Comprehensive Climbers Ranking

[VayaVayaVaya]

Did someone use Claude Code to create a poster?

 

[Netserk]

Did someone use Claude Code to create a poster?

good idea to ask here , i thought my response could already trigger a ban ( dear toby has previously been punished for responding to a bot )

 

[proffate]

Padun with the best performance of the year. What a *** legend! He was so good that Bahrain team left him home before the Tour! (apparently he was glowing in the dark)

Drive by shooting

I also love that Froome is like tier 7

 

[VayaVayaVaya]

And Remco ahead of Roglic and Contador and two tiers ahead of Froome, even though we decided in his thread he isn’t a climber.

 

[Red Rick]

And Remco ahead of Roglic and Contador and two tiers ahead of Froome, even though we decided in his thread he isn’t a climber.

Cross era's comparisons. Its like saying Alexander Zverev is a better tennis player than Rod Laver ever was

 

[Peyresourde]

Many here mean it is a joke, but the Remco talk is really getting ridiculous. A lot of riders on my Top 50 rankings and countless others fail all the time. I was barely able to scratch together 10 decent performances for several riders on the list (and that is in a 10+ year career). Riders like Sastre, Ullrich, Schleck, Berzin, Zülle, Leblanc and yes, also Contador etc. often disappointed.

It is the norm in cycling to fail more often than to succeed. Remco has at least one good climbing performance every year since 2022. This is already a lot more than many others. A lot of older riders you might think were good climbers have won almost nothing in their whole career and were 3 times more inconsistent.

 

[Krzysztof_O]

Cross era's comparisons. Its like saying Alexander Zverev is a better tennis player than Rod Laver ever was

Exactly. In Remco's era nutrition is much better than in the 2010s. Only in the 90s food quality was comparable.

 

[Samu Cuenca]

Exactly. In Remco's era nutrition is much better than in the 2010s. Only in the 90s food quality was comparable.

Back in the De Vlaeminck's day they had to eat the forks and knives because food hadn't been invented yet. He would have taken everyone to school if he had been racing today.

 

[CyclistAbi]

The way i see it is Rogla is listed in this metric, that is a plus, as for perceived tiers and beyond. Pogi is listed at the very top and i know that Rogla has beaten Pogi at a GT race, week long stage race and at a monument. And yeah, that is good enough for me.

All in all i am appreciative of the effort invested, that is in the end of much more value then some number.

 

[Netserk]

Many here mean it is a joke, but the Remco talk is really getting ridiculous. A lot of riders on my Top 50 rankings and countless others fail all the time. I was barely able to scratch together 10 decent performances for several riders on the list (and that is in a 10+ year career). Riders like Sastre, Ullrich, Schleck, Berzin, Zülle, Leblanc and yes, also Contador etc. often disappointed.

It is the norm in cycling to fail more often than to succeed. Remco has at least one good climbing performance every year since 2022. This is already a lot more than many others. A lot of older riders you might think were good climbers have won almost nothing in their whole career and were 3 times more inconsistent.

his best 30 minute climb was hautacam, no

 

[Peyresourde]

his best 30 minute climb was hautacam, no

?? Are you talking about Evenepoel last year?

 

[Red Rick]

Many here mean it is a joke, but the Remco talk is really getting ridiculous. A lot of riders on my Top 50 rankings and countless others fail all the time. I was barely able to scratch together 10 decent performances for several riders on the list (and that is in a 10+ year career). Riders like Sastre, Ullrich, Schleck, Berzin, Zülle, Leblanc and yes, also Contador etc. often disappointed.

It is the norm in cycling to fail more often than to succeed. Remco has at least one good climbing performance every year since 2022. This is already a lot more than many others. A lot of older riders you might think were good climbers have won almost nothing in their whole career and were 3 times more inconsistent.

Well, there was a fairly specific reason for riders from a certain era being inconsistent......

Recently, most riders fluctuate in level far less than Evenepoel does.

 

[Netserk]

?? Are you talking about Evenepoel last year?

yep

that or the second dauphine mtf

 

[Peyresourde]

Well, there was a fairly specific reason for riders from a certain era being inconsistent......

