Significant data variance

I have been digging into J. Friels books and have more recently started tracking ATL, CTL, and Training balance along with TSS which I have been doing for quite some time. I started out using Training peaks free trial and then decided to give a shot thanks to suggestions from this forum. Neither program would upload all my workouts for the past 90 days so I had to do a good bit of drag and drop to get them all in correctly. I checked and double checked that they were all uploaded and then compared the two thinking they used the same software. Apparently this is not the case as TP has approx + or - 15% higher and lower peaks and valleys on almost all data. All input is identical and overlaying the visual charts is also identical but the values? Even though Training peaks requires a paid subscription where as doesn’t I am leaning toward TP due to long term reputation. Anyone else notice this? 15% is a significant difference when working towards increasing and/or maintaining a relatively high CTL.

I don’t know what would account for the discrepancies, and I can only assume that the math is a bit different here and there. Strava’s Summit, Wattsboard,, TP, my own spreadsheet…are all close-ish, but never the same.

To be honest it doesn’t matter. This is kinda the same as the power meter accuracy discussions. Pick one, whichever you like best, and then stick with it while being mindful of your rate of change in CTL, instead of the actual number. Ramping up and ramping down are more important markers, particularly when paired with how you’re feeling.

3 Likes will pull everything you have on Strava if you look back far enough. It might take a little while.

If your FTP is the same (including on old activities) then the numbers for activities with power should be just about identical. Training load in is calculated the same as TSS. If you have activities with without power but with HR data then will estimate training load from the HR data. These might be different to TP. uses a exponentially weighted moving average for fitness (42 days, CTL) and fatigue (7 days, ATL) which is the “standard” way to do it. I did read somewhere that TP uses a different factor for ATL (not exponentially weighted). That would cause ATL to drop off differently. I haven’t verified if thats true however (don’t have paid TP account). has some nice features lacking from TP and other tools e.g. the age group power rankings, interval detection, annotations on the fitness chart and lots of features for “peer” coaching etc…


I gave it 2 days and lost patience. Training peaks wasn’t as bad as Strava but both only captured 60=75% going back to Jan. As long as it keeps current it’s done at this point. Thanks for the input.

Yes…that is what I thought also but not so. All data identical. Some workouts(primarily inside) with heartrate and some without but same data with identcal charts.