Race Day Tomorrow – Reedham Ten (Take 2, with predictions)

An update on my training, and some predictions for tomorrow's 10 mile race.

Take 2? That’s because the Reedham Ten was originally scheduled for the 22nd of January (two weeks ago), but was postponed until tomorrow due to icy weather conditions. Here’s my original ‘Race Day Tomorrow – Reedham Ten‘ post.

My Fit Scores are up!

The good news is that the extra two weeks of training and preparation looks to have improved my prospects for tomorrow’s race. All my Fit Scores are up! (Only a tad, but excellent for only an extra 2 weeks.)

That’s nice, but how does this translate into the real-word? To answer this, and with the assistance of TrainAsONE, I’ve collated some prediction stats…

Prediction Stats (the old school way)

I have not run a recent race, not even a Parkrun, so basing any predictions from one of them would not be valid.

I have (naturally) run my TrainAsONE fast assessments. So we can use the mainstay of race predictions, Riegel’s formula, against these assessments in order to estimate my performance tomorrow.

6 Minute Assessment

72 minutes, 57 seconds

Riegel’s formula, 6 min assessment

My latest 6min assessment was my best performance for a good number of years. In the 6 minutes, I managed to run 1,521 metres. A pace of 3:56 min/km. Using Riegel’s formula this equates to a time of 72:57 – a 4:32 min/km pace.

3.2 km Assessment

75 minutes, 6 seconds

Riegel’s formula, 3.2km assessment

My latest 3.2km assessment run was not my best recent performance, but still up there. I managed to run the 3.2 km distance in a time of 13:35. A pace of 4:14 min/km. Using Riegel’s formula, this equates to a time of 75:06 – a 4:40 min/km pace.

So that’s around 2 minutes of difference between the two calculations. About 3%, which based on my experience in looking at such numbers, is around expected. However, I ran a sub 73 minute 10 mile race in early 2022 (nearly a year ago), and I’m confident that I’m presently fitter over the distance. I did run my socks off on that race, but I still would be very disappointed not to beat it (discounting any differences in course and conditions). Consequently, I do feel that both those predictions (and especially the 3.2km based one) are a little under-representative of my ability.

Prediction Stats (the new way…)

Fortunately, I have access to TrainAsONE’s research and development, where a novel and unique race time prediction algorithm is almost ready for live release… The TrainAsONE algorithm takes all your recent raw running data and puts it through cutting-edge machine learning and artificial intelligence algorithms to generate a race time prediction. A great aspect of this approach is that the TrainAsONE algorithm does not rely on any races, benchmark runs, or assessments, just raw running data – how cool is that! And it works for all race distances, including ultra-marathons.

70 minutes, 48 seconds

based on TrainAsONE AI

So the TrainAsONE AI calculates that my 10 mile race time tomorrow is most likely to be 70:48 – a 4:24 min/km pace.

That’s over 2 minutes (or 4%) faster than the traditional Riegel’s formula on my 6 minute performance, and over 4 minutes (or 7%) faster than Riegel on my 3.2km assessment.

The other nicety of TrainAsONE is that it can provide more than just a single value, but also a likely error range for these predictions, i.e. given a group of similar runners it can illustrate how their race times are likely to be spread around the predicted value.

The box plot above displays the median (horizontal lines), inter-quartile ranges (boxes), and whiskers (vertical lines) for each of the above algorithms. To help those who are not familiar with reading box plots, the salient points are that the horizontal line is the most likely predicted value and the box demonstrates where the true value for 50% of people will actually fall. And the whiskers help to illustrate the spread of everyone else.

We’ll have to wait until after tomorrow’s race to know which could be considered the best in this particular case. However, I think the main aspect to take away from this chart is that across the sample population of TrainAsONE runner’s used to validate these algorithms, the spread of error was less with TrainAsONE’s novel AI approach over the traditional Riegel formula.

On looking at the box plots, one might reasonably conclude that a velocity of around 3.7 m/s is a ‘likely’ speed. This equates to a 4:30 min/km pace, with a time of 72:25 – a little faster than the 6 min Riegel prediction. However, given the source of that 70:48 prediction, I really have to aim for that… I feel it may be just a stretch too far, but I’m going to give it a go! (Fingers crossed that I don’t blow-up and hit that 3.2 prediction time!)

Let’s see what tomorrow brings…