Learning from Ferrari
Matteo Ferrari, the (in)famous bicycling coach, uses a very interesting system to compare climbs in the Tour de France across years and across routes. He’s measuring “Relative Power” through the use of a metric called “VAM” (Vertical Metres Ascended, per hour).
On the Science of Sport blog I first read of this and decided to put his formula to the test against 6 uphills that I had the following information about:
2. Time to summit
3. Elevation gained
With these few details I could calculate both my “VAM” and my “Relative Power” on these climbs in Watts per kilo of bodyweight. The higher the wattage and the more metres climbed per hour, the better the climbing performance can be considered. However, the “Relative Power” is the most reliable because it takes gradient into account. Neither can take into account weather, wind and terrain but on an uphill and in running these are reasonably minor factors. For my selection of races there were no extraordinary weather conditions that would overly affect the result.
Note Relative power is measured as Watts produced per kilo of bodyweight.
VAM is measured as vertical metres climbed per hour (here topping at 1021m/hour at Snowdon). As expected the graphs are similar because the ascent grades are similar on all routes ranging from 12.5% (Djouce) to 16.1% (Clohernagh).
How to read the graphs
If the two graphs look similar it’s merely because the ascent grades of your routes are also fairly similar.
The graph is actually extraordinary in the sense that is shows exactly what I would suspect to see: A linear progression upwards. At the Powerscourt Uphill race I could only produce 4.31 W/Kg while at Snowdon, the pinnacle of my season, I was at 5.07 W/kg. This is a difference of 12% which translates to a difference of 5:13 Had I thus been in the peak shape I had a Snowdon earlier in the season, that is the level I could have expected to run at (roughly).
To look more at the trend is very encouraging for my peaking strategy, from Djouce to Clohernagh to Lugnacoille to Snowdon there was a steady progression towards a peak performance. No doubt if I take a measurement at Mt. Leinster I will see a higher ratio than at Djouce and Clohernagh but lower than Snowdon as my form now starts to ebb off for a while (and my training too, of course, and these two are interlinked). From Djouce to Clohernagh I improved by 0.11 W/kg and then a further 0.31 W/kg from Clohernagh to Lugnacoille. I then had a final 0.16W/kg for Snowdon breaking 5 W/kg for the first time.
What about good old times or good old pace? Why not just use these as indicator. You can do this if you are running up and down exactly the same slope (such as comparing the ascent time from Lug 07 to Lug 09), however, on the hills even the exact route is rarely the exact route, so the VAM is more correct.
In general pace tells you nothing across races in the hills as the below graph illustrates.For the pace chart to show the improvement you would expect it should slow a steadily lowering graph (as times get faster, e.g. lower) but instead it shows the best average pace at Djouce, very unlikely to be my best performance, so we can safely discard the use of pace as a measurement of uphill performance. (also, of course, like in all other distances of running, the shorter the climb, the quicker the average pace will be and, incidentally, Djouce is shorter than Snowdon if longer than Lug and Clohernagh).
Historical records are even more enlightening. Lugnacoille 2007 (an uphill only race) allowed me to produce 4.31 W/kg compared to the 4.89 W/kg two years later (2008 can largely be discounted as I struggled with injury. My theory is that we would see my best ever outputs, better than the present, during March 2008). My peak performance of the hill-running season 2007 came at Snowdon and it was a 4.60 W/kg. This shows how peaking was consistent throughout the years.
More interestingly, it took me until Lugnacoille this year to again approximate the climbing ability shown at Snowdon 2007. From this race I came into a very strong marathon performance and then had my glorious Winter-Spring 2008. So the figures clearly show that only in the recent weeks have I managed to return to the climbing power I had achieved before my injury.
Relative Power as a Tool
I think this measurement is potentially revolutionising for comparing hill running performances across time and courses and I will make major use of it in the future. The caveat to this it doesn’t really work for downhills (there’s no climb) and the terrain plays a much greater factor. Your only real option here is to go by your actual pace and then make notes on the difficulty of the terrain.
So could one explanation for my improvement not be a lowering of body weight. Certainly if my overall weight is lower I can produce more energy per kilo (potentially). This could be true but the opposite seems to be the case. In 2007 I was weighing in around 70 kilos and while I had neared towards 68 kilos early in the Leinster League season, I bizarrely increased to 70.2 kilos shortly before Snowdon. Despite this increase in body-weight I managed to produce more Watts of energy per kilo.
But perhaps this was muscle mass enabling greater energy output? Again, this flies against the facts as my body-fat early in the season was down around 9% while it had upped to 9.5% towards Snowdon (probably through too much eating during the taper). However, it does explain some of the improvement from 2007 when my body-fat seemed to be around 10-12%. On photos from Snowdon 2007 I look very emaciated, though, so my body-fat percentage may have been lower then.
So the increase in performance comes either simply from increased endurance, strength and fitness or from a successful taper (its worth noting I did not have a taper for any other races on the list except Snowdon 07). My taper and mileage for Snowdon 07 and 09 are almost identical, so comparing these two may be the best gauge of improvement from 2007 to 2008 showing an increase in power output of 9%.
If I had weighed 69 kilos instead of 70 kilos I would only have needed to produce 350 Watts to reach the summit of Snowdon in 58:45 instead of 355 Watts. This is a difference of 1.4% or equal to 49 seconds worth of extra energy I would have had available. If I could return to my weight of 65 kilos with the same power output that alone (with no additional training) would yield an amazing 4 minutes and 7 seconds for a summit time of 54:38. If ever there was motivation for leaving that chocolate mousse in the freezer here it is!
While the VAM metric doesn’t account for gradient it is nevertheless interesting reading. For instance my best climbing speed per hour was Snowdon this year with 1021 vertical m/hour. At Djouce my climbing speed was a mere 899 vertical m/hour.
The average gradient on Djouce was 12.5% against the 13.4% on Snowdon. Add to this that I had to climb for 58min45sec on Snowdon and only 43min28sec on Djouce (mainly because Snowdon was a 7.6k climb against Djouce’s 6.5k), it’s easy to see that Snowdon constitutes the better performance and why the Relative Power figures above seem to give a representative view of performance.
This exercise is useful to predict future performances on routes. Since training is generally progressive, it’s interesting to calculate how quickly I could run other races if I was able to generate the Snowdon 09 output on each of the courses. The reason it makes sense to use this “peak performance” for prediction is that my future peak performances should be higher and that we know, for certain, that the power output from Snowdon is physically achievable for me (e.g. within the realm of the possible).
These figures seem to support that if I hadn’t been injured I could have continued my improvement from the top-20 and been vying for spots between 7 to 17 instead of regressing on last year’s performances. Snowdon now seems to suggest that I’m back on track. This should work as laurels to rest on despite this and it’s critical to do focused work both on uphill and downhill to ensure I start next season in the same shape as I was as Snowdon or better.
A good few runners, even those who train sub-optimally, will also improve, simply from the extra mileage they’ll have done, so the goal of competition won’t be to just keep this upward trend going but to create a swifter upward trend than the people ahead of me (incidentally, this is easier than you think, because the quicker you get the smaller your returns for extra get, meaning slower people who are increasing their training are always catching up on quicker people increasing their training by the same amount or less, in relative terms).
I’m happy with this analysis and will use it for specialised Uphill sessions as another tool for gauging performance. With this I can construct a 2-3km incline that run at the same intensity and then analysed through this metric will give me the same insights as a 1 mile time trial on the track (and more accurate for hill running since not all of the 1 mile speed transfers directly to the hills).