Friday, September 5, 2014

CTL as predictor of performance

A quick n=2 analysis of CTL vs kilojoules and other metrics for predicting performance in cycling.

Subject 1:

Elite female racer, (my wife) 12 months of data during the 2013-2014 bike racing season. All data collected with a powertap. Any workouts lacking power data have estimated TSS  entered by hand and kilojoules estimated from that TSS. PMC set with a 42 day time constant. For each month in the data set we plot mid month CTL value against the peak minute normalized power achieved in that month (within ~15 days of the CTL value).  45 minute normalized power was chosen as a measure of performance as it is known that the athlete in question would have all out efforts in that duration every month and it is a good proxy for general aerobic power (aka CP, or FTP)  Other performance metrics are shown in a chart of correlation coefficients. Including 1 minute power, 5 minute power, and powerfactor (which is just a weighted average of the other 3).  These same performance metrics were then compared to an exponentially weighted moving average of kilo-joules, using the same 42 day time constant as CTL, I call this ewaKJ.  This amounts to using the same approach as CTL but replacing TSS with kilo-joules

click to enlarge

click to enlarge

For this athlete, the two metrics (CTL and ewaKJ) do about as well as one another at predicting aerobic performance, and 1 and 5 minute power as well.

Subject 2:

Me, a cat 3 male bike racer, during the early 2013 season. Again all data collected by a powertap, same procedure as subject 1:

click to enlarge

This time the results are very different. My training period had a large shift in its makeup, from primarily long steady training rides to lots of racing and crits. ewaKJ was trending down during the season as CTL was trending up, and so was performance.  Being able to handle differences in training variability is exactly what is supposed to make CTL a better indicator of overall training load than simpler metrics. So this is a good sign.

I would love to do more sophisticated analysis like this, with more subjects, but data quality is paramount, as is a deep understand of the athletes training and racing history, so that performance metrics relevant to them can be devised.

No comments:

Post a Comment