is vs knowing nothing is worth about 7% gain.

The Lasso model which judges impact of individual riders is worth another 4.5% of improvement over the neutral model. So knowing the impact of all riders on probability of winning is about two thirds as valuable as knowing the ability of best sprinter.

My sense is this model is not ready for primetime as there is no correlation between the coefficient produced by the lasso model and how often riders are deployed with the best sprinters. It seems likely that judging a rider's ability to impact sprint results based off how their team deploys them yields the most useful information.

So any rider performance statistic has to be based on one of those three data-points: time gaps, points, or finishing rank. Each has its place.

Time gaps are a very poor way to evaluate success in a bunch sprint where 100 riders might finish on the same time, but they can be a good way to evaluate success on a mountain stage with an uphill finish.

PCS points have been developed into a widely used evaluative method which recognizes that success in cycling can be achieved in a wide array of competitions (GC, race wins, jersey competitions) and has dozens of different scales which are used for different quality of races, but fundamentally their point scales are opinions on the value of different results relative to each other.

Finally, finishing rank is often used to count victories, podiums, or top 10 finishes across the season, but is plagued by vastly different difficulty levels to achieve certain results (how good is 3rd in a World Tour race relative to 1st in a .1 race?). Ranks are often notoriously difficult to take averages of; Wout Van Aert's transcendent 2021 Tour de France yielded an average rank of 25th for a return of 3 stage wins, just behind Enric Mas's 6th on GC with nary a stage podium finish.

In recent months, I've developed my own tweaks to use finishing rank as an evaluative method, producing a stat I'm calling Log Rank. The handful of keys to make it work are:

1. All finishing ranks in a race are transformed by taking the natural logarithm. This produces a value system where the difference between finishing 1st vs 5th are large, while the difference between finishing 50th vs 100th is not as large. The red dots below show equal gaps between results; so 1st and 3rd are separated about as much as 3rd and 7th/8th. However 1st and 7th/8th are separated equally as 7th/8th and 55th. I think this is a fairly intuitive appraisal of the value of different finishing positions.

2. Using these transformed ranks, taking averages are much easier. For example, Wout Van Aert's final week of Tour de France where he finished 25th, 40th, 36th, 43rd, 1st, 1st (average 24th) are transformed into 3.2, 3.7, 3.6, 3.8, 0, 0 (average log rank of 2.4) which can be re-transformed back into average rank of 11th (by taking e^x where x = average log rank). Basically, this says we care way more about Van Aert's two victories than the fact he finished outside the top 20 in those other races. In fact, he could have finished 50th in those four stages (new average of 34th), but his log rank would only change to 13th.

3. The difficulty of different races are found by an objective system which looks at how difficult it is to achieve certain results in different level races. For example, in recent seasons it is roughly similar difficulty to achieve a 10th place in an U23 2.2 level race as a 27th place in a World Tour race. Using a host of these type of comparisons, I've created a Strength of Peloton rating system to judge all level of races against each other based on the difficulty to achieve certain levels of results. All that needs to be said here is that results are adjusted based on what type of races they were achieved in. For example, Ethan Hayter and Tadej Pogacar achieved very similar raw finishing ranks in 2021, but Pogacar did so against the 4th toughest pelotons and Hayter only around the 600th toughest.

2021 Log Rank Rankings

Applying those three steps yields the following top 10 for all 2021 results, just averaging all race results (ignoring time trials):

Building on Log Rank

The next challenge was to build on this basic Log Rank to add in parcours level impacts of things like the climbing difficulty and whether the race ended in a bunch sprint. For example, Enric Mas raced 66 times on the road in non-time trials in 2021. If we're judging how good of a rider he is we probably don't care about where he finished in the flatter stages which littered the Tour de France and Vuelta a Espana. However, we care a lot about how he performed in the tougher climbing stages of those races and others.

