Adjusted Sea Level Power (aSLP) is our latest development of climbing performance metrics based on ᵉw/kg. This article will explain how aSLP is calculated, what it represents and what limits this stat has. To read our explanation on how we calculate ᵉw/kg please see this article.
aSLP compresses every climbing performance into a single, directly comparable number, normalising for duration and altitude. The calculation process starts with the estimated ᵉw/kg as a base value. Other values necessary for the calculation are the duration of the effort as well as the median altitude of the climb. From this you can already guess that aSLP accounts for the duration as well as the altitude of a performance.
The first step in the calculation is converting ᵉw/kg to Sea Level Power (SLP). Sea Level Power represents what power a rider would be able to do with this performance if the entirety of the effort was at sea level, as it is more difficult to develop power at higher altitudes, due to the lower oxygen concentration. This Sea Level Power is calculated based on this study by Peronnet et al on the effect of altitude on running performances, see graph below:
The theorised effect of altitude in the study strongly correlates with the trends we were able to draw based on our database with over 9,000 climbing performances. Based on this we calculate how much power a rider would have been able to develop at Sea Level. This is what Sea Level Power represents.
What should be mentioned is that the altitude effect, which SLP adjusts for, is obviously for the average rider and individual riders may respond differently to altitude. Miguel Angel Lopez might be able to develop a higher Sea Level Power at higher altitude, because he is more used to this environment than the average rider. That obviously means his performance level is higher at altitude, which is what we are trying to measure here so it is not a problem. The same principle applies in estimating ᵉw/kg for riders with outlier aerodynamic drag, such as Remco Evenepoel, who may be able to develop a higher ᵉw/kg for his actual power on shallower gradients, compared to a rider such as Felix Gall.
Now the first step of the calculation is complete, as we have adjusted the raw ᵉw/kg for altitude. The next step to directly compare performances is to normalise for duration.
In this step we will adjust every performance to a duration of 60 minutes, which will give us a simple number to compare all performances. To be able to adjust the Sea Level Power to 60 minute efforts, we had to create a standard Sea Level power trend-line/power curve, which we created based on our data set, while taking into account the work of Valenzuela et al in his study regarding power profiles of male professional cyclists.
These Sea Level standard trend-lines differ from the standard ᵉw/kg trend-lines, as they have a smaller power drop off for longer durations. This is the case as the ᵉw/kg trend-lines were also created based on existing data and, on average, longer climbs have a higher altitude, which means the higher altitude for longer efforts was (at least partly) indirectly taken into account in those trend-lines. This also led to longer efforts at low altitude being quite overrated in those graphs, while short efforts at high altitude were quite underrated. This issue is obviously fixed with SLP and aSLP.
Below you can see a graph with the top performances at the Tour de France in SLP, compared to the standard SLP Trend-lines.
Based on the mathematical equations for the standard SLP trend-line, every effort is adjusted to the equivalent 60 minute power. This 60 minute Sea Level Power is the final aSLP value. In the last step, the value is simply multipled by 100, so the metric is easily distinguishable from ᵉw/kg and SLP. So an effort of 6,30ᵉw/kg for 60 minutes at sea level is a 630 aSLP performance.
For reference, 660 aSLP (6,60ᵉw/kg at 60min at Sea Level) is equivalent to the pink trend-line, 635 aSLP is the red trend-line, 615 aSLP for the yellow trend-line and 590 aSLP the green trend-line. Below you can see a list with all performances above the pink trend-line/660 aSLP.
If we exclude multiple performances on the same climb (so only one of 'Pogacar Plateau de Beille' and 'Pogacar Plateau de Beille Last 10,55km'), Marco Pantani still leads the list with 5 such performances, ahead of Tadej Pogacar, Jan Ullrich with 3 and Bjarne Riis, Santiago Perez with 2. Ivan Basso, Evgeni Berzin, Alberto Contador, Miguel Indurain, Pavel Tonkov, Jonas Vingegaard, Adam Yates and Alex Zülle have achieved this once.
What this metric offers is a more advanced comparative performance metric than ᵉw/kg. Nonetheless, aSLP is still far from a definite decider as to which performance is more impressive. Factors such as weather, prior stage difficulty and pacing still massively impact the aSLP output on the final climb. Future work, building on aSLP, will be to develop metrics that will also take into account these factors, but for now aSLP is our most advanced stat, which means a large part of rating climbing performances remains a (partly) subjective analysis.
I hope this article clarifies what aSLP is and how it can be used, as we will quote this metric more and more from now on. Any additional questions or feedback is welcome!
Gabriel Stròżyk (@NaichacaCycling)
Nice work. It is fascinating how much better Pogacar and Vingegaard were in 2024 compared to other years.
Yep but this also goes for some other riders like Yates, Landa, Almeida, Gee, Evenepoel etc to be honest.
What’s the margin of error? eW/kg is already an estimate and there are number of different assumptions in your analysis, so aSLP is likely to be even more inaccurate than eW/kg.
Given that these calculations are readily used to infer/suspect doping, providing some error bars would be prudent.
Is it not weird that none of Pogacar or Vingegaard’s performances from previous years are in the list? Did they just just make 1 big jump over all legendary dopers like Pantani & Riis in 1 year or am I reading this wrong? Apart from Adam Yates (which can be explained) this list is full of 20 something year old records that suddenly get blasted away. Seems off. This is not a doping accusation btw, I just dont know what has so drastically changed this year?
Yes of course it is weird and I can’t explain the reason for as I don’t know it either. It’s also not exclusive to the big 3, but rather the whole peloton improved massively in the last 12 months. We can be very certain however that this jump did took place, for whatever reason. Not only because of the calculations, as Jonas and Pogacar both also stated that they were pushing career best numbers by far. Adam Yates even mentioned that he was pushing 480 watts (at ~58kg) while getting dropped on Plateau de Beille.
Let me be a critique of this metric, and point out how the “other” factors mentioned at the end of the article matter:
If we look at the main battleground mountains which were not decided by a late explosive attack (Beille, Isola, Adet, Couillole, maybe even La Turbie), but rather sustained effort, which is exatlcy what SLP should be able to measure, we have a clear and definetive order of the big 3, the same every stage. With 2 exception: Jonas had a bad day on Isola, and Pogi probably held back on Couillole.
However, the metric indicated that Beille Remco had the roughly the same performance as Adet Pogacar. However, we know that roughly the same Remco had a .36 worse SLP on Adet.
So if the SLP has aa at least .3-0.4 range depending on other factors, that means at this point, it can’t really prove that Pogi Beille is better than the “pink trendline performance”, whixh he exactly hit on Adet.
Still, I strongly believe in the future of this type of work,and appreciate your effort!
Yes of course, this is why the metric is only an indicator. I think the pacing wasnt as hard on Adet and the effort was more negative split for the best guys, which is why Pogacar, Vingegaard and Evenepoel achieved less aSLP, while the rest of the top 10 were roughly on the same level as on the next day, as they didnt do such a massive negative split. as they couldnt accelerate to faster speeds than the previous pacing.
100% agree, I think pacing is the biggest factor for these kind of performances. Do you have any idea how that could be incorporated in future single metric? The usual effort from point x is a good start I think, but it will severly underrate the performances due to the tiredness gained during the “slow” part.