Changing Xert's bias

TL;DR: Is there an easy way to modify Xert to “prefer” an individual’s strongest aspect (PP, HIE or TP) when calculating new signatures? Essentially, getting the program to weight one of those three heavier than the other two?

For the most part, Xert seems to track the shape of my progression relatively well. However, I’ve come to notice that I’m more of a “HIE-strong” athlete, rather than having a particularly high PP or TP. This becomes more noticeable when I do a breakthrough effort on one of the 3-4min hills in my area. When Xert recalculates my signature to compensate, it seems to default to bumping my threshold value, rather than my HIE number, resulting in some very optimistic TP numbers (as much as I’d love to have a TP of 428W at 71kg, it’s definitely not right). I know I can go through and lock individual activity signatures, but this seems more like guess-work as to what values to put in and is a bit more fragile a solution, imo. Is there another way to get it to weight HIE heavier, so that my TP doesn’t skyrocket after these VO2-type blowout efforts?

Hi Nate,

There’s really no “biasing” the model. Xert uses your Fitness Signature (MPA) and power data to determine how you’re able to perform. In general, the model already weighs the input of PP, HIE, and TP into efforts through the work allocation ratios. (I.e. doing a PP sprint for a BT will result in an increase to PP, but have little to no effect on TP or PP. Conversely, a long 20min effort BT will have huge influence on TP, but no influence on PP).

I had a look at your data and made a slight adjustment (looks like your PP decayed a little from April, which will may have limited your HIE from increasing based on the loose correlation between PP and HIE). Also, remember that TP in Xert is not representing your 60 min power, but the highest point at which you accumulate and dissipate fatigue at the same rate. Sustaining that effort for extended periods of time will certainly be uncomfortable. Also, look at how closely your TP and training load are related. Seeing your TP at that level is not surprising, it seems that you were “overdue” for this breakthrough. HTH

1 Like

Sorry, but seeing my TP at that level is very surprising, since it’s quite a ways above my lab-derived LT2. Even the resulting LTP that Xert gives is close to that level, not to mention the 60min power the curve gives me…I know for a fact that me doing 415W (almost 6W/kg) for 60min is complete BS. I get that a PP BT affects PP and little else (except opening things up for a HIE increase, like you mentioned) and a long effort BT affects TP and little else, but shouldn’t the same hold true for a 3-4min BT - shouldn’t that affect HIE and little else?

I like the idea of the tool, but my data is littered with inaccuracies caused by these 3-4min BT efforts affecting TP instead of HIE. I’m not arguing that there shouldn’t have been a BT, I’m saying that the tool keeps adjusting the wrong part of my signature for those specific BT efforts. It should be primarily HIE, rather than TP.

Hi Nate,

Looked at things again. Looks like you found a better fit by upping the HIE in manual MPA? The curve looks much more sustainable, especially for a highly anaerobic rider.


Yeah, I’ve been going in and manually tweaking some of the “problem activities”, as I mentioned before. It’s a good-enough solution, but I’d much rather there was a more automatic way to handle those efforts.

After looking at some of your breakthrough activities, I can speculate on why this might be happening. A lot of your breakthroughs are single, hard 4 min efforts, which is right in the region where TP and HIE have overlapping influence on performance. However, if you look at your breakthroughs on July 16/23, these are much more meaningful to the signature extraction algorithm because they capture how you are able to perform after a hard effort…square wave efforts are decent, but Xert really shines with the variable efforts, as HIE will have a big influence on MPA during extended (or repeated) maximal efforts.

If your breakthrough efforts aren’t natural, i.e. your power doesn’t fall as does your MPA which is what normally happens to athletes when they approach and then reach their point of failure, then the algorithm will fit the unnatural data. As Scott says, square wave efforts (flat efforts that end abruptly) aren’t natural and thus may not yield best results. Variable efforts, especially when power continues as MPA declines are much better since they provide more info on how your body is handling the fatigue.

They’re definitely more variable efforts - most of them start at a higher wattage than they end.

I agree with Scott and that’s what I was saying previously. Those efforts are generally attempts at a Strava KOM; relatively easy ride to the segment, blowout effort on the segment (variable, not flat), easy ride back. Fits really well with my training, but like you pointed out, the time period is on the overlap between HIE and TP. So if I do a standalone, 3-4min full-gas effort, is there any way to tell the model to “prefer” HIE increases over TP increases within those overlap periods? Because right now, it seems like when the algorithm finds something in the overlap between HIE and TP, it tends to choose a TP increase for me, which then leads to highly inflated TP values.

An example activity is here:

Before I adjusted it, that ride caused the model to increase my TP to ~425W, while not increasing my HIE. A similar result occurs whenever I do a standalone 1-4min segment attempt, as detailed above.