merlinextraligh

I did a moderately hard workout today, and the Garmin decided to raise my FTP 110 watts to 350.
I know this is wrong. It would be 10 watts short of my all time best, set when I was racing a full calendar and training hard.

I only recently resumed training with power, and am coming back from surgery and a layoff, so I know my actual power is well below that. I'm assuming that the limited amount of data, and some pretty decent short maximal efforts today are throwing off the auto calculation.

I don't rely on the auto calculation feature so its more a curiosity.

Do others use the auto calculation feature, find it to be at all accurate compared to field tests? and upon what is the algorithm based ? or is that information proprietary to Garmin?

Seattle Forrest

It's pretty useful to me, but I know it isn't god's honest truth, it's more like an updated ballpark. Sometimes I agree with the number, sometimes I don't, depending on a lot of things like how fatigued or fresh I was when I was riding.

There are two things that will trigger this feature, according to Garmin's documentation:

1. If you set a new 20 minute MMP it will assume your FTP is 95% of that value. That's not perfectly accurate for all people, but it's generally accepted as a way to guesstimate.
2. It uses HRV to determine when you're at your threshold and looks for a consistent power output there.

WhyFi

I would suspect a problem with your power meter, not with the calculations. I'd look at the file and see if something is amiss, like showing averages of 500w+ for 3 min or something.

The calculations are in the right ballpark for me.

merlinextraligh

The power meter has been recently calibrated, and is consistent with another power meter I have, perceived effort, and other recent efforts. The results in the file don't look unusual and the maximum power for 5 seconds all the way 90 minutes don't look unusual.

A number of the power readings were all time bests. But that's in context of only about 6 weeks data.
Before today, a lot of the work I've been doing has been sub threshold, tempo (sweet spot) training.

Today had some maximal efforts, 20 seconds on 20 seconds off.

I'm guessing the previous auto calculation was low because the algorithm under weighted the tempo work, and today's is high because it over weighted the short max effort intervals.

merlinextraligh

Half of today's workout was 20 seconds on 20 seconds off maximal efforts. The other half was 3x20 strength training, done in a high gear, at a low cadence, around 50 rpm. My 20 minute power on those was pretty good, and by their nature your HR tends to stay lower than normal steady state intervals.

So based on the above, it seems that perhaps decent power at well below my threshold HR caused the Garmin to skew high.

Also full gas 20 seconds on and 20 seconds off for close to 20 minutes ends up producing a pretty high 20 minute power, but likely not the indicator of FTP, that the 95% of a 20 minute field test is designed to be.


Oh, and I did add one short maximal effort when I just had to catch and drop two punk kids on e bikes.

WhyFi

Hmmm. What kind of numbers are you seeing at 5min and 20min? I've never seen FTP calculated from less than a 3min average, not from Garmin or Intervals.icu, and that'd take some hellacious 20sec Intervals.

Seattle Forrest

https://www.firstbeatanalytics.com/e...old-power-ftp/

Sounds like they also consider your VO2max, of you haven't been training with part for a while that value may be off and the calculation may become more accurate when you have more (recent) data.

WhyFi

Ah, maybe it's not doing that for me because I'm using an arm strap, rather than a chest strap, which isn't HRV compatible.

merlinextraligh

best 5 minutes was 310. Best 10 290. Best 20 minutes was 250.

Ironically, the best 20 was not during a 20 minute interval, but 15 minutes of 20 seconds on 20 seconds off, and the subsequent 5 minutes recovery. ( the20 minute intervals were prescribed to stay within a range and not exceed a top end; the 20 second efforts were close to full gas.)

Only explanation I can think of is 20 minute power in the mid 200’s at a HR we’ll below threshold gets extrapolated upward.


Applying the 92% of 8 minutes or 95% of 20 minutes formula doesn’t get you close to 350.

merlinextraligh

My take from all this is that the FTP calculator with limited data, particularly with a workout that is targeted to work specific areas, as opposed to just going out and riding at your limit, is subject to pretty wide variations.

Perhaps, over time, with more data points, it comes closer to what field tests would show.

Bah Humbug

The FB calculations aren't based on 20' efforts alone (though it'll use that if you have one); they also work on power produced at steady-state threshold HR. In my experience it's quite accurate (at least, in the sense of agreeing with my other tests), but even with a calibrated PM you could get that outcome from a bad HR strap reading low.

I really think your situation is due to a GIGO effect, not finding a flaw that big in the algorithms themselves. They're not perfect, but that's at least a full order of magnitude bigger than any error I've seen from it before.

WhyFi

Wow, that's a pretty crazy estimate with these figures. One nice thing about Interval.icu is that they actually note which effort resulted in a bump to your FTP estimate. It would be nice if there were more clarity on Garmin's side for instances like this. I mean, they do make you choose whether to update or whether to skip any FTP change, so it's not going to wildly mismanage the tracking of your training load without noticing, but still.

Seattle Forrest

Is your max heart rate set correctly in Garmin?

Seattle Forrest

​​​​​​That's really cool!!

GhostRider62

It is easily impressed. It liked to flatter me by spitting out a VO2 max of 73 ml/kg. Not quite possible for someone into their 64th year. If it just lied a little, maybe saying 50 ml/kg, I'd get a brief rise out of it.

merlinextraligh

So today’s ride was just 2 hours of endurance at a moderate effort. With a normalized power of 240 for 2 hours at an average HR 22 beats below my lactate threshold, Garmin now tells me that my new FTP is 250, which seems a bit low for that power at that intensity level.

I think the truth is somewhere between the 2 days. It also seems that you need a fair number of repeated, sustained hard efforts to get the estimate very accurate.

To me though, it seems mostly a curiosity, given that you can do a field test to actually set your training zones. At best, it would appear to be a tool to tell you it’s time to retest if the number is going significantly up or down

burnthesheep

The best indicator of performance is performance itself. All I can say. If you care about a power at a duration, go try it. Shrug.

Troul 

only if that was true for calories being burned in general.

<Looks at a cookie, gains 4 lbs & type 2 ...

Seattle Forrest

​​​​​​FTP tests give you the best data (once you learn how to pace one) but they're hard and unpleasant, and take time to recover from. The AI FTP feature is (obviously and by necessity) less accurate than actually taking a test, but it doesn't wipe you out the same way either.

I take it as an opinion. I don't want to go take a new test every time my intervals start feeling too hard to easy. If I don't agree with the number it gives me, I ignore it and go on with my life. If I'm already kind of thinking the # I've been using is out of date and the Garmin says the same thing, I take it as I'm right and go from there.

GhostRider62

Of course, he is correct.

However, people use these data points to establish training plans. There is a 40 watt difference in my estimated FTP between Garmin and other AI packages. Rather than subject myself to 60 minutes of torture to invalidate one or the other, another alternative is to measure lactate. But of me, I do not trust the Garmin numbers.