No short cuts for Ms Bushell
Lynn Bushell ran every mile of the 2017 London marathon, and did so in a quite extraordinary way.
She passed each 5km checkpoint along the course almost exactly 30 minutes and 12 seconds after the last one, only deviating by a few seconds either way during each segment.
She had a plan for the race, and stuck to it religiously. It must have taken immense stamina and self-discipline, but her efforts won her a very respectable overall time of four hours and 14 minutes, and she took 827th place in her age and gender category.
Not one of the 41,300 other runners who completed the course that day, not even the elite runners, ran as consistently as her. Here is her average speed plotted on a chart.
Ms Bushell's race data is not missing any split-times from any of the checkpoints set up along the route to measure runners' progress, but virtually all marathon runners who do miss checkpoints do so for innocent reasons, either because of a technical failure or they somehow ran to the side of the sensors without cutting any distance off the course.
As an example of this, here's a chart showing the speed of another extremely consistent runner who missed a few checkpoints, which are highlighted in red.
But here is the race data for three other runners.
There is one obvious explanation for how somebody whose pace is flagging by the half-way point can suddenly seem to be running faster than the race leaders.
As the map shows, the London marathon route doubles back on itself in a couple of places.
A few runners may have found the temptation overwhelming to skip over the barrier separating traffic in one direction from the other and skip out much of the course.
The biggest short cut was taken by a runner who appears to have turned left immediately after Tower Bridge, cutting about nine miles or 35% from the actual route - their average speeds between checkpoints are shown in the first chart of inconsistent times above.
But the most popular place to cut the course appears to be further along, in the Docklands. Five runners seem to have dodged a loop measuring about three miles by taking a short cut near Heron Quays.
In all, my analysis of the marathon data turned up 11 questionable sets of split times, a tiny fraction of the tens of thousands of runners who completed the full course. It is possible that the runners involved all told their friends and family that they did not plan to run the full race, or maybe they were suffering from an underlying injury (although 23 miles on a dodgy knee is not much better than 26), or maybe some major (and very rare) technical malfunction is to blame for the discrepancies.
Over two weeks after the race, only two of those questionable sets of results have been removed from the official race data. In a marginal but meaningful way, the continuing presence of the others undermines the phenomenal achievement of Ms Bushell and her unimpeachable peers.
Not one of the three inconsistent sets of times shown on the charts above has been deleted.