22 watt increase since TR AI was released

TrainerRoad Software
1

I thought it would be good to post that the new TR AI does actually work if you stick to the plan. although I also have some gripes.

Below is my FTP History since 2025. (I’m based in Australia so that’s why it looks like I’m peaking in winter for the Northern Hemisphere people.) Looking back on 2025 I saw a decline in FTP because I was not resting enough. My TSS was consistent throughout the year. I still hit all my goals and felt great but was not happy seeing my FTP stay the same or decline.

Enter 2026 the end of the Aussie summer of racing and then expecting my first child in February. I started using the beta on Jan 10th.

You can see a 1 watt decrease. Since then I have stuck with the plan although it was tough to see the AI prediction ebb and flow. You can see after 4 weeks I saw a 2 watt decrease, but then the next detection was a 9 watt increase.

For me it did seem like the system knew I was on the cusp of being overtrained. There was a part of a build where I had the same VO2 workout several weeks in a row even though I was hitting the numbers and rating it hard or moderate but I trusted the system and came out on the other side faster.

327 (4.84 W/kg) is my highest FTP ever at the young age of 37. (I still don’t understand why people don’t consider this a valid FTP within the new system) If I’m still doing % of FTP at 20 mins of sweet spot how is it not valid? They didn’t change the workout structure.

Things I should disclose. I ride 7 days a week. I train roughly 10-13 hours a week. I do my easy endurance days inside, intervals are done outside as well as the longer endurance day.

I no longer tack on easy “miles” to my interval days.

My current prediction is 348. We’ll see what happens in the last week though.

One of the stranger things I saw was that I added Lazy Mountain the day before my last prediction and it was lowering my prediction by 3 watts. Which was 2 watts lower than my current FTP.

I seriously doubt pedaling at 150 watts would lower my ftp that much for the next 4 weeks. I can also prove it cause I then deleted the ride from TR.

Feel free to ask me anything.

The takeaway should be if you stick to the plan the system will work. I understand not everyone can ride the amount of hours I do.

I’ve followed TrainerRoad plans on 6 hours a week and still saw increase and fitness gains.

Screenshot 2026-06-16 at 7.53.43 PM
Screenshot 2026-06-16 at 7.53.43 PM1562×1356 197 KB

2

The simple test is to find analogous workouts with your lower FTP and your new FTP and compare them.

So, if you’re doing 2 × 20 at sweetspot (Eichorn, for example) with an RPE of 7 but you’re now doing them at 250 watts instead of 230 watts, yep - legit increase in FTP. Gotta be honest about that RPE, though. :slightly_smiling_face:

3

Because it’s not.

TR “ftp” is not the same as what Coggan defined it as and what coaches use (maximal metabolic steady state).

None of that is to say you have not improved, you evidently have, just that the ftp that TR uses is not the same as FTP.

4

*except in those cases where they happen to coincide +/- a watt or two

5

Yeah there’s no point looking at a multi-year history of “TRFTP” now. If you want to know if your fitness has improved you have to look at power records over a time period like 20 or 30 mins. Maybe that has improved?

6

I’ll bite. Why would we stick to the plan, if changing the plan around gives a higher predicted FTP?

I found it pretty easy to get the number to go up by making relatively small changes to the plan. Sometimes I can get the prediction to go up a lot by making bigger changes.

Feels like if the prediction works, then it means the plans I’ve been given have been wrong.

7

Because in my experience anytime I mess around with the plan(picking tougher workouts) I’ve hit a wall.

8

I find it hard to find a quick comparison because in Jan 2025 my SS level was 10 with a 323 ftp. It’s currently at 7.6 at 327.

Even looking at VO2 workouts the length of them was different. In 2025 I was doing 2 mins at 360. This past build I was doing 360 for 5 mins.

9

I meant comparing today to Jan 26, when your AIFTP was 305.

I have a 2 × 20 @90% from Monday this week that’s 11 watts higher than a practically identical ride earlier in the year for the same RPE with a marginally lower HR. I’ll take that as evidence of actual progress.