The find the impact of climbing difficulty and a bunch sprint finish I set-up a mixed effects model which can be run over results from a given period of time (eg, July 2019 to June 2021 to predict performance going into the 2021 Tour de France). The model was specified using three random effects involving individual riders attempting to find a) their general level of ability to finish with a good finishing rank in races b) the impact of climbing difficulty on their finishes, and c) the impact of the race ending in a bunch sprint on their finishes.

lmer(log_rnk ~ (1 + climb_difficulty | rider) +
 (0 + bunch_sprint | rider)

Using this model, we would expect a sprinter like Sam Bennett who struggles in the hills and mountains, but generally ranks highly in terms of finishing rank to have a smaller individual coefficient (indicating that he generally achieves high finishes), a larger climbing difficulty coefficient (indicating that as races get tougher in terms of climbing his finish rank get larger/worse), and a negative bunch sprint coefficient (indicating that he finishes with better ranks when the race ends in a bunch sprint vs smaller group).

The model results for July 2019 to June 2021 show Bennett with about the 50th best general ability to finish highly (a above), the 20th worst impact of climbing difficulty (b above), and the 2nd best bunch sprint impact (c above). Overall, he would be expected to finish with an average rank of 3.7 in a flat, bunch sprint race - 2nd best in world between Wout Van Aert (3.1) and in front of Caleb Ewan (3.8).

We can similarly look for hilly races not ending in bunch sprints (prototypical classics race) where Mathieu Van Der Poel had the best prediction at that time at 6.8 - essentially tied with Wout Van Aert - and ahead of Roglic, Pogacar, Van Avermaet, and Alaphilippe.

The top predictions in high mountains race were unsurprisingly the three main recent grand tour winners: Pogacar, Roglic, and Bernal. They were followed by Mikel Landa and Adam Yates.

Fundamentally, cycling races can be viewed as war of attrition. As I laid out during this post and Dr. Seiler discusses in his video here, races normally feature a long stretch of steady efforts, before the pace is ramped up towards the end. This steadily increasing pace towards the back-end of the race is what creates separation between riders in most races. The exception is in some primarily flatter races which just do not feature the type of topography which results in time gaps and so the peloton finishes together in a bunch sprint. In all other races, separation is typically created - particularly on hills and/or mountainous sections of the race - but also on cobbled roads, gravel/poorly surfaced roads, in crosswinds, etc.

So that is a fan's understanding, informed by some limited studying of power outputs on significant climbs across a large sample of races. However, we can leverage an even larger data-set of individual race segments on all types of flat, uphill, downhill, poor surface roads, etc. I've gathered a data-set of rider speeds on different length race segments primarily from 2020 professional season to do just that. There are 22,500 unique segments in this data-set covering 177 races.

What Produces Separation?

The metric of choice for showing separation is the time difference between 90th percentile in speed and 10th percentile in speed on a segment, divided by the median speed over that segment. Eg, if 90th percentile is 27 km/h, 10th percentile is 20 km/h, and median is 22 km/h the Separation Factor is about 32%. That is fairly high among all segments where the mean is 12% and median is 7%. The max Separation Factor for the average race is around 48% - typically a short segment.

Essentially treat the Separation Factor as the percentage difference in speed between riders racing the fastest and those racing the slowest. On the nine decisive final climbs in the 2020 Tour de France the Separation Factors averaged 29%, ranging between 18% for Col de la Loze and 39% for Orcieres-Merlette.

Separation is primarily created by higher gradients. This is maybe the most blindingly obviously statement I've ever made, but there it is. Flatter or downhill segments created very little separation among the group on average, while uphill segments create increasingly more as the gradient increases from about 3% to over 10%.

When comparing segments on cobbles vs similar gradient segments on normal roads the rougher roads show a highly statistically significant difference of about 5 to 8% larger Separation Factor for cobbled sections vs normal roads, depending on how it is modelled (a model with gradient included tends to diminish the impact as many cobbled/white road sections are also uphill). The impact here is roughly a Separation Factor of 9% for a flat, non-cobbled segment vs 14% or higher for a flat, cobbled segment.

And replicating the work done previously showing that power varied more in later stages of the races, segments further towards the end of a race provide for more separation than those earlier in the race, with the most significant increase in roughly the last third of the race.

Where is Separation Largest?

