Hi,
Please explain to me why Traineroad doesn’t make optimisation by default…
I just modified an planed Endurance workout to an easier one and my predicted FTP increased 3 watts.
In my opinion it should automatically adjust for optimal outcome - my change to something “easier” shouldn’t be necessary?!
Are you on a training plan? If so, keep in mind that short term goals (FTP prediction) and long term goals (target of the plan) may not necessarily align. The AI may have planned for more recovery than optimal in the short run to hit you with harder work later towards your end goal.
Yes, I’m on a plan - and I get the distinction between short-term prediction and long-term goals. My point is more about the mechanism: if an easier endurance swap immediately improves predicted FTP, that suggests the model is quite sensitive to short-term fatigue/recovery signals.
In that case, I’d expect the system to proactively adjust workload for optimal adaptation without me having to manually downgrade sessions. Otherwise it feels like the user is forced to “game” the plan to get better outcomes, which shouldn’t be necessary in an adaptive system. @martenk
You’re not understanding long term vs short term trade offs. If both the short term and long term benefit from the same training, then what you’re saying is correct. But if the long term benefits from a harder ride today while the short term benefits from an easier ride today, and the plan is optimizing for the long term, then of course tweaking the plan can result in better short term (28 day window) results.
Here’s a clear and obvious example: races. It’s best to come into an A-priority race with low fatigue (openers don’t add much fatigue). So TR will ease up your training load before an A race. However, it doesn’t do that for B/C races because although you will do better in the B/C race, it’s not optimal for doing your best in your A race. To optimize for the A race you have to race fatigued in the B/C races.
The TR interface is only telling you the short term (28-day or less) FTP prediction. You have no idea how the AI is predicting your manual changes to affect your long term performance. So you can’t say it’s making the wrong choice, unless your plan ends within that window.
To maybe give a another example of what was already mentioned by the others.
Say you biked 6 days a week, two hard workouts, 2 hour endurance, and some good volume during a build with a couple days strong 2.0-4hr high zone 2, low tempo. Instead… you cut down the long endurance ride and shortened others… these are not maybe building FTP gains much for an athlete with a long endurance history/base. Much less volume and TSS for a few weeks, BUT… now you wouldn’t be carrying fatigue as much into each hard workout. Would be able to really extend those to higher in range, one more interval, longer intervals… recovery would be much easier with shorter easy recovery rides between rides and endurance rides that were shorter. Short term gains sure, use existing base to its maximum in hard workouts! BUT… long term with low volume and less foundation of volume, could see long term slip in gains and endurance suffer in long events.