Recently, most riders fluctuate in level far less than Evenepoel does.

This explains it for some riders, but it is far from the only reason. I think for Ullrich that was the only consistent part of his prep...
Now we also have Hindley, Ayuso, Kuss who gets 0.5 w/kg boost in GTs, Landa etc. Often, there are easy explanations when riders underperform. For Evenepoel it seems more complicated sometimes. I also find it interesting to discuss the reasons.

As I have mentioned, consistency is worth a lot and so it is natural to mentally downgrade Evenepoel a bit from his place in my ranking. I would still stay that from my Top 50, only 10-15 riders actually climb(ed) consistently good over long stretches of their careers.

yep

that or the second dauphine mtf

If you are strictly talking 30+ minutes, Valmeinier in the Dauphine was the best (77, which is not bad and would be a Top 10 career effort for many decent climbers). Overall, it was Ganda (86). I see no real reason to make a distinction between 20 and 30 minute efforts, it has the same physical requirements.

 

[VayaVayaVaya]

Consistency is king. Roglic > Remco 95% of the time they climb. Maybe Remco has a couple climbs that are arguably better, but if you picked a random climb from their careers, odds are Roglic performed a lot better. Not suggesting your rework your list, which is great and very thoughtful. There’s no way to factor everything in and please everyone.

 

[Pozzovivo]

Long post sorry. I am posting a short summary of my takeaways from the discussion. For my own sake and in case others find it useful.

Your framework, Peyresourde, has struck me as thoughtful and internally consistent. The central ambiguity in these debates struck me as this: we all agree on the raw climbing times, most agree that w/kg can be estimated with reasonably small error margins, yet many of us do not think that these estimates map cleanly onto what is often labelled “performance.”

Estimated w/kg is an output derived from physics-based assumptions and available data. Within a given margin of error, it can be informative. However, it does not directly represent performance. Plotting it on trend lines as if it were a stable indicator across contexts ignores the extent to which outcomes are heavily shaped by exogenous factors that differ from race to race. Even within a single race, conditions are not perfectly equal. The w/kg value therefore requires interpretation and adjustment.

Several analysts attempt to adjust for relevant variables, which is reasonable. Your approach is particularly rigorous in this respect.. The well known Lanterne Rouge altitude adjusted metric also incorporates several corrections, though altitude has somewhat misleadingly become the headline feature. In my view, your index captures a broader and more defensible set of influences. By contrast, the W2W performance index has historically been much less convincing in its adjustments.

Even so, translating an adjusted w/kg value into “performance” remains erronous imo. W/kg, adjusted or not, is a physiological estimate. Performance in elite sport is inherently relational. It is defined by who you compete against and by how perform against them. A meaningful operationalization of performance should therefore incorporate both strength of competition and time gaps.

Take the 2024 Giro stage to Prati di Tivo, for example. If you matched the top ten from that stage against the top ten from Giro d’Abruzzo on the same climb a month earlier, I would expect the Giro group to win convincingly. Yet, they "performced worse". The Giro numbers, imo, reflect the competitive ceiling imposed by that field and race scenario. Making the top ten there represents a higher level of performance in competitive terms, regardless of whether the adjusted w/kg was marginally lower.

For clarity, I am separating three concepts in my head going forward:

1. High w/kg for a given duration. = Not performance
2. High w/kg adjusted for relevant contextual factors. A middle ground but without clear benefits over #3.
3. Performance, defined as the product of competitive strength and relative outcome, that is, who you beat and by how much, possibly weighted by event level.

Much of the disagreement stems from conflating these categories. Some analysts, including one of the LR commentators, regularly blur the distinction between physiological estimates and competitive performance. It's very misleading.

Overall, excellent work, Peyresourde.

 

[Red Rick]

Long post sorry. I am posting a short summary of my takeaways from the discussion. For my own sake and in case others find it useful.

Your framework, Peyresourde, has struck me as thoughtful and internally consistent. The central ambiguity in these debates struck me as this: we all agree on the raw climbing times, most agree that w/kg can be estimated with reasonably small error margins, yet many of us do not think that these estimates map cleanly onto what is often labelled “performance.”

