FTP vs. AIFTP- Huge difference

I am on the low volume SS base phase (with extra TR rides added when the time is available). Last week AIFTP increased my FTP to 230W which made sense since the SS workouts were 2- 20 minute intervals at 210W or so and the threshold workouts were higher wattage. After last night’s VO2Max workout send me a congratulations email saying that my FTP had gone up by 4W to 200W. So, the difference between the two is 30W which is 15% of value.

Has anyone else seen these values be so far off from each other? Any idea why? The TR values seem to be pretty close to the power curve that I see on the road.

1 Like

Intervals uses a single max effort. Have you done any max or near-max efforts >3 minutes lately?


No, I haven’t done a max efforts recently. Just what TR tells me to do. Plus some MTB rides as well.

Just like AI FTP has a requirement of 10 indoor rides on the trainer, Intervals has a requirement that you do at least one single max or near-max effort (my rule of thumb is greater than 3 minutes).

When I was using TR, I didn’t have any max or near-max efforts. So if you are the same, I would say you haven’t met the requirement to get a reasonable estimate from Intervals.


That won’t be enough for Intervals. The only way I get a good ftp reading from it is to do a all out effort Zwift race, where it comes back with a realistic value.

1 Like

Sure, but thats next level discussion.

Starting from the top you need a max or near-max effort greater than 3 minutes (I think thats the default). TR doesn’t work that way. Thats the primary difference between the two FTP estimates.


I found that was great for updating ftp when I was doing long threshold workouts outside and inside over 18 minutes. I recently did a max 5 minute test and it gave me a small bump as well, but that was after my ftp had dropped over a few weeks from just chill riding. Bottom line, you need to feed some worthy efforts to get it to recognize and spit out a realistic ftp number.


This 2-minute all-time power best from April was good enough to get within a handful of watts on estimating FTP:


Yes my rule of thumb is 3-minutes, but its not an iron clad rule.

on the other hand this 8-min near-max in May didn’t move the needle:


Algorithms - love 'em and hate 'em! Use your judgement!


Thanks for the input. I haven’t been doing anything special for (intensity, number of rides, etc). I just got the random email one day and was confused. I’ll try following the guidelines above and see what happens.

1 Like

Yeah, I was going to point to this as well. Intervals will tell you why it calculated what it calculated by sharing which effort generated the FTP. As Wind suggests, it’s usually a max effort and TR tends to avoid max efforts.

1 Like always seems to be very optimistic about my FTP. I can produce better 1-5 minute power compared to my actual FTP and thus it over estimates.

@djwalker1260 Just follow one system for measuring FTP so you can be consistent. If you are curious, you could feed the model at Give it the beans on a 5 minute interval and see what you get. (I’m guessing it would be more accurate with multiple efforts along the power curve.)

The above posts about how is not correct. My understanding is that uses a generalized formula from the 95% of 20min max test protocol. It works as low as a 3minute test, tho you can manually change the settings to ignore efforts below a certain time limit, say 10mins for example. You may want to increase the lower threshold if you’re a very anaerobic riders. For example, I have very high 3-5min power, so I can get a huge ftp in w/kg. I could not sustain that pace for 30-40mins.

So in other words, you just need to do one max effort at least 3min in length and it will give you a eFtp. However, if you don’t ever do max efforts that long, it will always be an underestimate by default (because it’s estimating off sub maximal efforts).

This is me recalling what the developer, Tinker, said a while back on this forum.

I need at least 10m to be accurate. 3m efforts overestimate my FTP 10-15 watts

1 Like

For me it was 30W for 5 min power. 3 min is close enough, but apparently my 5min power (when I was training properly) is outlier in PDC (the same for WKO).
Conclusion - if you are not sure and do not know yourself, never trust short power estimation.

:+1: I can believe that individual power curve differences will have an impact on the algorithm. FWIW I haven’t found Intervals eFTP to be that useful, and I prefer WKO or GoldenCheetah.

1 Like

I think that article is saying it is fitting your max effort for that duration to a curve from the set of curves estimated from the fast fitness* dataset. So you don’t need lots of data points, just the effort

1 Like

Alex of FastFitness dataset, he is a coach in the UK (not to be confused with Colorado FasCat). Alex/FastFitness is the guy that screen scraped data from TR users.

1 Like

I would expect different measurement techniques to produce different results. The crux is that FTP is an operationally determined value whose only useful purpose is to benchmark results over time, and to set appropriate difficulty levels for workouts.
If your FTP is X, and doing threshold workouts is hard, VO2Max workouts are very hard, and you can ride tempo all day (all values set as some percentage of X), then your FTP is set correctly.
To put it another way, if you’re using TR, use the TR methodology. If you’d prefer to use some other training program, use the methodology they believe works best with their system.

Yeah agree with the others you need good max efforts, preferably at long duration.

@djwalker1260 My intervals eFTP is also ~30W under my FTP, if I was basing training off that I would be riding sweet spot intervals at low tempo/high endurance power …

You and others eftp on is lower because you have not done maximal efforts to feed the algorithm. It’s not an ‘AI-ftp’ algorithm like TR. you need to actually give it a max effort and it’ll give you a eFtp. It’s assuming your sub-maximal efforts are maximal and that’s why it’s substantially lower than TR

In other words, you actually have to do a max ftp test, you just don’t have to go all the way to 20min like the classic protocol