Apologies if this has been requested already, couldn’t find it.
Runalyze has effective vo2max which determines race prediction on a day by day basis, taking into account how more or less fit I am over a season. It works well.
Can something similar be done here? I feel that it would be more accurate because intervals.icu is using my LTHR, which I keep up to date, and also considers resting HR.
Yes - any arbitrary distance which is what makes it cool. This is how it works:
effective vo2max is essentially the closest fit line of all your runs’ “performance” (eg, avg pace combined with avg HR, at its simplest). The fit line itself can be personalised with a correction factor, as you can see it fits very well my recent races/time trials (bigger circles on the right-hand side of the curve) and since my season ended, effective vo2max is going down (currently estimated at 42.9)
These are arbitrary distances that I’ve inputted, and Runalyze gives me the prediction corresponding to a 42.9 vdot (for each distance, to the left is my best performance, to the right is my current prediction)
NB: Runalyze also uses “Marathon shape” as a correction for distances longer than 15K or so, so my current marathon prediction is shown as 4h44m because I’m running low mileage and no long runs, but it would be 3h37m for a vdot of 42.9. It’s valuable (and if you can figure out a good measure of “long distance endurance” to correct fitness-based race predictions, that’d be awesome) but for now I would be happy with a simpler race prediction based on Fitness or Lactate Threshold Pace or something.
Tx for all this.I have creating a “pace vs distance” chart on the todo list (same as power duration curve basically). With any luck power models will work for this then it will answer best time for x distance.
They call it “Running Index”, but it’s just a vo2max prediction based on chunks of running longer than 12min where speed stays above 6km/h (but I remember reading that a couple of stoplights are also allowed). Anyway, it’s not terribly effective. On the plot above, the best predictions for 5k and HM are 18:50 and 1:26:30. I’ve run 17:00 and 1:18:20 this year.
In my experience, race times prediction works best when comparing real races. E.g. 5k <> HM conversion usually works for me up to few %
The neat thing is that the corresponding **time-vs-distance plot is linear ** - and the slope of the line would be the CS in m/s. So it’s very easy to find.
Critical speed is important, because it’s the theoretical limit one could run “forever” - and one could even try to use it predict marathon time (better trained runners can use a higher % of their CS). I suspect Runalyze (and Polar) are using something like this for their “effective vo2max” / “running index”.
Might be of use maybe? A Critical Speed calculator taking in best efforts (three distances chosen by user?) in a certain date range would be neat, although it wouldn’t be a moving average (eg, would not represent your daily race readiness, but your season’s or all-time CS).
You might find this useful for calculating your critical speed. I’d add a word of caution when using CS for predicting anything above about 10k though for us amateurs.
Hey that’s very cool - it seems like I was already registered though, and forgot the password. There’s no password reset mechanism, I feel like that is pretty important to implement
Is there a support email I can write to, to get it reset?
Just reviving this thread to let folks know there is still interest! Critical Speed is really bad at predictions if you have faster times in shorter distances. For example I’ve run a 5k at 6:03 pace, and because my 2mi record was also set during that run at ~6:01 pace, CS R2=0.9999 thinks I can run a half marathon at 6:06 pace, whereas Vdot would predict 6:35 pace which is obviously more reasonable.
Or just implement Vdot I ran a 5K 2 weeks before a HM, both all-out, and vdot predicted my performance on the HM within one second. I didn’t believe I could run it and started much more conservatively, but negative split to a 1:23:54 which was a good HM debut for me.
t2 = t1 × (d2 / d1)^b could work if you try to estimate what the exponent is for the person (I definitely think the riegel b=1.06 is too aggressive for me.
Vdot calculation is this, for reference if anyone looks later!
where: S is speed in metres per second T is time in seconds
My understanding is that to predict a different time based on this, one then best-fits several different times for the pre-set distance to approximate the S where the V02 max matches the athlete’s Vo2 max. Complicated, but extremely effective I’ve found! It’s how I set all my running paces and has led to success for me including PRs at 1mi, 5k, and HM for me in the past 3 1/2 months of racing.
A couple of months ago I was intrigued about the VDOT system and how Daniels’ paces come to be. It’s not that complicated and revolves around the two equations you shared. I created a notebook that tries to be comprehensive about the mathematics/approach.
The “official calculator” is most likely using a more sophisticated approach (particularly for the lower ranges of VDOT). After all, those equations are based on a book from 1979.
as i go deeper on running i really miss the intelligence of the bike metrics for evaluation and updates for training zones. as power is kind of a messy subject on running, there are plenty and very simple calculations based on paces that would save people a lot of time, specifically:
threshold (based on 3k+ efforts) and other zones
and predictions (5k, 10k, half, marathon…).
from there, would be nice to have things like:
timed deprecation, something similar to eFTP deprecation
visualization on fitness charts as we can have for eFTP
One solution would be to start with the Tanda Marathon Pace Predictor ( RUA: A simple relationship for predicting marathon performance from training: Is it generally applicable?). There are several online versions. However, the formula is quite simple:
Pm (sec/km) = 17.1 + 140.0 exp[-0.0053 K(km/week)] + 0.55 P (sec/km)
Pm = predicted marathon pace
K = (mean?) weekly run distance the last 8 weeks
P = (mean?) run speed during the last 8 weeks
Based on this formula, one could perhaps create a custom Wellness field or custom chart. Unfortunately, I lack the necessary programming skills.
I’d like to anti-recommend Tanda in favor of VDOT or other systems. The tanda formula is based on a very limited sample size, and crplot.com and other implementations don’t seem to update based on troves of data available. Offsets can be used but need to be customizable. For me, I’ve outperformed Tanda by 15% in my last two HMs, and by 7% (>15 minutes!!!) in my last FM. Only times I’ve underperformed Tanda are trail races (which makes sense) or when I’ve gone in races un-trained. Some of these points are even the run legs of triathlons.