Kiwicoach wrote:
Question is what do you feel the estimate is right compared to?
If you did a maximal 60min effort, what would think your power would be?
In the winter, the threshold estimate of Xert was compared to the FTP/CP models of WKO4 and Golden Cheetah as well as high intensity threshold efforts that were in effect maximal TT efforts between about 30 minutes to 1,5 hours in duration.
In the winter, all of the estimates from the different tools were largely in accordance with each other as well as the actual TT threshold efforts.
As I've changed my training to not include any threshold efforts of longer duration, instead focusing on HIIT with shorter efforts between about 30 seconds and 5 minutes, both WKO4 and Golden Cheetah are now reporting an FTP that's about 40-50 watts lower than in the winter, while my monthly TSS has increased and CTL has gradually moved upwards from a low point of 50 in the winter to a high point of 85 in the summer.
As noted, Xert now estimates that my FTP is 280 which is a little higher than its highest estimate in the winter and almost exactly the same as the highest estimates of WKO4 and Golden Cheetah in the winter.
I could do a longer threshold effort to see where I actually end up at present (about 260-280 depending on freshness and training state, would be my guess), but I suspect that it's highly unlikely that I've lost up to 50 watts at FTP while improving other parameters of fitness as well as wattages at 30 seconds to 5 minutes, as WKO4 and Golden Cheetah would have me believe.
Of course, much of the reported loss is to do with the 90 day window that I've chosen to use in WKO4 and Golden Cheetah, as their estimates shoot right back up to about 260-270, if I choose to use data from all of 2017.
But if I do that, I lose the responsiveness of the model estimate, which you correctly value.
The more likely possibility than both an actual drop in FTP as well as errors in data is that WKO4 and Golden Cheetah can't accurately estimate threshold, if your training regimen does not contain efforts of longer duration near threshold.
It's the model accuracy paradox of training that I outlined in my previous reply: If you already have the wattage data from the longer threshold efforts, the model estimates are more or less accurate but also more or less irrelevant, as you already have the data; while if you have no data from longer threshold efforts, the model estimates could be very relevant, but in this case they're sadly more or less inaccurate, as the models appear to require longer duration threshold efforts to achieve accuracy. This is of course a schematic simplification, but it get's to the root of the problem with the WKO4 and Golden Cheetah models.
I'm guessing the WKO4 model reflects the totality of the specific training and efforts done by the riders used as its statistical foundation, and I would guess as well that all or most of these riders regularly did longer duration threshold efforts, which means that this type of training is what the model is primarily tuned for.
In other words, the WKO4 and Golden Cheetah models can't accurately estimate FTP based only on data from HIIT training at shorter durations.
Xert appears to have no such problem and its threshold estimate is therefore actually useable for me, as well as in line with my winter performance.
If Xert does turn out to also be inaccurate in its estimates based on my HIIT training, it seems reasonable that it's at least much more accurate than the estimates of WKO4 and Golden Cheetah, when they only have my HIIT training to go on.
I also suspect that my circumstances are far from unique, as many others vary their training or tend to avoid FTP testing or longer threshold efforts for whatever reason. I'm no coach (and as an aside, I've actually just cancelled my Xert subscription), but Xert seems great for easily getting an FTP estimate from riders you're coaching who are not regularly doing longer duration threshold efforts for whatever reason, and for whom the estimates of WKO4 and Golden Cheetah are therefore possibly inaccurate.
If the training state FTP or tsFTP from the Xert Garmin app could be incorporated into the Xert web tool to supplement its optimal state FTP or osFTP, that would be ideal, as the tsFTP seems to me to correspond more closely to the estimates of Golden Cheetah and WKO4 than does the osFTP of the web tool, while both Xert models are useable for shorter duration HIIT training, which WKO4 and Golden Cheetah are not (with regards to the model estimates, at least).
It's worth noting that the tsFTP of the Xert Garmin app is also the most responsive threshold estimate of all, as it reflects your FTP on any specific day that you use it, just as an actual FTP test consisting of a longer threshold effort would, without actually requiring you to ride longer duration threshold efforts.
None of the other models come close to this level of responsiveness and ability to reflect your actual daily training state FTP, while simply doing maximal efforts at any duration.
In short, the Xert models appear to have very minor practical constraints compared to the other models, and this has several practical benefits, as it means that Xert is compatible with and accurate for both a wider range of consistent approaches to training as well as larger variations between training types throughout the year.