Related
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

955 massively underestimating my ability

loltracking
over 1 year ago

My predicted race times and my "stamina" meter are way off and have been over the course of 3 months of consistent use. For example, my race predictor gives an estimated 5k time that is only a few seconds faster than the 5ks it has recorded from tempo runs. When doing any speedwork, my stamina will regularly go to 1% halfway through each interval. At the same time, it does seem to be recording improvements, as my VO2 Max and estimated times steadily improve as my workouts get faster. 

While I haven't raced with this device yet, I'm quite sure it's very wrong, based on past experience with workouts and races. 

What could be causing this issue?

  • 0 over 1 year ago in reply to Sergey_Kessel

    To my knowledge, all the Garmin race prediction and similar algorithms are run on the watch, not the servers. And Garmin doesn't even have to licence anything, since they bought FirstBeat in 2020.

  • 0 over 1 year ago in reply to Sergey_Kessel
    Sergey_Kessel said:
    Also personally it is hard for me to beleive into that 5% error rate. Really - my mistakes are bigger. I am running more or less same pace (1.34-1.40 for half), but in May it was 56 (loads of 1000m intervals at 3.45) and today it is around 52 (I just do less short intervals - almost none). Topicstarter had 5 min+ mistake for 1:24 race - that is 6% difference. I had vo2max of 51 in spring with that 1.35 time, prediction was around 1.41-1.42 - that is 7% mistake. That is still impressive (no joking) but totally unusable. 

    The 5% error claimed by Firstbeat is the "mean absolute error". To calculate this, they measured the error for each participant in their study, converted the errors to absolute values (so negative errors and positive errors didn't offset each other), and took the average.

    Roughly speaking, this means that only half the participants had an error of 5% or less. The other half had an error that was at least 5%. (I'm glossing over the difference between mean and median). In other words, an error of 6% or 7% isn't at all surprising.

  • 0 over 1 year ago in reply to Sergey_Kessel

    It does definitely consider the length of your long run, thus it very probably also considers mileage.

    I did a long run and the race predictor said, in regards to an upcoming marathon, something like "great job,! your last workout was beneficial and improved your predicted race time". And the prediction dropped by 1 min or so

  • 0 over 1 year ago in reply to 7dfcd1938c0146f1804397d33ddc9c14

    Why you think that it was milage that was counted and not TE aerobic score? It may be te, or workouts longer then X with te higher then y. 

    Your may be right, but what you stated is a guess. 

  • 0 over 1 year ago in reply to martinleroux

    Didn't know that, assumed they used normal distribution and trusted interval (which is likely to be used by stryd as mentioned above since they state +-half a minute). 

  • 0 over 1 year ago

    Actually I think links/ posts/ quots from this thread are really useful and thank you to those who posted that. I'd say post it totally different from "fix it immediately" that I often see in the posts here. 

    Thanks for everyone posting here and to the starter as well