eFTP is off by 1 watt for 20 minute effort

Basically it says:

FTP +13 to 248w from 20m at 260w

but 260*0.95 gives exactly 247.0 (both Garmin and TP agreed on that new value).

Is this a bug or intervals use different model? It’d be weird because 0.95 for 20-min effort is well established multiplier.

Intervals does use a different model

It is, but it’s often far from the truth…

2 Likes

Fun little article about trying different FTP tests as a little thought provoker.

2 Likes

Well repeated from Hunter Allen’s use.

Some well known coaching companies use a range from 0.90 to 0.95. Diesels at 0.95, those with large anaerobic capacity at 0.9. And others somewhere between the two.

The multiplier is an estimate. Treat it as such. And ftp is a small range, something like 10W for your ftp. Treat it as such.

1 Like

It is, but it’s often far from the truth…

Advanced athletes know that and use their own multipliers based on their aerobic/anaerobic abilities, but this one looks like a off-by-one bug in a system to be honest. Those kind of inaccuracies/bugs make Intervals look to average users not so well design and reliable system.

(For example, the calendar page looks pretty good in English, but very ugly in other languages. I was trying to “sell” it to others, but they said it is unusable because of that. Turned out to be very long word replacements cause this issue. It was easy fix for me, but impossible for an average user to even report it.)

It’s not that FTP tests are wrong. The result depends on athlete’s physiology. The only true FTP test is 30-min MLSS test.

Let’s go back to your original question:
You ask if Intervals uses a different model. Actually Intervals ‘uses’ a model, a scientifically researched and accepted model. A quit arbitrary multiplier of 0.95 is not a model. It’s an averaged, rounded number that returns a value that is valid for the ‘majority’ but can be far off for people with different physiology.
Comparing both results and having only 1 watt difference, is actually a confirmation that, in this case, the FTP test has a high likeliness to be valid.
I don’t get why you would call this a bug and certainly not why you keep pushing this!
Power meters have, best case, +/- 1.5% accuracy, which could mean up to 8W difference for the 260W. Hydration, heat, motivation, etc… will all yield differences that are a multiple of that 1 W.

When it comes to translations, all initial translations were done with an online tool. Several languages have been reviewed by native speaking people and they put in the work to make the languages usable for others speaking the same language.

I would strongly suggest to show some more respect to the people around here that are helping out other users on regular bases. That’s why I flagged your last respons.

9 Likes

@MedTechCD it looks wrong for newbies taught FTP is exactly 0.95 times average power for 20-min TT.

As an engineer I remember thinking the same, and then I learned what FTP really is. So I understand his point, but believe the better answer is education (without getting into a discussion of how to implement). And telling the user how to edit to make it align with any other analytics platform being used.

And that includes explaining why even LAB (not field) tests of MLSS can give different results on different days of the same week. Because there is no true test. Just a bunch of ways to estimate in the LAB or in the FIELD.

Heard this Jack Hagg interview yesterday, somewhere early on he talks about how easy it is to screw up a lactate MLSS test:

sorry I don’t remember exactly where he says it.

FTP by definition is the maximum output over 60 minutes. So any test shorter than that is an approximation.

that is not the definition. The guy (Coggan) that created the term FTP will tell you that over and over and over and over and …

1 Like

“the highest power that a rider can maintain in a quasi-steady state without fatiguing for approximately one hour.”
– Andy Coggan

Is that better qualified? :man_shrugging: My point was about the approx one hour anyway.

A quit arbitrary multiplier of 0.95 is not a model.

WKO also has a model, but they use 0.95 for 20-minute effort and not 0.9538… like in my example.

I don’t get why you would call this a bug and certainly not why you keep pushing this!

Because I see those one offs everywhere on Intervals unfortunately. You may remember on Auuki thread it was discovered that ZWO exporter code was doing back and forth conversion causing Auuki showing lower power targets than it was created in Intervals. Eventually it was accepted as a bug and fixed. If someone exclusively uses Intervals then they don’t know about these differences, but I see that Garmin/Strava/TP/etc. agree on something (power, lap duration, pace, etc.), but Intervals may show lower or higher number, just by one watt/second/etc. These small imperfections may push back some new users from using Intervals because they see different numbers in the system they currently use.

