# New Adaptive AI not always picking the optimal workout **URL:** \ **Category:** Training\ **Tags:** plan-modification\ **Created:** [February 27, 2026, 4:22pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522 "2026-02-27T16:22:53Z")\ **Posts on this page:** 17\ **Page:** 2
**Author:** ![LaFarrapona](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/lafarrapona/32/14704_2.png) [@LaFarrapona](https://www.trainerroad.com/forum/u/LaFarrapona)\ **Post date:** [February 27, 2026, 8:39pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/21 "2026-02-27T20:39:22Z")
Its hard to see how a higher FTP in 28 days is worse long term than a lower FTP in 28 days. ---
**Author:** ![BenB](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/benb/32/7721_2.png) [@BenB](https://www.trainerroad.com/forum/u/BenB)\ **Post date:** [February 27, 2026, 10:03pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/22 "2026-02-27T22:03:25Z")
If I add a VO2 workout in the middle of my recovery week, my ftp prediction goes up. I don’t think anyone would say that’s the best long term plan though. So I believe there’s some logic in the model to recognize short term vs long term gains. ---
**Author:** ![jrfmricher](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/jrfmricher/32/132170_2.png) [@jrfmricher](https://www.trainerroad.com/forum/u/jrfmricher)\ **Post date:** [February 27, 2026, 10:31pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/23 "2026-02-27T22:31:11Z")
Yeah that wouldn’t be ideal. But this is me switching to an easier workout and increasing FTP prediction. So it seems odd it would not have picked that one. Just trying to understand the logic of it all. ---
**Author:** ![NigelTufnel11](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/nigeltufnel11/32/160164_2.png) [@NigelTufnel11](https://www.trainerroad.com/forum/u/NigelTufnel11)\ **Post date:** [February 27, 2026, 10:49pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/24 "2026-02-27T22:49:31Z")
Think of it like cramming for an upcoming exam: You might do better on the exam than studying more deliberately, but the chances that you’ll retain that knowledge are much less than taking the deliberate approach. So, if you’ve got a event in a month, maybe getting the largest short-term benefit would be the best way to go, but if you’re event isn’t for six months, maybe not so much. And, as for having lower FTP at the next prediction, in this particular case we don’t have any background as to whether, or not, the AI thinks this rider is overtrained/overreaching right now, and might not be able to maintain that trajectory. ---
**Author:** ![Niv24](https://avatars.discourse-cdn.com/v4/letter/n/76d3ee/32.png) [@Niv24](https://www.trainerroad.com/forum/u/Niv24)\ **Post date:** [February 27, 2026, 10:52pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/25 "2026-02-27T22:52:56Z")
Its not necessarily easier just because the workout level is lower. My understanding is that the levels were set by the TR coaches estimating relative difficulty. As they are estimates some will be set too high and some too low. It’s not an exact scientific measure. ---
**Author:** ![jrfmricher](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/jrfmricher/32/132170_2.png) [@jrfmricher](https://www.trainerroad.com/forum/u/jrfmricher)\ **Post date:** [February 28, 2026, 12:30am UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/26 "2026-02-28T00:30:41Z")
It might not be easier I agree. But if that’s the better workout for me then why isn’t AI picking it? I want to understand if it’s using other parameters besides predicted ftp to determine what’s best ---
**Author:** ![Niv24](https://avatars.discourse-cdn.com/v4/letter/n/76d3ee/32.png) [@Niv24](https://www.trainerroad.com/forum/u/Niv24)\ **Post date:** [February 28, 2026, 12:45am UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/27 "2026-02-28T00:45:35Z")
Because it’s looking and balancing training loading, fatigue and long term gain. one reason why they provide a slider for how aggressive the training plan is. Some people can handle a fast, aggressive plan, some can’t. ---
**Author:** ![JamesRoad](https://avatars.discourse-cdn.com/v4/letter/j/9fc348/32.png) [@JamesRoad](https://www.trainerroad.com/forum/u/JamesRoad)\ **Post date:** [February 28, 2026, 3:33am UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/28 "2026-02-28T03:33:23Z")
How do we know its not perfectly optimal? the AI is smarter than us so I dont see a reason to question it. ---
**Author:** ![GoLongThenGoHome](https://avatars.discourse-cdn.com/v4/letter/g/439d5e/32.png) [@GoLongThenGoHome](https://www.trainerroad.com/forum/u/GoLongThenGoHome)\ **Post date:** [February 28, 2026, 7:23am UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/29 "2026-02-28T07:23:18Z")
Is your TTE predicted to be higher or lower in four weeks? FTP wattage is just one measure and not the only one for progress. ---
