Paula Radcliffe's Marathon World Record

photo credit: PA

Today (April 13th) is the 13th Anniversary of Paula Radcliffe's marathon world record of 2:15:25 set at the 2003 London Marathon. It was one of the greatest marathon performances of all-time. It is a record that no one has come close to since and no one appears close to being able to break in the near future.  In this blog I want to share some of the reasons why I think that is and what it will likely take for someone to break this record (without PED's) in the future.

All distance runners have what I term a "predisposition" in their running.  This is simply our physiological and psychological make-up that makes us better suited for certain race distances.  If you have a long distance predisposition for example, you'll likely be better at the longer races (HM and marathon) than you will be the shorter races (5k-10k), and vice versa, if you have a shorter distance predisposition then you'll be better at the shorter races than the longer ones.  This predisposition is determined by our physiology, things like bone structure, muscle fiber make-up and musculature, and psychological make-up and personality, which includes the type of discomfort we tolerate best and they types of challenges we can most easily embrace.

It is helpful to look at the predisposition of runners in terms of a bell shape curve with the majority of runners being at or near a neutral predisposition, and lesser numbers as we go further away from the mean. To put this in terms we can use in running, my studies have shown that most distance runners (i.e the mean) slow down in pace an average of 4.5% each time the distance doubles between 5k and the marathon. This means their 10k pace is 4.5% slower than their best 5k pace, and their 20k pace is 4.5% slower than their 10k pace.  This assumes they are equally as well prepared and trained for each race distance and on similar course and conditions.

Statistics rules say that 68% of people will fall within 1 standard deviation of a mean.  In our example I think 1 standard deviation is roughly 0.3% meaning that 68% percent of all runners will likely have a slowing rate of between 4.2% and 4.8% when equally as well prepared for each race distance.  And 95% of all people would fall within 2 standard deviations of a mean so that means 95% of all distance runners would fall between a 3.9% and 5.1% rate of slowing when equally as well prepared for a each race distance. Often the remaining 5% of the population are considered an outlayer, but even then all but a few (i.e 99.7%) will be within 3 standard deviation of the mean.

If you look at Paula Radcliffe's personal records for all distances between 5k and the marathon what you see if is that she slows pretty uniformly at a rate of 3.5 - 3.6% each time the distance doubles. This makes Paula's predisposition a very strong long distance predisposition or about 3 standard deviations from the mean.  This is exactly the type of athlete you would expect to hold the world record in an endurance race, someone who is an outlayer and has uncommon (i.e 3 standard deviations out) suitability for the event.

Edited Note (4/14):  there has been questions as to my math on this, so her it is in a nut shell: 
5k PR:  14:29
10k PR:  30:01 - a 3.6% slowing in pace as the distance doubled
Half Marathon:  1:05:40  - a 3.5% slowing in pace as the distance doubles (adjusting and calculating out for the extra 1.0975 kilometers) 
Using the 3.5% rate of slowing that would predict a marathon time of 2:15:55 - Paula ran slightly (30 seconds) quicker with a 2:15:25.  The 30 second differential is certain within the tight realm of performance fluctuations or could be attributed to her having male pacers in the marathon and not having them in her other races.  Using 4.5% as the mean and 0.3% as the standard deviation then 3 standard deviations from the mean would be 3.6% which is roughly where Paula's demonstrated predisposition over the course of her career is.  This makes her predisposition likely to be only had by ~1% of the running population. 

I believe there are 2 main reasons why we have not seen any runners come close to breaking Paula's record (even runners who later proved to be PED users).  There are very few runners (probably less than 1%) who have a 3 standard deviation predisposition towards the longer races, and I think it will take a runner with that type of predisposition (similar to what Paula has) in order to approach her record.  And even if a runner has this type of predisposition they have to still be willing to develop their shorter distance skills in order to fully get the benefits in the longer races.  Given that Paula was so predisposed to the longer races, what is more astounding to me is not that she could run the marathon that fast but rather that she was dedicated to and developed her skills at the shorter races (5k/10k) that she was not as well suited for.  But subsequently when she did turn to the longer races she was astonishing and broke records by large margins. We will not likely see her marathon record eclipsed until we find a runner with a similar predisposition, talent and dedication and that will be a very rare find indeed.

Also it is fair to point out that the longer the race is, the greater the chances for things to go wrong and the less opportunity the athlete has to race that distance.  Even if the predisposition is there, to get a marathon on par with your what your 10k or half marathon time suggests for your predisposition is harder to come by and less likely given the fewer number of chances and longer period of time for things to go wrong.

Side Note:  Incidentally, I think that Paula's clear and consistent predisposition across all distance (consistent slowing rate from 5k to marathon) over the course of her career is one of the greatest unbiased arguments for her being a clean athlete (i.e her performances were not likely aided by Performance Enhancing Drugs - "PEDs").  PED usage would tend to help certain events more than others because of different limiting factors. It would statistically be expected that a PED user would have an inconsistent rate of slowing between events slanted towards the events that the PED help them most at.  I know of no PEDs that would help a runner uniformly across all race distance 5k to the marathon