Race predictions way off

Sorry, my mistake. The close reference for LT values is NOT an HM, but a 10k. I will update my post above.

Jens said:

Incidentally, is the Stamina metric based on LT too?

I don't think so. The stamina seems to be closely related to the dFRC with some modifications for boundaries and I conclude it is mostly based on the VO2/Pace/Duration relationship that the watch maintains.

By the way, the look of your stamina graphs is a good indicator of whether that relationship is well established. The actual stamina should stay "hanging" to the potential stamina up to threshold levels. Repeated anaerobic intervals should depress the stamina faster than the rest intervals fills it back in. A good anaerobic workout should bring your stamina close to zero at the end.

In my experience, pace is a huge factor in stamina. When I run 400m max speed repeats on a steep hill, the stamina (and the training effect) is largely wrong, whereas 400m max speed repeats on a track look great.

Jens said:

The actual and potential are exactly together, and they both zero together, always before halfway.

Having the *potential* stamina bottom out before the end of a race would indicate a huge bias in the way your watch tracks your pace/duration/VO2.

Since stamina seems to be indeed a cousin of dFRC, the problem would come from a lack of maximal effort recording at various time intervals (for example, 1s, 4s, 15s, 1mn, 4mn, 15mn, 1h - I am using the scale of the power duration model in WKO5 here but the idea is to have maximal effort intervals span the range of neuro muscular, Anaerobic, Aerobic systems). You can do this by substituting a "normal" interval by a "maximal till you drop" effort during some sprint, anaerobic, threshold or aerobic interval repeat.

I had recently some "negative" dFRC and zero stamina on my indoor cycling. I solved this by doing a 25mn maximal steady power workout. That completed my power duration curve on the indoor cycle and my stamina looks normal now. I am guessing this should work as well for running.

Try this: do a new LT test, then try to actually run steady pace for 20mn at 105%  of the LT speed.

Would that be LT pace as per the watch or LT pace as per my 10k race pace, as you mentioned it should be earlier? There would be 35secs+/km difference. If it's the latter, then no, I wouldn't and possibly couldn't do it outside of a race environment. 

I have had the 965 over 6 months and really, I now only use it for GPS and power, which blows out any argument for spending the money on it in the first place, but i don't regret buying it because it's a good looking watch - that's it! I have no idea if power is correct and I don't care. I understand it is proprietary to the watch and brand. It seems plausible and most importantly, consistent. I input those metrics into my own spreadsheet for analysis and that is far more helpful (for me) than anything else the watch cares to tell me. Just about every other feature is rubbish in my opinion. It tells me my hill endurance is low and my hill strength is low. Low compared to whom? The elite or other runners of a similar ability to me on the flat? I recently came 2nd in a hilly 10k. Hills are my strength, but not according to Garmin.

Jens said:

Would that be LT pace as per the watch or LT pace as per my 10k race pace, as you mentioned it should be earlier? There would be 35secs+/km difference. If it's the latter, then no, I wouldn't and possibly couldn't do it outside of a race environment. 

First I wrote a bit too fast. The real world LT test is to be able to run steadily for at least 20mn at 105% of the supposed LT speed. This should be doable but should feel like a maximal effort. The reasoning is this approach is the inverse of the traditional LT field  test (run a 20mn maximal steady pace interval, take 95% of HR, pace or power during these last 20mn to get your LT values). I will correct the previous post, but the idea is to verify that the LT assumed by the watch is correct. If it is not, the usual suspect is the HR Max, or poor data during the test (hills, not using a chest strap for example).

Second, I don't quite understand you. In the sentence above, it seems you are saying you wouldn't be able to run your 10k pace for 20mn? If you run 10k at a given pace, I would expect you could run for 20mn at this pace. Am I missing something? 

Your 10k speed is expected to be very close but slighlty above your LT speed; some people prefer using the 15k distance benchmark, arching back at the 1h conventional duration. The fact is that this will depend on the individual and what % of VO2 Max is their LT performance. But these distances best effort benchmark are good enough to check what the watch is telling you from shorter efforts estimates.