Clearly segments further through the race have the highest separation between fastest and slowest riders. But where does the moment with the largest separation occur in these races? For this sample of 177 races, the key moment on average is 88% through the race, with about 40% of races having this key moment in the last 3% of the race (or last 5km for a typical 180km race). It's important to note a segment is counted as occurring based on where it ends within a race.

Again, to any fan of cycling the knowledge that the largest time gaps occur near the end of a race - particularly on summit finish climbs - is not novel. However, this data does show how rare it is for the segments which produce the largest time gaps to occur anywhere in the first half of the race.

So far this has largely been exploited only as a social media activation tool for NTT (eg, on Twitter @letourdata). But knowing the position of every rider with their speed is obviously powerful information. For example, who was pulling in the lead group to try to extend the gap on stage 7 of this year's Tour de France? How large was the group at the bottom of each final climb? How much time did Zakarin lose to the leaders on stage 8 on descents? Which Jumbo Visma domestique drove the pace the hardest on the climbs?

Leveraging this data, I've analyzed ten of the hilly or mountainous stages of this year's Tour de France to look at the probability of staying with the front group (defined as the group with Primoz Roglic as he was in yellow for the lion's share of these stages) over the stage. I've decided to ignore riders who spend the stage in the breakaway, but anyone who attacked away from Roglic (eg, Pogacar in stage 8) counts as surviving as well.

Survival Probability for Notable Stages

Stage 8 was a short 141 km stage with three major climbs - Col de Mente at 82 km to the finish, Port de Bales at 37 km to the finish, and Col de Peyresourde at 11 km to the finish. Col de Mente did little to shake-up the peloton and almost all riders were able to come together to the bottom of the Port de Bales - a 12 km HC climb. That was where the major selection on the stage came; by the end of the climb, less than 20% of the riders in all rider types except Climbers had been distanced from the GC group. About 60% of Climbers survived Port de Bales with the GC group.

The selection for climbers came largely on the Peyresourde and about 30% of climbers survived with Roglic to the end of that climb (with nine riders finishing on the same time from the GC group).

Compare that with Stage 4. Stage 4 was not particularly selective before the final climb with a handful of category 3/4 climbs leading up to the 1st category climb to Orcieres-Merlette. At the end of the final warm-up climb at 20 km to go at least 50% of domestiques and sprint train riders were still there along with upwards of 75% of puncheurs, mountain helpers, and climbers. The non-climbers were distanced quickly on the final climb, but it wasn't until the final few kilometers of the stage that the selection was made among climbers and even then over 60% of them came to the line with Roglic (with 16 riders finishing on the same time).

Stage 17 had two HC climbs - the Col de Madeleine summit came with 64 km left and the race finished on the Col de la Loze. The major selection here came very early on the Madeleine where already only half of climbers were left in the front group with 5 km to go on that climb. Riders were steadily distanced on Col de la Loze until the leaders came over the line with massive time gaps. The first six riders came in alone and there were 17 different groups in the top 20 riders.

Most Selective Climbs

The four most selective climbs for Climber rider types were Col de la Loze on Stage 17 (50% of climbers at beginning vs 12% at end), Montee du plateau des Glieres on Stage 18 (72% of climbers were left at beginning of climb and 22% at end), Col de Peyresourde on Stage 8 (54% at beginning and 27% at end), and Col de Madeleine on Stage 17 (100% at beginning and 50% at end).

For the full peloton, Col de la Loze (Stage 17) was the most selective overall, cutting the peloton down to less than a sixth of its size before the climb. The Madeleine (Stage 17), Port de Bales (Stage 8 penultimate climb), and Glieres (Stage 18) were the next most selective - each reducing the peloton to a fifth of its prior size.

With enough data it would be interesting to tease out the most important factors to make a climb selective. Is it the length, the gradient, a combination of the two, the position in the stage? Based on this limited sample of 50+ climbs, the two most important factors are the length in kilometers (long climbs are more selective) and the overall difficulty in terms of vertical gain (gradient * length). The difference in length and vertical gain between a typical HC climb like the Col de Madeleine and a 1st category climb like the Orcieres-Merlette climb is about five times more important than the difference in distance to the finish between a climb 100 km from the finish and one which is a summit finish. However, that is a very weak claim with only a dozen stages worth of data.

