281 - Age grading 2 | Scoins.net | DJS

281 - Age grading 2

I find many people asking whether a time and distance combination is 'good' and increasingly I find this irritating. In an attempt to have a prepared answer, I look at that general problem here.

Of course, 'good' is a personal self-reference usually, and so requires you, the individual, to have an idea of your previous data and therefore how the most recent data compares with this. Personally, I have found that age-grading provides this well, since it removes the degradation that goes with advancing age. ¹   

Male, 35, 5km road

  Time         Male         Female

24 mins     55.25%     61.93%
23 mins     57.66%     64.62%
22 mins     60.28%     67.56%
21 mins     63.15%     70.77%
20 mins     66.30%     74.31%
19 mins     69.79%     78.22%
18 mins     73.67%     82.57%
17 mins     78.01%     87.43%
16 mins     82.88%     92.89%
15 mins     88.41%     99.08%

As a general guide, I find general agreement with these labels: 90% = world class, 80% national class, 70% regional class, 60% local class —  where we understand that the 'local' class is self-referencing, from 60% and upwards. Which puts me solidly in the 'regional' range, borne out by the associated experience. I'm not at all sure what 'local' means, but it would be unlikely to produce a win in any race. A male of 35 at 5km is looking at a time difference of two to three minutes for the 10% between 60 and 70% [table inset to right] so we might expect around 3.3% per minute improvement and at around that age one adds 0.4% for each additional year. At my sort of age it's a lot more like 1% a year, 20-30 seconds per year on the parkrun distance.

Those without a history of exercise will have what I could call an initial age-grading figure which is surely going to improve with time. For example, I have a friend who has now been running for 4 months (she is approaching 50); her first and significant success is actually managing to complete such a distance without stopping. Whatever time she clocks for her first 5km, let's say around 40 minutes and therefore about 40% age-grading, will immediately become a target for improvement. [36:30 45%, 33 mins 50%, 29:30 55%] and as time goes by I expect her (people in her position) to experience improvement, rapid at first and then tailing off to her personal limit. A different friend (male around 30) is reaching the end of his first year of running and I met him at parkrun while he was stuck at my standard, around 45% for him. One week he suddenly improved by a whole minute (from him seeing me ahead to me seeing him ahead) and now he runs consistently at 63-64%; out of sight. I suggest that this is his current target level and that, unless he changes his routines (training, diet, targeted resource), this percentage is his measure of a 'good' run.

I think it is sufficient to know how your own performance changes, but I appreciate that for many they want to know where they stand in comparison to everyone else. So how does 60% or 75% on age-grading fit into the population? ³ This requires us to have large data sets and we must immediately recognise that the best we can do is sample the population that has chosen to turn out in a timed event, whose data is then shared. I'll repeat that in other words; the people who contribute to the available data is a self-selecting group, so this is not everybody, it is everybody who chooses to run. Even then, these are people who have chosen to run in an event that publishes data. 

A male of 35 at 75% does a 5km in 17:11   and at 60% takes 21:29 [grading calc], the faster time represents just inside the 99th percentile [data table] and 21:29 is in the middle of the 88th percentile. Repeating that in different words, the youngish representative guy at 75% is in the top percentile of those who run, so he might, on average, be among the top five of a run with 500 people in it.      A male of 65 at 75% would take 21:52 and be in the 99.1th percentile, while at 60% would be on 27:20 at in the 92.4th percentile. While both of these are good runs, in the sense that they lie inside the top decile, that 15% difference in performance (speed as an expression of the corresponding world best at that age) represents around ten times the number of folk in front (0.9% vs 7.6% or 1% vs 10%).  This large dataset is from US 5km races, which is not the same as, say, parkrun, nor club runs in the UK, though we're rather assuming that they are very similar.

But the percentile expression doesn't tell you anything about what that means for any particular race, unless the results are themselves graded by age. At 75% this means that at a local level an age-category win is quite possible, but someone at that standard might well already recognise the other runners around that standard. I can testify to this myself; where I finish on age-grading compared to everyone else depends very much on who turns up.  At 60%, much further back in the field, you would recognise only that these people around you are varied in body shape. I have a local aged runner twenty years older than me running at 63-68% (i.e. pretty good) — but he is in the last five or ten in the parkruns we both attend. His standard is high but his age is sufficiently large that 'good' running doesn't equate to 'good' for all the other males who turn up, since he is usually the last male to finish. I say he is fantastic to be still running in his late eighties but the same speed from a 35-year old is really not very good at all. ² 

DJS 20190612

Also see Essay 263 

top pic from wikipedia

1. I stand around 75% on age-grading irrespective of my age at the time. There is surprisingly little change in that, from a peak of 85% such as when running for 360/365 days a year, to 72% when running perhaps twice a week. I only count the runs that meet some internal standard that differs from a 'jog' but includes any attempt to go running. A sub-70% performance represents being (provably, observably, patently) unwell, injury included, and probably happens around once a year.  Turning that around, it also tends to mean that I won't participate in a run that will produce published results if I think I'm going to perform badly, sub 70%.


75% translates to 5km times of around 22:15 at 65 (which proved fairly rare in 2019), 20:30 at 55 (often, at 62), 17:30 at 35 (probably achieved, 4:40 for the 1500m and 2:15 for the 800 occurred often and are close to 75%). At 17 (when I left school) I was doing around the same and 75% would be 4:47 and 2:20 respectively. I used this calculator for these figures.  I note that the parkrun calculator says 22:30 is 75%, but I think that uses a finer age adjustment than the runbundle one linked.

2  At 85, 37:30 is 63.4% on age-grading but at 35 it is 34% for a man and 37% for a woman. I've gone round the two lap park-run course a third time after my post-run chatter in that time. But at least those running at 30% have turned up and had exercise: very few park-runners are at this standard for long.

3 If or when you have run enough to have results recorded on the RunBritain rankings, then off to the right there is a list of event rankings expressed as  a % of those in the UK. I failed to find a precise explanation of what this actually means, but I think it has to be using the population of results, i.e. the available data. So my better times rank around the 10% mark, which I think means I'm achieving around the top 10% of all those providing data at that time (year) and distance. I did discover that I'm not the only person looking for this answer. One respondent observed that he ranked 4% at a parkrun and 32% in a marathon. He concluded that either his marathon is rubbish or that there is a far greater volume of data (if you like, a better class of runner) for the marathon. My own figures support the 'better runner' hypothesis, since in large population events my % ranking drops to more like 13%. For a long time (maybe from 30 to 55) I would go to an event and expect to finish at the 4% mark — if there were 1000 runners, I'd expect to be about 40th. A suitable example was running in CapeTown, a field of about 10,000 and yes, I finished very close to 400th, as expected. With smaller fields, the apparent variation increases. Maths students will think that obvious.

[2] http://www.bigdatarunning.com/tag/5k/

[3] https://www.goodrunguide.co.uk/AgeGrading.asp

Covid            Email: David@Scoins.net      © David Scoins 2021