Note that we went into the LT discussion because longer maximal efforts are as important as ultra short and short maximal efforts to solidify the ventilation/distance/pace relationships that the watch is using for stamina, performance condition, race predictions, targets for daily suggested workouts, etc.

Jens said:

I have no idea if power is correct and I don't care

I agree with you. the absolute value of the power is not very important. It is more important that the measure is reliable, repeatable and predictable. I think it is, except in corner cases like very steep hills, technical terrains and short burst accelerations.

Jens said:

It tells me my hill endurance is low and my hill strength is low.

Garmin doesn't publish the details of how they measure that score. Reading the "details" about endurance and strenght,  I am guessing this score is an application of the grade adjusted pace they introduced earlier. They might compare your VO2/pace/duration relationship with your VO2/GAP/duration relationship across the aerobic and anaerobic systems. Complete guessing here.

https://www.garmin.com/en-US/garmin-technology/running-science/running-dynamics/hill-score/

Jens said:

Hills are my strength, but not according to Garmin.

You know, if they are your strength, I am guessing you run hills quite often which could introduce biases in your VO2/pace/duration core relationship. In turn, that could explain why your hill score is disappointing because you compare yourself running hills with yourself running hills....

That would explain also your issues with other metrics, assuming your HR Max is accurate.

I found that hill running distort data quite a bit. As I described above running 10x400m max speed repeats on a 15% slope gave me a tempo TE, while running them on a track gave me the expected anaerobic TE. The stamina graph was looking like a tempo run instead of an anaerobic one. The performance condition was terrible. "Everything" was off. Since Garmin introduced dampening in the VO2 max changes, I didn't see an impact of that workout in the VO2Max estimates, but I saw it during the summer as I ran for a couple of months on hilly trails. So, clearly running a lot on hills introduce a bias. I cannot reconcile this with the introduction of the hill score which tends to encourage hill running.

What a detailed reply. Thank you. I think perhaps is the case that Garmin does make assumptions and go on averages, or look ups. I also think that the averages are more based on men and that women have different strengths and weaknesses, which is obvious really. I am relatively good at hills mainly because I don't have much weight to shift. Like you suggested, I also run a lot on hills. Garmin should see that. However, what Garmin doesn't see, is that a large proportion of my runs in winter are on the treadmill, all of which on incline. Garmin might see HR and time, that would be about it. 

I have also a strong bias towards the longer distances. Garmin doesn't recognise that, although it does give me a fairly decent endurance score. My 5k prediction is not bad, about 20 secs slow, 10k about 1:30 too slow and HM, over 4 mins too slow. This leads me to why I would find it hard in training to run 20 mins at my 10k pace - it's because there is not that much in it between my 5k and 10k paces. I would not be wearing carbons, I wouldn't be well rested and I wouldn't have the race adrenaline.

If it is any relevance, on my previous watch the 745, the predictions were too optimistic. For 5k perhaps 1:30 too fast, 10k around 2mins too fast, but the HM almost correct. Strange, isn't it? 

Hmm, I’ve always know the traditional LT test to be 100% of the last 20 minutes of a 30 minute max effort, not 95%. 

I stand corrected indeed.

If the entire 30mn is maximal steady effort, use 100% (Joe Friel's). If it is only 20mn (Coggan et al), then 95% is the typical adjustment for LTHR, LT power (FTP) and LT Pace (CP). I have seen different protocols, but most often there is a 10-20mn warmup before the 20mn or 30mn all-out interval.

Jens said:

My 5k prediction is not bad, about 20 secs slow, 10k about 1:30 too slow and HM, over 4 mins too slow.

Using Stryd ranges of values as a reference, it looks like the predictions are getting worse when the distance increases in your case. I tend to focus only on 5k and 10k distances and my training history reflects workouts for these distances.

What about your HR Max? How do you know it is the right value for you on the watch?