Other Survival Probabilities

Next I'm going to explore power outputs over the course of a race by exploiting power files for climbs throughout a race. I have over 15,000 unique rider/climb power files showing power output, gradient, distance, and position of each climb throughout a stage. These cover over 300 riders, for nearly 200 unique races, and over 650 different climbs within those races.

Where does power output diverge?

A simple model of a bike race is explained clearly by sports scientist Stephen Seiler in this Youtube video. Over the course of the race, riders raise the level of the race by raising power output. This steadily winnows the pack down. In a very selective race it may winnow down to 1 or 2 riders; in a less selective race you may go to the line with most of the peloton remaining. In the case of the 2020 World Championship Road Race in the video above, the race was very selective leading to a final group with six of the best riders in the world.

Exploiting these power files, we can draw a curve of the power required throughout the average stage. Interestingly, the curve does not follow this simple model of steadily increasing power. There's a spike in the first quarter of the race where presumably the breakaway is being established, but power declines in the last 20 percent of the race.

However, this does not invalidate Seiler's point as this graph considers all riders in the peloton. Of course, as riders are shed by the peloton because they cannot maintain the steadily increasing pace, they drop their power output and continue on to the finish at some lower pace. This is best illustrated by the gruppetto concept in stage races; each rider doesn't struggle to the finish at the best pace they can maintain. Instead, they are happy to reduce their power output and save their energy for another day.

When I re-create the chart above only stratified into riders finishing top 10, 11-25, 26-50, and 51-100, we see a clear divergence in power over the course of a race. And for those in the top 10 we see a steadily increasing curve from around 4.80-5.00 watts/kg in the first half of the race to about 6.00 watts/kg in the final stages. For riders finishing outside the top 100 we see a steady decline in power output from again around 4.70 to 4.80 watts/kg in the first half to around 4.20 watts/kg in the final stages.

That divergence really appears around 60-70% through the race. Riders in the top 50 but not top 25 can hang on until around 65% through the race, while riders in the top 25 but not top 10 can stick until about 80% through. Of course, nothing is ever this cut and dried considering different parcours with more or fewer climbs, but this gives an idea of the averages across pro races.

Power output by rider type

Leveraging the rider cluster types introduced here and further discussed here we can draw similar curves for six basic rider types. You can see three clear groups: climbers, mountain helpers/puncheurs, and sprinters/sprint train/domestiques. These shows obvious trends where non-climbers are declining in power in the last third of the race and climbers are increasing in power over that time.

Implications

Most interesting is how this model can be applied to identify the toughest obstacle for a rider in a race. A lot of cycling commentary focuses around whether certain riders can overcame a tough climb with the final group - often because that rider has a faster sprint and will beat those riders if they can stick around. These findings can be applied to identify where that most critical obstacle to overcome is located within a race.

To do so, we have to establish a baseline level of power output. This should not be the average level or the level maintained by the peloton at the start of the race, but instead some lower level that is theoretically the floor for a rider in the World Tour/Pro Conti peloton. I would propose using that roughly 4.00 watts/kg level that riders outside the top 100 finishers maintain in closing stages of a race.

We can then scale power outputs relative to that baseline 4.00 watts/kg level. To stick with the final group we can look at what is required for top 10 or top 25 finishers on final climbs. On average, this is in excess of 5.50 watts/kg up to 6.00 watts/kg. This is about 1.5-2.0 watts/kg higher than our baseline. We can also look at what is maintained by riders in the first half to two thirds of the race. That 4.70-4.80 watts/kg level is about 0.7-0.8 watts/kg higher than our baseline. To bring it all together, the curve below approximates the power required to get over the average climb with the lead group based on that climb's position within the race - all scaled relative to a final climb.

We can see the first 60% of the race or so requires about 35-40% of the power over baseline compared to the final climb. This increases steadily then from that point. The first 25% is likely slightly higher as the breakaway is established here in many races.

