Trainer Road FTP Prediction Alternative

TrainerRoad Software
1

How about just get rid of predicted FTP and only give us our current FTP estimate.

It’s impossible to estimate an accurate FTP (it’s an old method to get a guide to set training) and it’s very variable day to day. I’d say there is no such thing as an accurate FTP. Everyone get’s hung up on it. Give us a our current estimate so we can brag to our mates (or not) and see it slowly creep up (it will!).

For the prediction give us a predicted Power curve instead, with bands of likely outcome. For example Most likely + 3 to 5% / Likely 0 - 7% / Unlikely -5 % - +10%. The curve can still update as we progress through workouts but it’s a band rather than a fixed number. You can click on any day in the next month and get your predicted band for that day. The bands would be narrower closer to todays date as there is more certainty. I’d see most likely as if you do all the workouts and pass them all, maybe a couple of back pedals etc. Likely is you skip a couple. Unlikely is you don’t do any workouts or start taking EPO and smashing everything. You would hope that the current predicted FTP would correspond to the likely band starting at ~30mins on the low side and exiting to the unlikely band at around ~75mins.

Added bonus is that, say, you are doing base your FTP won’t change but the power curve at the right side would creep up. You would be able to see the benefits to endurance even though your FTP is static or heaven forbid :scream: going down.

With the current predicted FTP number you can change out a work out and get an FTP increase of 1 watt. Which although that might make you marginally fitter will it translate to 1 watt at ~60mins or 1 watt at 3hrs or 1 watt at 8mins. Does 1 watt matter?

2

That’s not true. Just think of the weather forecast: my weather app includes estimates 7 days into the future. The farther out you go, the larger the uncertainties typically. And even on the very same day, the forecast need not be correct.

Moreover, we don’t know what the uncertainties are, and if they are too big, you basically cannot make any meaningful estimates of your FTP.

Even if the uncertainties are small enough to make the prediction meaningful, I am still quite skeptical that it was a good idea to surface predicted FTP to the riders by default.

When you train models, you need to define what FTP is. Within that definition, you can verify its accuracy. Not knowing any of the secret sauce, I reckon FTP enters as a fit parameter that allows TR’s algorithms to minimize the likelihood that within the next four weeks you will fail workouts. (AI FTP does not use FTP, but FTP is used to translate workouts in TR’s workout library to workouts with absolute power numbers.)