- Stage 17: 4h 18' 50" (actual), 4h 32' 07" (prediction), 13' 17" slow (5.13% error)
As with Stage 16, I was slow on this stage. I am obviously much happier with a 5% error than yesterday's 16% error! Stage 16 is still a mystery to me. That stage was mostly uphill in the rain, and the winner's average speed beat every other stage winner's average speed (except that in the team time trial of Stage 2).
The Norwegian Edvald Boasson Hagen had the following average speed today:
- Stage 17: 11.53 m/s (25.6 mph)
Despite the near 50-km downhill near the end of today's stage, Hagen's average speed was nearly 10% less than Hushovd's Stage 16 average speed. Again, what happened on Stage 16?!? I welcome your thoughts in the comments.
Stage 18 has three monster climbs and two great downhill segments. By the time riders reach the end at Galibier / Serre-Chevalier, they will have gained 2.29 km (1.42 miles) of elevation from their starting point. Here is my Stage 18 prediction:
- Stage 18: 5h 51' 23" (prediction)
Can Thomas Voeckler hold the yellow jersey after tomorrow? It should be a wonderful climb to the finish line!
I had a look at your model. You point out the model "is most sensitive to the cyclist’s power input" but the power input estimates seem higher than the numbers I'm seeing at http://media.sbs.com.au/cyclingcentral/tourtracker/ . The tour tracker could be improved but some riders have live power data which becomes available when they get close enough to the front. Early tonight Irizar was coming up mostly a bit below 300W. The highest I saw was 474W.
ReplyDeleteGreat to see your model.
ReplyDeleteAt 3.2km to go Andy Zeits came up with 253W, again lower than your estimate. Not sure where he finished though as the results aren't up yet. You said you can't get good enough profile data as "The profile data available on the Tour de France website is, unfortunately, not as good as ..." but if you use Google Earth you can get any mesh density you want.
The stage 17 time of 6h 7m 56s is +5% on your prediction.
Ken: Great comments! Regarding power input, my model uses published research done on cyclists in laboratory settings. Each rider is obviously different; my model seeks to find the winning stage time, which means using the top power outputs. Drafting, motion in a peloton, and other factors alter the drag on a rider. The rider's orientation for uphill and downhill biking also affect drag. What my model does is use published power results to estimate what the winner will output on average over a given segment of the race. It is true that you will find a wide range of power estimates. In short bursts, cyclists are capable are rather large power outputs, even upwards of 500W (see "High-Tech Cycling," edited by Edmund R. Burke).
ReplyDeleteAs for the mesh I use, I am employing Google Earth to help me add a few extra points. There is a balance between too fine a mesh and too coarse a mesh. I certainly don't want to know about every cobblestone and stick a rider may pass over. But, I need to know important points where the road slope changes.
Thanks again for the great comments!