Both I think. It is more likely that you overestimate your threshold. But even if you don’t, I think your riding style is more likely to rely on your anaerobic capacity, because short bursts over FTP feel easy. In group rides and similar, that means you can often ride above your ability for quite a while, as long as you can draft/recover between bursts.
Last year, you had it right when you wrote: “Think both FRC and FTP come from the same fitting equation, so maybe if your FT is set too low, your FRC might come put too high?”
Since they’re simultaneously determined, from an estimation point of view, if one is biased high the other is biased low.
I poked my nose into the Xert app for their free trial the last two weeks.
It’s quite interesting to have all the different variables automatically mapped out/updated. It did jump wildly around when I tried to get the settings right so might be something you need to not trust wholly.
Anyway, it says my HIE is 13.5kj. that’s lower than I’d thought previously, but based on recent racing and training, it’s probably not crazy to think. I can do multiple hours at 0.8-0.9, but less than an hour with constant efforts over FTP leave me completely cooked.
I dare say it models a diesel quite well, just that not many people are doing super long efforts.
ETA: I do wish TR could do a slightly deeper analytics suite so that I could have just one ecosystem. Xert, like Join, has lots going for it, but doesn’t have the other side of things
Estimation from models, right? I have somewhat high (for my age) short power that falls off steeply after ~15 seconds, and run into the “modeling problem” from time to time. Xert always appears to over-estimate my HIE / W’ (see above: What is your W' bal(/anaerobic capacity) in Kj? Did you work to change it? - #20 by WindWarrior) while WKO usually appears to under-estimate FRC / W’.
Ok, they (HIE, W’, FRC) aren’t the same, but bear with me. One reason is that it is challenging to do all-out short efforts. Mostly mental, I find it really hard to keep pushing beyond 15-sec. Probably part mental, part physical. Whatever.
Right or wrong I’ve accepted what some coaches say - that W’ changes slowly if at all. Here are some estimates from WKO:
Models can be noisy, right? Especially models that aren’t always fed good data 
What isn’t shown are the time spans where I know the model has good data.
So with that coaching assumption, I wave my hands like a good engineer and declare “my FRC is 18kJ” which seems reasonable if you agree it changes slowly if at all.
And then I hardcode WindWarrior’s “eFRC” along with modeled dFRC and use that on graphs for workouts that reduce anaerobic capacity, like last night:
and of course I normally leave dFRC hidden and only show WindWarrior eFRC on that graph.
Yes I know the model is a rough estimate and there isn’t a good algorithm for replenishment of W’/FRC. And a side note: about 2 weeks ago modeled FRC was 16.7kJ, and since then a really hard, maximal type ~15-sec sprint fell out of the 90-day window used to estimate FRC. So FRC dropped as expected.
Here’s what I do: I take the best power by duration over the last six or seven weeks or so, convert watts and duration (in seconds) into Joules, then I plot Joules on the y-axis and seconds on the x-axis. They ought to be a very straight line up to maybe 1800 seconds or so (=30 minutes). Then I cherry-pick the upper envelope of points (i.e., if the there’s an obvious drop-off at one point that’s “below” trend for both shorter and longer durations, it probably means I was slacking off so it wasn’t a true max effort. Then, since the work-duration line is so very straight, I fit a regression line. The intercept is W’, the slope is CP.
Here’s the thing: I can successively drop suspicious points where I may have been dogging it, and see what happens to W’ and CP. If they move around, then I know my CP and W’ are not very well estimated. This is sort of akin to what you were doing.
This also shows why the estimates of CP and W’ are related: they’re simultaneously estimated and if one is biased high, the other is likely to be biased down.
thanks, I’ve read up on CP and am aware of that technique and how that model works.
Do you happen to know if that is a standard chart in Golden Cheetah? I poked around WKO and couldn’t figure out how to do that. This is more for curiosity, I don’t have any really issues using WKO to help estimate short power stuff. Just looking for another tool to confirm/triangulate what I appear to see using other methods.
In GC, if you go to the CP chart and hover in the upper left corner you’ll see “More …” If you click on that, you’ll see “chart settings…” Then the “data series” is a pull-down menu. The default is “power” but if you change it to “work” it’ll show you what I do above.
Another option for “data series” is “Veloclinic plot” That’s a robustness check on the CP and W’. You can google up Veloclinic plot where Mike Puchowitz describes what he was thinking when he came up with it and how it skews left or right when W’ is “off” compared to CP. So for the OP from last year, he could look at the Veloclinic plot as a diagnostic for whether one or the other was way off.
