A summary of my Reedham Ten road race, including a discussion of how my performance compared to predictions - spoiler, TrainAsONE's AI prediction was only 1% off.
As outlined in a previous post, the Reedham Ten is a 10 mile road race starting and finishing in the village of Reedham in Norfolk, England. It was the inaugural race of the 2023 Sportlink Grand Prix calendar, no doubt a contributing factor to it being a popular event and fully sold-out.
My race
I had not run this race before, but understood it to be undulating around country roads, with minimal hills to contend with. On walking up to the start line, the wind was quite strong and had a definite chill. My thought was that it was certainly a day to take the course as it comes: make use of the down hills and tail winds when possible, and not to push too hard when conditions were the contrary.
TrainAsONE had calculated I could go sub 71, and I was going to give it my best shot.
Upon starting, my focus was sheer determination that I was going to hit my TrainAsONE race prediction(see my pre-race post regarding this). If TrainAsONE calculated I could go sub 71 minutes, then I was adamant that as I got tired over the latter stages, I would not end up loosing focus – and fail.
The very first couple of kilometres felt fast! It was a good downhill that seemed to last quite awhile. I recall it then turning into a flat stretch, but the elevation chart suggests there was an uphill section too – my mind has wiped that from my memory. This first section was all with a tail wind, hence the fast start. Whilst the terrain then became more undulating, it took a few more ‘k’ for the wind to change direction, but by then I had settled down into a good pace. At the half-way point I was just over 30 seconds ahead, and considering this, my legs were still behaving themselves.
As one might have anticipated, it was from this point that the wind decided to be less friendly as it subsequently always seemed to be head on. I’m sure it wasn’t, but rather a case of my legs starting to feel the previous 40 minutes of hard running. They were starting to put seeds of doubt in my mind. But I pushed on.
With 5km to go I was still hardly slowing, but I knew I was running close to my limits. Only a Parkrun to go… Only 4km to go… Less than a dreaded 3.2 km assessment to go… With only 2 km left, I felt I could push for that finish, but common sense prevailed and I resisted. Until the mile point. From where I picked up the effort for one last gritted performance. Unfortunately this did not last the full mile, as at the last 400 metres my legs were spent and I only had enough left to prevent a person behind snatching the finishing line before me. It certainly was not my strongest of finishes, but I had done enough to nicely come in under my time. 69 minutes and 54 seconds (though I am still awaiting to hear the official times…) Job done.
Analysis
Overall, I would say that it was a (hard) but well run race. I averaged 4:17 min/km for the first half and 4:24 min/km for the second, about 3% difference. On paper this might not be the best splits, but would fit with my recollection of the conditions.
I over-performed TrainAsONE’s prediction by 1%!
So I exceeded the TrainAsONE prediction by 54 seconds. That’s a 1% over-performance, which is a great match. For comparison, I outperformed the 6 minute based prediction by over 4%, and the 3.2km based prediction by nearly 8%.
Above, I have over-layed my race speed onto the box chart from my previous post, which nicely illustrates how each of the prediction methods faired against reality. Reality nicely intersects TrainAsONE’s inter-quartile range, just above the target. Similarly my actual speed fell within the inter-quartile range for the 6 minute assessment prediction. However, as suspected before the race, the 3.2 km prediction is further off, though I did still fall just within the the upper whisker, so not a complete outlier.
Whilst TrainAsONE’s AI certainly looks to be more accurate, I think it would be unfair to say that Riegel’s formula is bad. There are a few of reasons why I say this.
Firstly, their inter-quartile ranges are not too far off the mark, and secondly (and probably more importantly) we are not applying the formula as intended. Riegel was designed to use a race as input, and we are using an assessment / time-trial. And despite the fact that I have been running such assessments for many many years, and so very experienced at them, they probably just do not bring out the best in me. I suspect this is the same with many other people, hence the long whiskers.
Algorithm | Lower Whisker | Quartile 1 | Quartile 3 | Upper Whisker |
Riegel (6 min) | -7.9% | -3.4% | 7.6% | 20.8% |
Riegel (3.2 km) | -10.5% | -4.5% | 0.1% | 7.9% |
TrainAsONE AI | -2.6% | -2.3% | 2.7% | 3.9% |
The additional problems with using Riegel’s formula against an assessment or a race is that it does not consider specificity of the target race and ignores all the training from the assessment / test race up to the target one. Especially for the more amateur runners, where training is a little more sporadic (due to the juggling of personal life, and perhaps motivation) I think this is very important.
Parting Thoughts
If I did not know my predictions beforehand, how would I have performed?
Personally, I think I could well have been 30 to 60 seconds slower, putting me even closer or spot on TrainAsONE’s prediction!
A massive shout out to the race organisers, volunteers, and all those involved. A thoroughly enjoyable event. Thank you.
Till next time.