Estimated w/kg is an output derived from physics-based assumptions and available data. Within a given margin of error, it can be informative. However, it does not directly represent performance. Plotting it on trend lines as if it were a stable indicator across contexts ignores the extent to which outcomes are heavily shaped by exogenous factors that differ from race to race. Even within a single race, conditions are not perfectly equal. The w/kg value therefore requires interpretation and adjustment.

Several analysts attempt to adjust for relevant variables, which is reasonable. Your approach is particularly rigorous in this respect.. The well known Lanterne Rouge altitude adjusted metric also incorporates several corrections, though altitude has somewhat misleadingly become the headline feature. In my view, your index captures a broader and more defensible set of influences. By contrast, the W2W performance index has historically been much less convincing in its adjustments.

Even so, translating an adjusted w/kg value into “performance” remains erronous imo. W/kg, adjusted or not, is a physiological estimate. Performance in elite sport is inherently relational. It is defined by who you compete against and by how perform against them. A meaningful operationalization of performance should therefore incorporate both strength of competition and time gaps.

Take the 2024 Giro stage to Prati di Tivo, for example. If you matched the top ten from that stage against the top ten from Giro d’Abruzzo on the same climb a month earlier, I would expect the Giro group to win convincingly. Yet, they "performced worse". The Giro numbers, imo, reflect the competitive ceiling imposed by that field and race scenario. Making the top ten there represents a higher level of performance in competitive terms, regardless of whether the adjusted w/kg was marginally lower.

For clarity, I am separating three concepts in my head going forward:

1. High w/kg for a given duration. = Not performance
2. High w/kg adjusted for relevant contextual factors. A middle ground but without clear benefits over #3.
3. Performance, defined as the product of competitive strength and relative outcome, that is, who you beat and by how much, possibly weighted by event level.

Much of the disagreement stems from conflating these categories. Some analysts, including one of the LR commentators, regularly blur the distinction between physiological estimates and competitive performance. It's very misleading.

Overall, excellent work, Peyresourde.

Good post.

I also want to commend @Peyresourde on the body of work presented, also because I may seem overly critical at times.

I think this is very much a case of all models are wrong, but some models are useful, and this models is absolutely interesting. But as @Pozzovivo said, the index numbers aren't the final indicator of performance, and I try to push back against that a little bit.

As for other commentators or websites, I mainly refer to Watts2win, not because I think it gives the most accurate numbers, but because it's by far the most easily accessible, which is also why I think it's gained quite a bit of popularity quickly. I also think this is why the LR numbers got as much attention as they did, they were just guys with a bigger platform for their numbers estimates and were usually one of the fastest ones to post them.

Ultimately, I think that due to cycling being an outdoors sport it gets so hard to estimate conditions it becomes virtually impossible to adjust for every factor because we're bound to miss many. In addition, I wouldn't know where to begin gathering and scraping all the data to begin with. Plus I don't even think the adjustments are that necessary as long as you're using the index numbers as a starting point for talking about performance rather than a final conclusion.

 

[Netserk]

id rather make the distinction between performance and capacity

performance is what can be directly measured , what a rider actually does in a given situation ( with the full context )

capacity is a latent variable that constraints performance -- to infer the capacity of a rider ( at a given point in time , for various kinds of performances ) , you infer an upper-bound on performances ( more complex than just a power curve )

pdt in the giro was a very limited test on the sustained climbing capacity of the riders and the performance from bottom to top was poor . even the finish was somewhat tactical ( iirc ) where riders didnt perform to the limits of their capacity , yet the performances in the sprint did test an aspect of their capacity

 

[Red Rick]

id rather make the distinction between performance and capacity

performance is what can be directly measured , what a rider actually does in a given situation ( with the full context )

capacity is a latent variable that constraints performance -- to infer the capacity of a rider ( at a given point in time , for various kinds of performances ) , you infer an upper-bound on performances ( more complex than just a power curve )

pdt in the giro was a very limited test on the sustained climbing capacity of the riders and the performance from bottom to top was poor . even the finish was somewhat tactical ( iirc ) where riders didnt perform to the limits of their capacity , yet the performances in the sprint did test an aspect of their capacity

I do love my flat track mountain sprinter bully.

