Why Garmin Connect so optimistic on race prediction?

In my case, the prediction is rather accurate: 5K - 8 sec difference, 10K - 2 sec difference, ½Marathon - 58 sec difference

The prediction projects the ideal race time. It does not take in account such factors as the terrain, weather, shoes, your technique, the pacing strategy, nutrition & hydration, mental state, your will power, etc. Also please note, that typically you'll do quite a bit better time in a true race, than when you run alone. And then, if you do not use a HRM chest strap, the prediction which relies a lot on the VO₂max estimate, may not be as accurate as it could.

EmBer said:

is there a way to trim these values?

The predictor times are just estimates, telling how fast you could possibly run at your fitness level. It does not have any negative impact if they do not match with your real performances, so there is no need to trim them. If you want to set your own predictions or goal times, you could create an 'Event' for each of the distances with a goal time.

I agree. All these estimates are also way too optimistic to me. Similarly even the daily suggested workouts are nonsense: they give me runs that are supposed to be base, but with a pace faster than any of my runs in the past months, most of which were categorized by the watch as  threshold or tempo

I'd take it on the positive way - it sounds that you have big reserves, and with your physical state and fitness level, you can run much better if you work on your technique, strategy, pacing, nutrition, etc. So there is a lot of room for improving your performances.

In contrary, in my case, with accurate predictions, I seem to already have exploited these reserves, and my only way to improve is working hard on increasing the fitness, including the VO₂max.

That makes no sense. The race prediction should be for if I ran today or next week and not reflecting my genetic abilities that I can achieve in 6 months. And if they meant to predict my genetic max, then why is it changing all the time 

flocsy🤠 said:

And if they meant to predict my genetic max, then why is it changing all the time

It is not really directly the genetics, it is mainly your current fitness level and your training history that determines the prediction (though the genetics plays certainly a role in both of them). However, the real performance then depends on many factors that cannot be taken in account in the prediction - whether you will have a good or a bad day, what weather there will be, what terrain you'll run on, whether your technique will be sloppy or excellent, whether you run with full stomach, which shoes you take, the pacing strategy you employ, whether you prepare mentally and physically, etc, etc.

For example my 5K perf can vary from 22 min to 40 min, depending on those previously mentioned factors, and there are days the slow run is more exhausting than the fast one a few days before. However, when I go to a race, prepare correctly, and manage to keep my warmup, technique, strategy, pacing, and nutrition on the top level, then I indeed reach the predicted times rather accurately. I do not expect to reach them on every my run every day.

Otherwise my predictions are rather stable (see below), and evolve only relatively slowly. If it is not the case on your side, then there are minimally two possibilities - the entry data is not accurate enough (i.e. you do not consistently use a HRM chest strap and the OHR does not work well for you), or your performances are chaotic, not resulting in stable metrics, and consistent results.

Garmin's race predictor can spot general trends (do your times go up or down), but like so many other Garmin metrics (body battery, sleep score, ...), the times themselves are fantasy numbers for most people. Not for all, though, because even a broken clock gets the time right twice a day :)

What makes Garmin's race predictions way off:

(1) Garmin uses its own VO2max estimate, often wildly inaccurate because your watch is unable to measure oxygen consumption. Worse yet, Garmin's VO2max calculations (based on accurate time, somewhat accurate GPS distance, questionable heart rate) converge to the average for your age and sex. If you're much fitter (or way less fit) than average, Garmin's VO2max calculation gets it wrong. The same problem bites Cooper tests and BMIs: useful for a large population but not for an individual.

(2) The relation between VO2max and race times is rather loose and useless for individual race time predictions. That's why all those online race time calculators get it wrong most of the time. Predictions based on aerobic and anaerobic thresholds are much better, but because Garmin doesn't know these numbers it has to guess your VO2max instead.

3) Garmin doesn't know if race day will be hot or cold, humid, windy, or if the course has lots of curves, hills, crowds, asphalt or dirt, etc.

Garmin should add a "lab test" activity type to its app, with data fields to enter your lab-measured AeT/AnT/VO2max speeds and heart rates, and use those measurements to correct its guesstimates.

I see the estimates updating through the year according to my training levels and, yes, they're usually a little bit faster than I'm immediately capable of. However, if I'm doing an event rather than just some training then I'm more likely to consider nutrition, hydration, rest, sleep and general "race" preparation in the week or so leading up to it in order to achieve a faster time.