I only run on average one marathon a year, but as I alluded to in a previous post (“my first time“), running the marathon distance was what helped me develop a running habit in the first place; and the way in which I continue to validate myself as a marathon runner. You can imagine my surprise then, when I confirmed a few weeks ago, 7 years and as many marathons later, that I’m not very good at it!
And I don’t mean “not very good” compared to someone else. The elites, or even what we would call “proper athletes”, are too far ahead of me to permit any comparison, but on the other hand my time in London last year was in the UK’s top 22% (without even being a PB), so I am not that far back either. No, I am comparing how I do in marathons to my performance over shorter race distances.
Most of us are familiar with the race predictor tool, found on Runner’s World and elsewhere, which allows us to estimate our race time for a race distance, based on a given time for a known distance.
It is based on population averages (i.e. the degree to which race times get slower as distance increases) and the main caveat is that you train appropriately for the distance you want to predict a time for. E.g. just because you can run a fast 5k, this doesn’t mean you can achieve your predicted marathon time without any marathon-specific training.
Unfortunately, the accuracy of it’s forecasting, in my case at least, drops quite significantly for races from a half marathon and upwards, despite the amount of training I may do. Now, this could be because I have a natural bias to shorter, faster races (the second caveat RW mention), or it could be because more of my runs
(taken as a total during a year) are up to half marathon distance than above; or it could be because I am influenced by the predictions of the calculator, and set off at a pace which I cannot sustain for the entire race.
The geeky stuff:
Very helpfully, Runner’s World gives the formula for the predictor on their site, which means that one could list all distances between 1 and 27 miles (its accuracy drops for times under 3 minutes and over 4 hours) in an Excel worksheet, and use it to plot a line of predicted finish times, based on a known race time. Below is an example based on my 5k PB. Using my 10k PB produced exactly the same results (there was a difference of 1 second at the predicted marathon time), even though the 5k and 10k PBs were set 4 ½ years apart:
The formula behind the calculator is Tp=Tk*(Dp/Dk)1.06, where:
Tp is predicted time, Tk is time of known race distance, Dp is the race distance for which you want to predict your time and Dk is the race distance for which you know your time. The exponent 1.06 is there to reflect the fact that your pace will slow as the distance increases (so your 10k time won’t be double your 5k time etc). The greater this number, the more the calculator will adjust for distance (i.e. the slower your 10k predicted time will be from your known 5k time) and vice versa.
And this is the interesting thing: 1.06 must be accurate for most people (which is why it’s used in the generic version of the formula), but when I then plot my own PBs on the chart above you notice that my PBs for longer races (from half marathon onwards) tend to diverge from the line (the kinks in the line in this chart are caused by the extra data points between full miles; e.g. 13, then 13.1 for the half marathon, then 14):
In fact the divergence increases with the distance, not only in absolute values, but also proportionally: so with a 5k known time, my 10k time is spot on RW’s prediction; The half marathon takes me 3 minutes 1 second or 3.09% longer; my 20 miles PB is 7 minutes 56 seconds or 5.19% longer; and the marathon a massive 19 minutes 11 seconds, or 9.42% longer than RW’s predicted time!
Which is what led me to my conclusion at the beginning of this post, that, compared to my 5k and 10k times, I’m not particularly good in this marathon malarkey!
Why does this matter?
- Choosing your training: I think that anything that gives us an insight into ourselves as runners is very useful in helping us decide which elements of our training we should concentrate on, and how to adapt our training programme to help it address our specific needs. This insight for example, has confirmed that in the latter stages of a marathon I fade much more than the “average person” with an equivalent 10k PB as me would. So I have decided to turn as many of my LRs into progression LRs (especially in the “peak” phase of my plan), to help with my discipline in the early stages of a race, and my speed endurance in the latter third.
- Choosing your pacing: before a long (half marathon onwards) race, I tend to use a recent race result to get a predicted race time, to gauge my form and decide how to pace myself in the early stages of the race, when most paces seem easy. But if the RW calculator tells me that my predicted finish time is 3h 20′ when my potential is in fact closer to 3h 40′, I then set off at too fast a pace (in fact 28” per km / 46” per mile too fast!), chasing an unrealistic objective. No wonder I fade in the end! Could it be that by setting off at the 3h 40′ pace would mean I have more energy at the end and actually achieve a better time than that?
- Choosing your battles: Perhaps you want to concentrate your racing on distances you are comparatively good at; or you are aiming for your next club standard, and you want to choose the race distances that are most likely to give you a good result, compared to the standard. Knowing whether you have a bias towards shorter of faster distances will help you decide whether to sign up for a 5 miler or a marathon!
Some other considerations:
- We all have faster and slower periods / years. So initially I had plotted my best performances for each year, in case my bias was different from one year to the next. Turns out (for me at least) it isn’t, so I’ve used PBs.
- Beware of outliers: I mentioned before that my 5k (3.11 miles) and 10k (6.21 miles) PBs are consistent with the calculator, and that my performances begin diverging from the half onwards. But if you look at my PBs for 3 and 6 miles, they lag behind: this is because while I’ve ran many 5k and 10k races, I’ve only ever ran one race on each of the 3 mile and 6 mile distances and, the 6-miler in particular, was quite hilly. I therefore decided that these are just outliers, and disregarded them.
- But don’t dismiss unusual races altogether: for example all my 20m races were on the same, hilly course. This could be used to explain why I am slower at that distance than the calculator would predict, but given that my divergence at that distance is consistent with those for my half and full marathon PBs (both on fast courses), the course difficulty probably isn’t a factor.
So what’s this about a better race predictor then?
So what if instead of using one race to predict performance in another (e.g. a 5k to predict half marathon performance) we could take account of all our PBs (or Season Bests if we preferred) for different distances and came up with a prediction that takes account of how our pace is affected by the race distance?
The good news is that I have created just such a tool! And what’s more, it includes an option to also use a recent race time, over any distance for which you have an established PB, to see how your current form affects your predicted race time (this can be a breakthrough PB in any distance, or being slower than your prime)!
That was relatively simple to do for my race history. What was much, much harder was to then create a template which can be distributed publicly and which – in theory! – should work for most other runners. It involved a lot of experimenting with Excel formulae which I had never used before, a lot of mumbling to myself, very late nights and perhaps some wine for company.
But finally…. (drumroll please)…
Now, please bear in mind that this is only the first version, and I’ve tested it on myself and one other runner! But by all means, have a play round and let me know if it works for you. There is a “mailto” link in the file, or use the form below to send me your feedback: