Race prediction to improve after you race a classic distance

Yeah I have noticed things like that on the odd occassions before. Have you had your 945 long and/or have you recently completely reset it? I do find it took a few weeks before it learnt me. Maybe also if you do a track 5000 (good time BTW) that is a slightly fastest environment that your other recent runs and so the 17:20 is more reflective of that environment.

Also how much distance did the 945 actually record for the 5000? I have found the new Track Run feature to be a great way to get a lot lot better distance accuracy on a track.

VO2Max is the primary input into the algorithm. If you dig through the Firstbeat whitepaper, they claim accuracy with their VO2Max estimation to within about 5%. Your predicted time was about 4% of your actual time.

However, it is also worth noting in that white paper that the accuracy of the VO2Max estimation declines if it doesn't also have an accurate fix on your maxHR. If you don't already have a good idea of your true VO2Max, some further field or lab testing here may improve your race time prediction.

Thanks @timgrose. have my 945 since July 2019, did a couple of hard resets since then, the last one recently. For this 5000 m race device recorded 5,15 km with a (fake) personal record on the gps distance of 5 km at 15'46" (obviously not real). I will try the new feature that Garmin is going to release in the next update release

Yeah if you've seen 195bpm during short races, assuming it was at the end during all out sprint... i'd say 196 would be a smart input.  I think 195-197 would be legit inputs.

My guess is that the vo2max hasn't fully caught up to your abilities.  For me it seems I always get increases to vo2max score when I do a sustained 18-25minute hard Tempo run when it is fairly cool.  Gives the calculations a good chunk of time to analyze HR and pace pattern at a strong effort.  Long enough also that any smaller hills get washed out in the pace vs HR trends.  Pace that gets HR up to 82-85% of max maybe in first mile, then slowly rises to 87-90% and somewhat plateaus.(obviously depending on your own lactate threshold and aerobic fitness).   I've found helpful to cut cooldown short (at a couple minutes)...and start a new activity ('Trail Run') that won't effect or sway vo2max calc, to finish off the last mile or two cooldown.  Since after a good warmup and strong threshold run depending on caridiac drift (heat, dehydrate,etc)... if it sees you running an easy pace with a HR that doesn't drop enough... it thinks Oh-wait-maybe this guy isn't that in shape. lol (just my over thinking things...)  

I've had way to many issues with my old watch giving bad vo2 data from when overheated, hungover, etc.  So i've done a lot of reading on it and thinking about ways to prevent the one off bad readings or how to hopefully ensure a good reading.  

I think the algorythm already take in to account the runs considered as "relevant" and cut from the VO2max calculation those odd or short runs like cooldown or warm up...