You really don’t have anything similar? The workout levels are largely meaningless. You need to find intervals of similar duration at a very similar %age of FTP and at the same RPE.

10

Well, that’s obviously outstanding progress! :smiling_face_with_sunglasses:

11

Let the AI pick workouts, that’s what it was designed for. But set the plan yourself, because the AI doesn’t even try to optimize plan structure.

For example, my specialty block looked a LOT like my build block. So I changed one of my interval days to vo2, and prediction went up. Changed the type of vo2, and it went up more.

Aligns better with both what the AI predicts as optimal, and traditional coaching recommendations to sharpen top end during specialty phase.

12

To me that’s a bug. If the system is working as intended, it should suggest the best possible schedule for the time you’ve got.

13

Nice work! Keep it up, and just shout if we can help with anything along the way! :+1:

14

That’s not a bug, that’s how the system was designed.

Stock plans are user selected (schedule, frequency, length, category of riding, etc). This populates the plan with a list of ride categories or rest for every day. AI picks the best individual ride it can, based on those constraints. Changing the constraints makes a material change to the final results.

Think about how slowly the AI picks specific rides. Now imagine it trying to do every possible combination of rides for all 28 days.

You can stop reading here if you don’t like numbers. But because I was thinking about it, here’s some interesting details:

The interval category with the largest number or rides is VO2max, with 1,682. Endurance has 427. If we have 3 interval days, and 3 endurance days each week, we end up having to pick 12x1,682 and 12x427 rides. And it picks them one at a time, in sequence. Total search space is 25,308 rides to consider.

If we instead try to optimize the entire 28 day window, you’d have to include every ride category for every day, plus rest. 5,350 different options. For 28 days, that comes out to 5350^28. Which is this monstrosity:

247,688,530,525,344,057,542,676,572,001,549,391,222,888,832,349,679,215,699,806,809,425,354,003,906,250,000,000,000,000,000,000,000,000,000 possible combinations.

No one would try to build that system, it’d never finish running. So you’d have to perform aggressive pruning similar to how chess engines try to pick which paths are worth exploring in detail, and which aren’t. Also means that no matter how good a plan is, its always possible to make a better plan. Never mind that this is only looking at 28 days, which is too short term to make a good long term plan.

15

Update: A 32 watt increase. :upside_down_face:

IMG_5394
IMG_53941178×1060 46.1 KB

16

Also a member of the “All-in for TRAI” club. Just landed a 36W increase since the beta on the day before my A race :slight_smile:

IMG_8674
IMG_8674906×714 37.9 KB

I still get spooked by the numbers. My brain still thinks I’m stuck in the high 200s. Yesterday morning was an easy 45min endurance ride and for the first time my high Z2 was over 240. 240W used to kind of hurt and I still kinda see that number as being challenging until I get into it and realize my body can do it easily. Still surprises me!

Advance apologies for those not seeing results they want, not trying to gloat, but the system really can work if you let it. I have also had my struggles and I’ve had my down months and disappointments but there are principles that really help the system work for me and of course practices that undermine my progress. As long as I stay on the right side of the ledger though TR has been astoundingly effective for me.

Huge kudos to the team and especially Eddie who wades into shark infested waters regularly to try to rescue those who are feeling a bit lost in the open water :+1:t3: :shark:

17

Thanks for the kind words, and congrats on your rising fitness!

I think a lot of us have some form of imposter syndrome, and it’s often surprising to see what we can actually do when we unleash ourselves.

Keep pushing and enjoy the ride! :raising_hands:

19

Congrats man!! Riding Z2 at 240 is a blast :zany_face:

20

What’s your training schedule, approach and plan look like? Those are great results!

I’m just curious how successful athletes are utilizing TR.

21

Nice increase and congratulations.

I too an a believer in letting TR do its thing. I was having some good gains this season but had a minor set back in May.

Things are back on track again and it looks like the fitness is coming back swiftly.

IMG_1541
IMG_15412360×1640 284 KB