That’s why I flagged your last respons.

That’s fair, but it was not a topic about what FTP is, how to find it, etc. It was created under “Bug Reports”. It was about one off.

With Age, one actually starts to “not care as much”. I’m not here because I’m a world class athlete and I’m chasing every marginal gain and the 1s here or 1w there matters. But I do understand your point. (there has been users who has been miffed about not getting recognized for their eg: 10k run PB because it was off by 1s - but these are perhaps <5% of the users??)

Having said that, you are flagging a 1w difference and that could in all likelihood comes down to rounding. There’s a bit of rounding inconsistencies around these parts :-p

not always

237 x 0.95 = 225.15 but mFTP = 220W

The WKO model maps your MMP to a family of power curves. At least for my physiology, WKO usually is 0.95 of 20-min max power but not always,

Exactly this. With Intervals, you select among the five different modeling approaches. The default one is from Morton’s paper: https://doi.org/10.1080/00140139608964484
and the raw formula is
eFTP = Critical_Power + (Anaerobic_Work_Capacity / (t_exercise - t_exhaustion_at_max_power))
(If anyone has a fully parameterized 3P formula from that paper, please post it.)
The abstract states it’s a hyperbolic curve with “time asymptote significantly less than zero” [??]. Other discussions on this forum imply that the parameters are based on real world data, which of course sites like Intervals would have in great volume.

Just because some other source uses a crude roundoff to 95.00% for 20m doesn’t mean that Morton/Interval’s model happens to intercept the 95% line precisely at 20m. I’ve been approximating with a log fit against ratios I’ve seen come out of Intervals, and fwiw, the 20m value seems to be about 95.46%. I haven’t seen what the values look like between 20m and 60m, but they might be >100%, if Morton isn’t using 60m as his baseline.

To me, all of this is just a proxy for LT2, which is not defined by any particular time period. Athletes vary greatly in the time over which they can maintain power levels just under LT2.

You can see Morton’s equations in this paper:

It’s just curve fitting to get the best parameters for the curve shape. Whether those parameters reflect the real world is another discussion. And whether the points in the MMP are accurate for the Athlete’s ability.

I have a spreadsheet coded that has the model for comparison. I’m trying to make is more user-friendly and I can share it eventually.

I’ve been looking at replicating a 1h effort at FTP and boy it wasn’t an easy task to accomplish. Why wouldn’t that be the best way to determine the FTP when such data is available? Sure, not everyone wants to go all out for 1h, but if such test is done, shouldn’t that value prime over all the others?

Intervals.icu doesn’t use a fixed multiplier like that. It uses a bunch of curves derived from real power data to estimate FTP from a single effort (longer is more accurate). The big advantage of this is that if your effort is less than 20m (you went out too hard or there isn’t a safe place to do 20m in your area or you just felt good and the effort wasn’t planned) you will still get a number. If you get to 20m and have spare gas keep going and you will get a higher number etc..

As I view it, the question for power testing duration is how long I want to be miserable, and to be making sub-optimal use of that misery for training, to measure a thing that is only a sorta-proxy for a reference power level that I actually want to be basing my training on.

What I want to approximately measure and use as a baseline is MLSS/LT2. The TP folks note that trained athletes can sustain MLSS for anywhere from 30-70 minutes. In a perfect world, a single pedal-stroke could be precisely measured and somehow predict MLSS. Then I could measure the somewhat separate question of how long I can maintain LT2. In practice, it seems to require at least a 3 minute effort.

So I’ll go with the 3-20 minute tests and apply the 3P model. Then see how long I can sustain that inferred MLSS.

1 Like

Thanks for the URL!

Yes, please tag me when you have a spreadsheet ready to share, even if not especially user friendly.

1 Like