Thanks, appreciate that info. I started using Golden Cheetah and got sidetracked wanting to hack the FIT import code to properly auto-fill the workout names. I’ll go back to trying Golden Cheetah again, its much improved over 4 years ago. FWIW I definitely know my data (where the good max efforts are), and understand your point about cherry-picking the upper envelope of points.
Xert will always have a higher number for HIE since it takes into account efforts that aren’t perfectly paced. If you’re using MMP data that is from regular field data and not test data from maximal efforts to determine your highest power, you’ll find this difference more noticeable.
Also, HIE isn’t a capacity that can be fully depleted. MPA gets in the way of depleting it fully at a given intensity. This same reason is why W’bal models are easy to break. When W’bal is zero, your power doesn’t drop to CP and thus W’bal dips below zero every time.
So caution when you use other models that rely on MMP data. For this reason, most coaches that use them as part of their practice use test data, not field data.
“AI detection” can be pretty reasonably good much of the time, but it can’t always know when you really busted out a max effort. GC actually lets you flag those efforts and then fits on them. That’s nice.
LOL, I prefer a mix of “natural intelligence (NI)” and some models to keep the NI honest. I’m a coach of 1 athlete - myself - so that greatly simplifies the problem.
Seems like you could benefit from some good old heavy squats.
any control over the graph y-axis? I clicked around and swore it used to exist.
and with filter bests enabled:
Had some 15-sec and 1-min and 20-min all-time PRs in that ~100 day window. Its hard to see but the closest points are at 23-sec, 1-min, and 20:20 min. The points at 1:47 and 5:34 are fairly close too.
and the Veloclinic plot that I’ll need to look up later:
Thanks for the quick start tips, really appreciate it!
@RChung thanks I played around with GC late last night looking at some older data where there were enough near maximal efforts to represent “good enough” data for a model. The extended CP model basically gave the same result as WKO’s mFTP, that was a bit of a surprise as so many say that CP is higher than FTP. FWIW it seems I can feel out the border between stable and unstable, just using breathing and heart rate. All of my 2017 and later FTP estimates that I did a spot check on, they all line up with the extended CP model. Some were exactly the same, like when I did a long 50-70 minute effort at FTP. And some like training leading up to a big climbing ride I self assessed at 263W (using WindWarrior’s NI™ power-vs-HR, plus breathing on above threshold efforts) and GC’s Extended CP gave me 265W and WKO gave 261W (for Aug-Nov 2017 time period).
Wasn’t sure about the model’s default search interval for aerobic at 420-1800 seconds (7-30 minutes), and long aerobic 4000-30000 seconds (66 minutes to 8.3 hours). Will go look at source code, check GC site, or drop into the google group.
The Veloclinic plot made perfect sense after reading the article.
Thanks again.
It depends on your event targets.
For a road racer 33-36Kj is extremely high. That’s more what you’d expect to see in a track sprinter. I’m a road sprinter myself and don’t think I’ve ever had mine that high.
If it’s modelled correctly, you could absolutely bring up your aerobic power.
Thanks.
How - on 6-8 hrs/wk?
It’s currently 14.8Kj (last 42 days data) according to Intervals.icu. But not something I consciously work on, as I don’t need a sprint in my events.
It’s a long term change.
You’ll simply need to do a ton of aerobic work. On 6-8hrs you may have to do it right on LT1/VT1 due to the low volume.
The more the better. It won’t happen quickly. We’re talking months, even years to really see substantial gains.
Essentially, nearly all amateur athletes are underpowered aerobically. The result of being time poor and the predisposition to do far too much intensity.
You’re obviously gifted anaerobically. You can afford to let that drop in relation to your aerobic power. Additionally, just a few weeks of anaerobic capacity work close to an event will be ample to ramp it back up.
My power curve for the last year firing the numbers into here ( Critical Power and W’ Explained For Cyclists (inc. Critical Power Calculator) — High North Performance) for 1min, 3mins and 20mins gives me
I have 5k kJ in intervals probably from the last time I bothered to calculate it; I’ve no idea if its an improvement or not or if its even accurate in the first place.
Currently sit a 22.6kJ from intervals and 21.1kJ GC.
2022 was pretty insane, 27kJ and 26kJ respectively.
I guess that is also reflected in power curve, according to intervals in 2022 I was 92nd percentile and 88th for 1m and 5m, but down at 79th for eFTP. w/kg that is, raw power is a different story, I am rather light.