 

[Peyresourde]

id rather make the distinction between performance and capacity

performance is what can be directly measured , what a rider actually does in a given situation ( with the full context )

capacity is a latent variable that constraints performance -- to infer the capacity of a rider ( at a given point in time , for various kinds of performances ) , you infer an upper-bound on performances ( more complex than just a power curve )

pdt in the giro was a very limited test on the sustained climbing capacity of the riders and the performance from bottom to top was poor . even the finish was somewhat tactical ( iirc ) where riders didnt perform to the limits of their capacity , yet the performances in the sprint did test an aspect of their capacity

Capacity is a good concept. You could call my climbers tier list a ranking of their capacity to climb fast. Due to various circumstances, riders often don't/aren't able to use their full theoretical capacity and thus perform below their limit.

I average 10 performances for my ranking to determine the 'repeatable climbing ceiling' of a rider. It has always been the case that this ceiling is not reached more often than it is due to various circumstances (form, tactics, conditions etc.). But it still gives good insight and also some level of historical comparison between eras. The results still need to be interpreted (correctly), of course.

I also deliberately did not give riders beyond the top 10 an exact ranking. 2-3 points of difference (less than 0.1 w/kg) does not mean much. But in general, I think my ranking closely corresponds to observations regular cycling viewers make anyway (E.g. Climbing of Pogacar > Vingegaard > Roglic, Evenepoel, Almeida). In the last 15 years, no rider except for Pogacar and Vingegaard has reached an Index of 95 or higher even once. Vingegaard has multiple in the high 90s, while Pogacar has ten of 100+. Froome and Nibali had performed at a very similar ceiling for multiple years without a single outlier. So my conclusion is that my Index does generally work across different races and years (at least to an extent).


@Pozzovivo thanks for your long reply. We are still a bit on the same issue as page 1. To be honest, it would be great if someone implements your idea of a model based on gaps and competition. I personally do not believe that it would work better than my Index. There are also many examples where rankings like my Index were predictive of future results (E.g. Vuelta 2019 better than Tour 2019, Evenepoel+Vine in Norway 2022, Roglic and Bernal in 2018 Romandie, etc.). In these cases, a model like you suggest may not work as well.

And relying on gaps is in my opinion not that accurate. The W/kg calculations already incorporate gaps (+ additional information) and all my high level performance had big gaps anyways.

 

[Peyresourde]

Even the most 'inconsistent climber in history', Remco Evenpoel, has shown a consistent ceiling in the mid-high 80s of my Index for 4 years in a row (22-25). Maybe part of his problem is that his ceiling is relatively consistent and not improving (much) unlike most other current rider's capacity which seems to be improving year to year.

 

[Red Rick]

Even the most 'inconsistent climber in history', Remco Evenpoel, has shown a consistent ceiling in the mid-high 80s of my Index for 4 years in a row (22-25). Maybe part of his problem is that his ceiling is relatively consistent and not improving (much) unlike most other current rider's capacity which seems to be improving year to year.

The discourse around Evenepoel is usually because he talks about challenging the top 2 while not consistently being the 3rd best climber and often dropping far worse results. It's not really that much about his highest number in a given season, but even then I would argue high 80s does not seem to be enough, especially when it's a sub 25 minute climb in a one day race at under 7% average (I assume it's Lombardia?)

He's basically done 2 really good GTs, but outside of those 2 races and Volta Catalunya 2023 he's really been consistently bang average or worse in stage races. In one day races I think his rouleur ability enables him to fatigue much less than his direct competition on rolling terrain and that's why his index numbers can look so good in these. Plus I think the stage races where he did get great results also happen to be the ones that happened to be really kind to the indexes.

Lastly, I was curious about 2 more stages I didn't see. The 2013 Giro MTT and the 2023 Monte Lussari MTT. Do you happen to have those (and how bad are they?)