An example from 2019 Il Lombardia is shown below with climb difficulty estimates for each climb. These climbing difficulties are based on gradient, distance, and elevation of each climb so that a typical category 1 climb in a grand tour is around 10.0 and a typical category 2 climb in a grand tour is around a 5.0.

Lombardia had six climbs in 2019, two of the 1st category difficulty, three of 2nd category difficulty, and one of the third category difficulty. However, because of how the easiest climb (Battaglia) was placed in the race, it was likely to require more power to overcome than the three 2nd category climbs in the first 180 km. The Sormano climb is the objectively tougher climb without considering position compared to Civiglio or others, but Civiglio's later position in the race makes it a much more equal comparison.

By Time Duration

Leveraging over seven thousand power files, I can link rider types to power outputs over different time durations. I chose to look at 10, 30, 60, 120, 300, 600, 1200, and 2400 seconds which covers the full spread of efforts from sprints to longer efforts like the final of a one day classic or high mountain climb. For each power file I extracted the best power output from these time durations, calculated watts per kg using weights from procyclingstats.com, and adjusted those relative to average for all riders. I had data from 169 riders with at least 10 power files from 2019 and 2020.

An example of the curves produced for four riders are above. Smith packs a better sprint than the other three riders, but tails-off on efforts outside two minutes. Kamna excels on 10 minute plus efforts. De Gendt is second best at pretty much all points. Declercq is well off the highest outputs between 1-5 minutes, but is close on longer efforts.

29% of the data came from domestiques, 29% from mountain helpers, 16% from sprint trains (so 74% from helpers), 12% from puncheurs, 8% from sprinters, and 6% from climbers (so 26% from leaders).

I looked at both the 80th percentile of power output (so the better performances for a rider) and the median. As you would expect, the 80th percentile data produced wider spread between power outputs versus average.

Sprinters produce over 10% more power than average riders in 10 second efforts - while puncheurs and sprint train riders were both above average here.

Puncheurs were consistently above average at all time periods, while domestiques were consistently below average.

Climbers peaked with about 7% more power than average in 20+ minute efforts, while mountain helpers were about 4% higher than average. At 10 second efforts, climbers were about 13 percentage points behind sprinters, while at 1200 seconds climbers were about 13 percentage points ahead of sprinters.

None of this is earth-shattering information; if anything, this shows the validity of rider clusters based on simply result data because we're seeing the expected power outputs. Classifying riders as members of sprint train or mountain helpers is a valid distinction; they are producing different power outputs over different time durations.

By Stage Types

We can also break-down overall power output in a race based on the type of stage it is. I've simply broken down the races into three types: those ending in a bunch sprint (20+ rider group), non-bunch sprints on hilly parcours, and non-bunch sprints on mountainous parcours. The dividing line between hilly and mountains is roughly Fleche Wallonne or Giro dell'Emilia.

The metric here is relative weighted average power - so power output relative to a rider's own average across all races. In this case, 120% is basically max effort - the efforts required of a winning breakaway rider or top 5 in mountain stage - and 80% is a low effort day like a flat bunch sprint finish in a grand tour. For example, the three big breakaway days for Neilson Powless in the 2020 Tour de France were 114%, 116%, and 114% efforts, while he did 82%, 89%, and 78% on three flatter days in the bunch.

Of course, flatter stages require lower power outputs in general for all riders as discussed in a previous post. But, we can identify some significant differences between the clusters. Climbers are clearly different from other clusters in their mountain/bunch sprint outputs, while mountain helpers are clearly different from sprinters/domestiques/sprint trains.

Climber types have the widest gap between performance by parcours. In bunch sprint races they produce ~92% of their average weighted average power. In mountainous races, they are over 105% of their average weighted average power. Domestiques have the narrowest gap between bunch sprint days and mountain days.

Cluster analysis is regularly used in other team sports to define roles - even in sports with more defined positions. This paper from the Sloan Sports Analytics Conference from 2012 discusses clustering based on roles in the context of the NBA. This talk from Opta Pro Forum in 2015 discusses clustering based on player types in the context of football. There have been many more advanced and refined attempts at clustering in both (and other) sports since. Clustering is most easily done either with the K-means method or with hierarchical clustering. Both operate by feeding certain features for each row of data into the algorithm. For K-means, you have to pre-define the number of clusters you're looking for (this can be optimized so it's not necessarily arbitrary), but for hierarchical a tree is built which steadily divides the data into smaller and smaller clusters.

