[westward]

Differences in wall-to-wall distances at middle and longer distances are less important than in shorter distances. The article uses a 100m race. But consider a 50m race. 3 cm over 5,000 cm is a .06% difference. That's easily 1/100 of a second in a 24 second race.

But in longer races, there are many other factors that add noise to this. For one, swimmers aren't competing in a pool where one has to swim 1,500m and another 1,499m pool. They are making turns. A 3cm different is well within the 'noise' of a flip turn.

Also, there is drafting and bow waves, and other things going on that have a bigger effect than the 3cm per length.

[kevincox]

I don't think the length matters because you do the same lane multiple times. So the error should always be a consistent fraction of the distance.