Same as Joe - Ramp test (if I ever do them) is generally my brag number, not that anyone actually cares
for my training number I use AIFTP, I would rather be 10% low in training versus too high. It just wipes me out after a few weeks when its too high.
I agree generally, but in order to train effectively you need some number to use. Some people like deferring all training thoughts to TrainerRoad, but I still want to be involved and understand what is happening at some level. This is all a pursuit to improve my performance and unfortunately, the best way to measure that is still FTPā¦
Oh 100%. I think having a number to train off of is absolutely essential. I just think that any test that spits out a number is no better than AI. Itās all just an estimate is my point. Keep smashinā!
Then do a Kolie Moore style test. Or, do a 2x20 at your current FTP with a short rest. Results will be similar.
I feel like people forget that FTP is not an all out effort. Itās a proxy for oneās maximal steady state. That is an RPE of like 6-8 (maybe starting at 6 and ending at 8). A ramp test allows a certain phenotype to game the test. Even a 20 minute effort can be gamed with the last 3 or 4 minutes all out.
I had a similar situation about half a year ago, at similar power levels. Training was followed to the T, ramping up volume and AI gains slowed and then while I felt absolutely the strongest I had ever been, it gave me a -1W. The loss (while insignificant) was totally unexpected. I did a ramp test and blew the AI number out of the water. I wasnt doing any VO2 blocks and Im not terrible gifted in that area, so I just took the ramptest number. In the past AI and ramp had always produced numbers in the same ball park, but at that time they had significantly diverged. I was able to complete training fine with the ramp test FTP and this started waking up the AI as well. Trust your gut on this one.
For the record, this occurred in Feb when I went from 274 to 273, by AI FTP. I ramp tested somewhere around 286-7, forgot the exact number. Many similar comment in the thread, about how ramp tests are not valid etc, and my fitness had just plateaud etc etc. Waking up the system was absolutely needed, and now here several months later, im happily riding with an FTP of 307 still progressing steadily. Im glad i questioned that data point at that time.
Thanks for the info! Could you be a bit more specific about what you did? It is unclear to me if you kept the ramp test value or went back with the AI FTP value, and then the AI prediction started agreeing with it?
I used the ramp test FTP, then closely monitored whether my power zones were still correct based on RPE and heart rate. The Saturday 2h (for me) threshold session is also a good benchmark in my opinion to see if FTP is correct. If you can complete PL 4+ workouts you are likely in the right ball park.
FWIW, I donāt think FTP numbers are that accurate. FTP is a range that can fluctuate day to day. A 12 watt difference between AI and ramp test is probably within the margin of error.
I donāt think it is insignificant and day to day variability, and Iām certain it wasnt in this case. It was a ramp test after a 3 week training block with no recovery. The real problem was not the number, its the trending of progression (AI FTP) starting to diverge from what we were experiencing/guestimating. Anyway not going down the AI FTP rabbithole agian. Im glad I questioned it, and iām pretty certain I wouldnt be where i was today had I not.
Well for me the difference was 26 which is an 8.8% increase. I think thatās bigger than a margin of error.
Yes, I was talking about the 12 watts which is around 4% in the above case.
If you want to know for sure, just take what you think your real FTP to be and then do a 2x20 workout or a KM test and see how it feels.
I completed an FTP detection about 3-4 days ago and got a result of 274. I then did a ramp test and got a 296. I accepted the new value of 296. When Looking at the AI FTP detection screen, I realized it is actually inaccurately stating the 296 is from AI FTP detection. It should be pulling the last AI FTP detection rather than currently set FTP.
Iām pretty sure it also says this even if you add a manual ftp number on your ftp history page.
Not sure why this got merged into this post. It was put in separately to notify TR of this bug and help them improve their system. It is a valid bug, and should be fixed by the TR team.
For sure, but the argument doesnāt hold up as well for bigger differences, which makes it not very good.
The solution is to always do a long form test. Then you know for sure.
Your screen shot shows a detected FTP of 298?
Yes, but itās a bug. My last detection was 274. The 298 is from the ramp test.
If anyone is curious. Just now got an AI FTP detection of 297. Hereās the overall timeline:
| Date | FTP | Source |
|---|---|---|
| Sunday, August 3, 2025 | 298 | Ramp Test |
| Thursday, August 7, 2025 | 280 | Manual Entry |
| Thursday, August 14, 2025 | 281 | AI FTP Detection |
| Friday, September 12, 2025 | 285 | AI FTP Detection |
| Tuesday, October 7, 2025 | 291 | FTP Adjustment |
| Saturday, October 11, 2025 | 297 | AI FTP Detection |
Dropping my FTPback to 280 was probably the right move. Iām a little surprised that I got an FTP adjustment just 4 days ago, and then when my AI FTP detection renewed it gave me another 2% increaseā¦ Intervals are going to hurt for the next while.
I have always been able to beat my ai ftp number by a few watts but using the ramp I burn out in flames after a few weeks. Ai is just right. Your numbers look similar so Iād say just use the ai number and enjoy good results.
Joe