Clustering in Pro Cycling

K-Means is the method I'll use here. The key to using K-means (and any clustering method) is defining the best features for your data so that there are obvious ways for the algorithm to divide the data. For this, I've defined season long average values for 2017-2020 for four statistics:

1. % of points earned in bunch sprint finishes (of all points earned) - where points are earned decay from 1st place earning the most to a cut-off between 15th and 50th place depending on the strength of the peloton earning the least

2. Overall points per race-day - with the same definition of points

3. % of race-days finishing as #1 rider on your team (must also finish in top 20 in the race)

4. Difficulty of the parcours weighted by points earned - where tougher mountain stages are high difficulty and flat stages are low difficulty

These four features define 1) whether a rider earns points in sprint finishes, 2) whether they are finishing high in races, 3) whether they are leading the team, and 4) whether they fit best on flatter, hillier, or mountainous races. We can generate other features like how often a rider is in the breakaway, their performance in time trials, whether they're successful in tough conditions, or how strong the races they participate in are, but this gives a good start and have strong data availability going back 3+ years.

Performing the Clustering

K-Means can be optimized using several methods (elbow, silhouette, etc) to find the correct number of clusters. Sometimes the number will be obvious and sometimes a small range is appropriate. For this data, between 4 and 7 clusters was the best fit. After fitting the model, six produced the most explainable clusters.

The six clusters produced can be broadly defined as three leader clusters and three helper clusters with the three levels corresponding to mountainous or flatter parcours.

1. Sprinters - the easiest cluster to define; these riders are most successful in bunch sprints in flatter races and are often the leader

2. Climbers - these riders get few points in bunch sprints; rather they earn points in mountainous finishes and are often the leader of the team

3. Puncheurs - these riders are best on hillier parcours and can win from the bunch or in smaller groups

4. Climbing helper - these riders earn fewer points and are leaders less often, but are more often successful in mountainous/hilly stages

5. Sprint train - these riders earn points often in bunch sprints finishes, but are rarely leaders

6. Domestiques - this is the catch-all group for riders who aren't successful in mountain/hilly stages, nor do they earn bunch sprint points often; these can be road captains or super-strong men like Tim Declercq whose work is done before the pointy end of the race.

So about 28% of riders fit into one of the three leader clusters, another 39% in the two specialized helper clusters, and 34% in the more generic domestique cluster. Said more clearly, in an eight man grand tour team you'll normally have two protected riders, three specialized helpers, and three less specialized domestiques.

Visualizing Clusters

This visual lays out how this looks at the team level with colors denoting clusters, % of races as leader on x-axis, and parcours fit on y-axis. Below is Bora Hansgrohe - one of the most successful teams in the World Tour in 2019.

They had three primary sprinters in 2019 who are clustered on the lower right and two climbers in the upper right. They have a number of puncheurs of whom Max Schachmann is the prime example. The clustering isn't perfect here; Formolo is more of a climbing helper and Postlberger is more involved in the sprint train, but because of the mixed roles they get classified here. Muhlberger is certainly a climbing helper though. In the bottom left are numerous support riders of which Schwarzmann, Archbold, Burghardt, and Selig are seen as sprint train and most of the rest are domestiques. You can argue Bodnar and Oss are more likely sprint train than not (and Oss is clustered with sprint train for 2017, 2018, and 2020).

In general though, these plots give a strong overview of which roles riders are fulfilling in a team for a given season.

A generic plot of where all riders fell in 2019 is below.

Applications

This clustering has numerous applications like:

1. does having more sprint train domestiques predict more success for sprinters / same for climbers and their helpers?

2. how does power output differ across clusters on different stage types?

3. which types of riders are most successful on different parcours?

4. which teams are most and least balanced (high or low percentage of riders as leader clusters vs helper clusters)?