It’s the Trowse 10k tomorrow, and being only a 10 km race it will feel a lot shorter than my last one (the Wymondham 20 miler). It is hosted by the City Of Norwich Athletic Club and is the fifth race of the 2023 Sportlink Grand Prix.
Like all my races this year, I have never run this event. The course is described as ‘mixed terrain’, but I think apart from a short ~ 500 metre section of track it is on roads. The route is basically a two lap circuit around the village of Trowse (Norfolk, England) , with around 85 metres of ascent / descent, making it a little bit hillier than both of my other 10 km races this year.
Since the Wymondham 20, training has been relatively uneventful, with one long run (of around 20 km) and a small Interval session (consisting of 3 repeats of 3 minutes at a planned pace just slower than my estimated vVO2max), interspersed with easy runs of around 5 km.
With a mixture of recovering from the Wymondham 20 and late nights working, I’ve been feeling tired and not completely ‘on form’. However, no doubt a result of the easy taper (courtesy of TrainAsONE), I started to pep-up yesterday and now feeling much better. Fingers crossed this bodes well for a great race.
Surely race preparation takes priority over painting!
The only negative is that I woke up this morning with a bit of a sore back. I’m sure this is from painting a ceiling yesterday. I wonder if Lou (my wife) will accept that as an excuse not to continue with it today… Surely race preparation takes priority over painting!
Race Time Predictions
As with my previous few races this year we’ll compare how TrainAsONE’s unique AI race prediction compares to the estimate provided by the use of Riegel’s formula.
Assessment / Race
|Pace (min/km)||Time (mm:ss)|
|Riegel (3.2 km Assessment)||32||4:39||46:30|
|Riegel (6 min Assessment)||46||4:22||43:43|
|Riegel (20 mile Race)||13||4:16||42:50|
|Riegel (10 mile Race)||62||4:13||42:11|
|Riegel (10 km Race)||34||4:07||41:13|
Consistent with the pattern of the year, TrainAsONE is predicting the fastest time. I put this down to the fact that it is the only algorithm to take into account latest training. I am continuously training, and so (at least in theory) getting faster relative to my previous race(s).
Following on from the theme of continuous training, conditions accepting, one would expect my time tomorrow to be a fraction faster than my last 10 km (which was only 34 days ago). This fits with TrainAsONE’s prediction, however with the course being hillier than the last race, I’m not pinning hopes on another PB.
The TrainAsONE prediction is not considering elevation changes, and an additional calculation would suggest that the hills will add around 20 seconds to my time, so around 41:25.
Some observations following a review of the box plots.
Given that the Riegel (10 km Race) prediction is estimating a 10 km time based on the results of another 10 km, one might expect the error to be evenly spread around the predicted value. However, on average it seems that people perform worse, the median error being around 9%.
Only the Riegel (20 mile Race) has a median error that is greater than the prediction, i.e. on average, all the algorithms except the Riegel (20 mile Race) overestimate race performance.
It does appear that the TrainAsONE algorithm is the most accurate
In just about all respects, it does appear that the TrainAsONE algorithm is the most accurate: the median error is the smallest, along with the inter-quartile range.
For this race I have obtained 2 further predictions: One from a well known sports watch company, and the other a well known athlete data analysis platform.
|Source||Pace (min/km)||Time (mm:ss)||% Difference from TrainAsONE|
|Athlete data analysis platform||4:26||44:20||8% slower|
|Sports watch company||4:18||42:56||5% slower|
Like TrainAsONE, both are considering all my health and running data, and I think they provide an interesting comparison. I believe recent history already tells us how accurate they are likely to be.
I was prompted to look at the above following a question from a TrainAsONE user (unrelated to race predictions). My understanding is that both of these services use heart rate as a prominent, if not a major, component of their predictions. An aspect that I believe is a cause of the issue. Part of my reasoning is nicely exampled by my current ‘fitness’ calculations in these platforms – they show an unexpected pattern. I intend to write that up in a separate article in due course.
On to tomorrow! Let’s see if I can get another PB!
(I’ll let you know how it goes.)