What does TrainAsONE predict my time will be for tomorrow’s 10k race?
Race day again tomorrow! Another 10 km. This time it is the Mike Groves 10k hosted by Coltishall Jaguars Running Club at RAF Coltishall, just north of Norwich. The route is around the perimeter of the airbase, which as one might expect, is quite flat. “Good for PB’s except when windy as it is very exposed” apparently. The forecast is predicting light rain with a gentle breeze. So fingers crossed, PB-friendly conditions.
It’s been three weeks since my last race (the Valentine’s 10k), and whilst I have kept up the consistency of my training, it has not been as focused as I would have liked or hoped. Pressures of work with resultant late nights have necessitated me to cut short and downgrade workouts. (I use the term ‘downgrade’ to mean changing a TrainAsONE scheduled speed session or hard assessment into an easy Economy workout.)
This did mean that I got out of doing a dreaded 3.2 km assessment run. However, given that I am also training for the Wymondham AC 20 miler race on the 26th of this month, I was actually a little annoyed that I was unable to fit in a long run scheduled for early last week. C’est la vie!
Despite my tiredness, and in a effort to help gather more valuable data for the TrainAsONE machine learning algorithms, I did force myself to run my scheduled 6 minute assessment. As is often the case, once out, the cold air woke me up and I had a good run, managing to cover just under 1.5 km at a pace of 3:52 min/km. I was rewarded for my stomach churning lactate inducing run with my best 6 minute performance for nearly 5 years, reflecting how my recent races have gone, and bodes well for tomorrow.
In addition, a friend twisted my arm to run a Parkrun. I got a course PB. My fastest Parkrun in quite a while, and again good signs for tomorrow.
My Race Predictions
Two weeks ago, I discussed 10 km race predictions in my article I used AI to travel back in time, and re-run my 10 km race.
So what are my predictions for tomorrow?
Algorithm | Days Since Assessment / Race | Pace (min/km) | Time (mm:ss) |
Riegel (3.2 km Assessment) | 53 | 4:25 | 44:10 |
Riegel (6 min Assessment) | 12 | 4:20 | 43:21 |
Riegel (Parkrun) | 15 | 4:17 | 42:57 |
Riegel (10 km race) | 21 | 4:14 | 42:15 |
Riegel (10 mile race) | 28 | 4:10 | 41:40 |
TrainAsONE | n/a | 4:09 | 41:32 |
So as depicted in the table above we have a range of predictions from the slowest time of 44:10, down to the fastest time of 41:32.
One would think that given that my 6 minute assessment and Parkrun were very recent, and best recent performances that they would provide good predictions. Unfortunately, their predictions are slower than my last (and very recent) 10 km race, so not as good as one would have hoped!
Who will be closer? Reigel’s 41:40, or TrainAsONE’s 41:32?
TrainAsONE’s prediction of 41:32 is particularly interesting (not just because it is the fastest), but because of the 41:30 prediction a couple of weeks back as discussed in my previous article: I used AI to travel back in time, and re-run my 10 km race. To be honest, I was hoping this latest prediction would be faster than 41:30, but let’s see what tomorrow brings.
Above are the box-plots for the various prediction methods. I have not included the Riegel (10 km race) as it is a little nonsensical (predicting a race time based on a race time for a race of the same distance).
A point of note is that the inter-quartile range for the Riegel (10 mile race) is narrower than that of TrainAsONE, whereas the extent of the whiskers is greater. So it seems that in this case, whilst Riegel might be a little more accurate for 50 % of people, TrainAsONE is more accurate for the whole population.
I’m setting my sights on that 41:30 time. I’ll let you know how I get on.
And remember, if you are running or spectating at the Mike Groves 10k at RAF Coltishall, Norfolk, tomorrow, and have a camera to hand, look out for me and you could have a go at winning yourself a TrainAsONE premium subscription and T-Shirt by playing ‘Spot the Sean‘.