Rotation speed is a decision problem wearing a tracksuit

Teams that rotate fastest are usually not the fastest teams, they are the ones whose players start moving earlier.

Speed is the wrong word. Two defenses cover the same ground in different amounts of time and the difference is almost never how quickly the players run. It is when they started.

Break a rotation into two parts. There is the latency between the pass being released and the helper's first step, and then there is the travel. Travel varies between players by a modest amount. Latency varies enormously, because latency is a function of whether the defender had already decided what he would do if that pass was thrown. The prepared defender moves on release. The unprepared one moves when the ball arrives, which in practice is a step and a half behind, and no amount of athleticism closes that.

You see it clearly on the second rotation, the one after the first help. First help is instinct and almost everyone does it. The second man, filling the space the first help vacated, is where teams separate, and it is a pure recognition task performed by somebody who cannot see the whole floor.

The practical implication is that you improve rotation speed in the film room and in walkthroughs, not in the weight room. Rules help more than effort does. If every player knows in advance that the corner is filled by the man one pass away and never by the man at the top, latency collapses, because nobody is deciding anything.

Where this gets misdiagnosed is with older teams. A veteran defense is often described as slow because the players are slow, and the eye test agrees, and then you look and find the rotations are actually arriving on time because the decisions are pre-made. Meanwhile a young athletic team gets shredded while running faster than anyone.

The measurement problem is that we can time the travel and not the latency, because latency requires knowing the exact frame at which the pass left the hand and the exact frame at which the helper changed direction. Both are computable from tracking data and neither is published. If I had one defensive statistic released tomorrow, that would be it, and I think several established defensive reputations would not survive it.

Effort gets you to the right place late. That is worth almost nothing.