**Author:** ![BothThingsCanBeTrue](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/boththingscanbetrue/32/171866_2.png) [@BothThingsCanBeTrue](https://www.trainerroad.com/forum/u/BothThingsCanBeTrue)\ **Post date:** [February 28, 2026, 11:10am UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/30 "2026-02-28T11:10:35Z")
> [@Niv24](#): > > My understanding is that the levels were set by the TR coaches estimating relative difficulty. I don’t think that is correct. PL were set by whatever LLM/Ai they initially started using. Before when the plans were static they may have been set by a human, but when they rolled out personal PL all of the old static plans/workouts were recalculated. I remember looking back at my static training after the recalculation and the workout PLs were all over the place. ---
**Author:** ![brenth](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/brenth/32/12893_2.png) [@brenth](https://www.trainerroad.com/forum/u/brenth)\ **Post date:** [February 28, 2026, 5:41pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/31 "2026-02-28T17:41:06Z")
Because it was designed by humans. Humans are not perfect hence the multiple iterations from TR ---
**Author:** ![kochx137](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/kochx137/32/7746_2.png) [@kochx137](https://www.trainerroad.com/forum/u/kochx137)\ **Post date:** [February 28, 2026, 7:31pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/33 "2026-02-28T19:31:22Z")
I respectfully disagree. PLs were a step better than the previous workout system and TR’s new workout selection modeling is yet another step better. Continuous, incremental improvement doesn’t imply previous versions were a bust. I was thrilled with my first cell phone, as it was a great improvement over landlines. The advent of smart phones didn’t suddenly make early cell phones a bust. ---
**Author:** ![Niv24](https://avatars.discourse-cdn.com/v4/letter/n/76d3ee/32.png) [@Niv24](https://www.trainerroad.com/forum/u/Niv24)\ **Post date:** [February 28, 2026, 7:36pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/34 "2026-02-28T19:36:06Z")
![Screenshot_20260301_082320_TrainerRoad](https://global.discourse-cdn.com/trainerroad/original/3X/f/d/fd130b0d9307a70902070fc4f8b20d9f8a4dc76e.jpeg) ![Screenshot_20260301_082452_TrainerRoad](https://global.discourse-cdn.com/trainerroad/original/3X/6/d/6d49a8ed25db9b49097ae31258e76aff0065e991.jpeg) The top image is a 4.0 threshold workout I have coming up. The second image is a 3.9 alternative if I ask for a harder workout. Based on my 6 week power curve, the lower 3.9 workout requires a much larger jump in 6 week power records. And therefore the Ai does not recommend I try it due to the risk of pushing me too far and I fail. But, Thats not to say however that the PL score is grossly wrong, it’s just that both workouts produce two different power curves. ---
**Author:** ![jrfmricher](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/jrfmricher/32/132170_2.png) [@jrfmricher](https://www.trainerroad.com/forum/u/jrfmricher)\ **Post date:** [February 28, 2026, 7:44pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/35 "2026-02-28T19:44:05Z")
That is interesting. So maybe the numbers really need to be taken with more of a grain of salt like everyone said. I ended up doing the workout with the lower number since it had more long repeats and it ended up being exactly moderate as it was predicted. It’s also interesting to see how much the answer to the survey changes the upcoming days. Answering easy (which is truly wasn’t but i was just playing around) made the upcoming days much hard and predicted a higher FTP but i think the accumulated load would have not been ideal for me and in the end i would not progress as much. Regardless of the questions i’ve been having, I will say the training plans with the new AI seem very dialed and I’ve made some big jumps this year and feel like I just keep improving! ---
**Author:** ![system](https://global.discourse-cdn.com/trainerroad/original/3X/4/a/4a85e74b36d55fe725a41cbc921dd136eecd6eae.png) [@system](https://www.trainerroad.com/forum/u/system)\ **Post date:** [March 1, 2026, 1:52pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/47 "2026-03-01T13:52:02Z")
This topic is temporarily closed for at least 96 hours due to a large number of community flags. ---
**Author:** ![Caro.Gomez-Villafane](https://sea2.discourse-cdn.com/trainerroad/forum/user_avatar/www.trainerroad.com/caro.gomez-villafane/32/129856_2.png) [@Caro.Gomez-Villafane](https://www.trainerroad.com/forum/u/Caro.Gomez-Villafane)\ **Post date:** [March 3, 2026, 2:18pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/48 "2026-03-03T14:18:52Z")
Hi, @jrfmricher 🙂 Just some insight on how TR picks workotus: _TrainerRoad now looks beyond just Workout Levels, factoring in your fitness, recent training, and overall trends to pick the right workout for you that day._ _When a workout looks different from what you might expect based on past recommendations, it’s reflecting a broader view of your current fitness and capabilities._ _That might mean getting a workout that looks like a bigger step—especially if your recent training shows you’re ready for it.._ ---
**Author:** ![system](https://global.discourse-cdn.com/trainerroad/original/3X/4/a/4a85e74b36d55fe725a41cbc921dd136eecd6eae.png) [@system](https://www.trainerroad.com/forum/u/system)\ **Post date:** [March 5, 2026, 1:52pm UTC](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522/49 "2026-03-05T13:52:16Z")
This topic was automatically opened after 4 days. [Previous page](https://www.trainerroad.com/forum/t/new-adaptive-ai-not-always-picking-the-optimal-workout/111522.md?page=1)