 

[mikks]

I also deliberately did not give riders beyond the top 10 an exact ranking. 2-3 points of difference (less than 0.1 w/kg) does not mean much. But in general, I think my ranking closely corresponds to observations regular cycling viewers make anyway (E.g. Climbing of Pogacar > Vingegaard > Roglic, Evenepoel, Almeida). In the last 15 years, no rider except for Pogacar and Vingegaard has reached an Index of 95 or higher even once. Vingegaard has multiple in the high 90s, while Pogacar has ten of 100+. Froome and Nibali had performed at a very similar ceiling for multiple years without a single outlier. So my conclusion is that my Index does generally work across different races and years (at least to an extent).

More or less how I would interpret these results. Useful tool. Interesting. It shows general trends and the broader picture, but the specifics are debatable. In some cases (like Evenepoel index 88.1 and Roglič index 87.7), the difference is so small, basically within the margin of error, that even putting aside the consistency vs inconsistency debate, it would be stupid to interpret that Evenepoel is a better climber than Roglič.


In some other cases, I would say Valverde vs Nibali and Froome — their careers had enough overlap, so we know how they competed against each other and what results they had. At least I’d be a bit skeptical to say, based solely on this index, that Valverde was a better climber than Froome and Nibali. I didn’t see it on the road.

 

[Peyresourde]

More or less how I would interpret these results. Useful tool. Interesting. It shows general trends and the broader picture, but the specifics are debatable. In some cases (like Evenepoel index 88.1 and Roglič index 87.7), the difference is so small, basically within the margin of error, that even putting aside the consistency vs inconsistency debate, it would be stupid to interpret that Evenepoel is a better climber than Roglič.


In some other cases, I would say Valverde vs Nibali and Froome — their careers had enough overlap, so we know how they competed against each other and what results they had. At least I’d be a bit skeptical to say, based solely on this index, that Valverde was a better climber than Froome and Nibali. I didn’t see it on the road.

Valverde's 2 best Indexes are from 2004 and 2005, they do the heavy lifting. I think his ceiling on his very best day was a bit higher than Froome and Nibali, but overall he was definitely worse than Froome in the 2010s and maybe also worse than Nibali (though that is more debatable and more depending on the type of climb/stage).

The discourse around Evenepoel is usually because he talks about challenging the top 2 while not consistently being the 3rd best climber and often dropping far worse results. It's not really that much about his highest number in a given season, but even then I would argue high 80s does not seem to be enough, especially when it's a sub 25 minute climb in a one day race at under 7% average (I assume it's Lombardia?)

He's basically done 2 really good GTs, but outside of those 2 races and Volta Catalunya 2023 he's really been consistently bang average or worse in stage races. In one day races I think his rouleur ability enables him to fatigue much less than his direct competition on rolling terrain and that's why his index numbers can look so good in these. Plus I think the stage races where he did get great results also happen to be the ones that happened to be really kind to the indexes.

Lastly, I was curious about 2 more stages I didn't see. The 2013 Giro MTT and the 2023 Monte Lussari MTT. Do you happen to have those (and how bad are they?)

I probably won't talk about Evenepoel anymore until there is new information. All in all, he has a consistent climbing ceiling that he reaches very inconsistently and is worse on steeper gradients. (Though the last 2 k of Ganda where he dropped everyone are steep).

I do not have the 2013 TT. It is a climb in 2 parts that is difficult to get right. I might calculate it in the future. On Lussari, the performances definitely were not that great, but it is also difficult to be certain with the rough surface and a descent.

Primoz Roglic | 80 (+1): 6.37 W/kg for 28:50 on Lussari ITT (Giro 2023)

The other riders Kuss (76), Thomas and Almeida (74) were not that good.

The (+1) adjustment results mainly from the way I adjust for regularity of climbs and I also gave a bonus point for the chain drop. Otherwise it would be a slight negative adjustment.

(To adjust for irregularities like extended flat sections and descents, I slighty adjust the watts up directly and also add bonus adjustments to the Index, as irregular climbs affect both the watt calculation and the pacing of the rider.)