The section "SRM Data Analysis" says the data is from training and competition. The "Results" section say the data was collected from 1727 trainings and 481 competitions. The clue to it being race data is in the 1st 4 hour value being so low....
I bought the article the other day since my lad was first year junior this year and I was most interested to compare data and see what his tour podium chances will be in six years ;-)
My impression of Pinot's data is it all looks exceedingly plausible ( i.e. non-doped ) since Pinot's second year junior data is not so different to what I'm seeing, and a fifteen percent improvement to U23 is what I guestimated. I'm still struggling with the SRM a bit though, and wonder what data they had to throw out and for what reasons.
This was the first season we collected a full year's worth of data and really started analysing it. The max values we got for intervals of up to one minute were all set in training. i.e. Sprinting up a short incline when the rider is fairly fresh and rested. These are pretty reproducible. Pinot's values here are nothing special, and the spike in 2012 is probably due to him doing a block of sprint training that he doesn't normally do. At the other end of the scale it's no surprise that Pinot's 4hr value improves so much - Juniors don't race 4 hours, so that first 2008 value would be a long training ride. Whatever, it's those 5 to 20min max values which are interesting for Pinot.
Despite "training hard", all the 5 minute to 3 hour max values we collected are race values, which reduces the number of "interesting" datasets considerably. This is also the area where the values get pretty variable. Certainly there are a few wet / cold races where power was maybe 10% down, and there were also a couple or races days which perhaps shouldn't have been race days where the power was down. But on the other side we have 3 files which are 10% up on all the others, and these are the maximal between 10mins and 1hour. Sure, the lad was really on a good day and was riding like his bum was on fire, but was it really that good? You look a file like this and check : was the SRM calibrated - yes - normal calibration errors wouldn't give you 10%. had the SRM just undergone large temperature change (e.g. come out of hot car) - no - Had the chainrings swapped - no.
The paper says they corrected for "artifacts, incorrect calibrations and speed sensor faults". Well certainly speed doesn't affect power readings, but I wonder what constitutes an "artifact". Is there anyone else out there with experience like this? Do athletes sometimes have "float" days.