How Play-By-Play Logs Became The Basis Of Analytics

Basketball analysis changed less because of new ideas than because of new access. Once event-level logs became available, questions that had been unanswerable became routine.

A box score hides the sequence

Totals tell you what a player accumulated but not when, against whom, or in what game situation. Two identical stat lines can describe entirely different performances.

Every contextual question — performance in close games, production against particular lineups, efficiency by score margin — requires knowing the order of events.

The play-by-play log preserves that order, with a timestamp, a score state and the players involved for each event.

Lineup data was the first breakthrough

Because substitutions appear in the log, the players on court at any moment can be reconstructed, which allows performance to be attributed to combinations rather than individuals.

This produced the family of plus-minus measures, which ask what happens to the score while a given group is playing.

Those measures have well-known limitations, but they were impossible before the log existed, and their existence reframed how rotations are discussed.

Possession counting made rates comparable

Team statistics per game are distorted by pace, since a fast team generates more of everything. Comparing teams requires knowing how many possessions each used.

Possessions are estimated from the log by counting the events that end them, which is a reconstruction rather than a direct measurement but a stable one.

Expressing production per possession rather than per game removed pace from the comparison, which changed how offenses and defenses are ranked.

Shot location changed the strategy

Attaching coordinates to attempts allowed expected value to be computed by court location, which is how the case for the corner three was established quantitatively.

The finding was not subtle once the data existed, and its adoption across the sport is one of the clearest examples of measurement altering play.

It also demonstrates the general pattern: the analysis followed the data collection rather than motivating it.

What logs still cannot see

Event logs record actions, so players who influence a possession without recording an event are invisible to them entirely.

Tracking data addresses this by recording positions continuously, but access is far more restricted than play-by-play, so public analysis remains event-based.

The shape of what basketball believes about itself therefore follows the shape of what was released, which is a fact about data policy rather than about basketball.

The phrase you hear in selection meetings is that a player is one-dimensional. It is said as a criticism. I have never fully understood why, given that the format is short enough that a team only needs a handful of dimensions in total and can distribute them across eleven bodies.

Consider the batter who can only hit down the ground and only against pace. He is useless in the middle overs against two spinners on a turning surface. He is also, in the right six deliveries, the most valuable player on the field, because the thing he does badly never comes up and the thing he does well comes up repeatedly.

The counter is obvious and it is about flexibility. If a wicket falls at the wrong moment, a specialist walks into a situation he cannot handle, and you have burned a slot. That is real. But it is a scheduling problem, not a talent problem, and scheduling problems have solutions that talent problems do not.

What actually happens is worse than either. Sides pick a squad full of players who are competent in every phase and outstanding in none, then wonder why they keep posting adequate totals and losing to somebody who had one man go berserk for four overs. Competence across the board is a way of guaranteeing you are never the reason a match turns.

I would rather carry two narrow players and accept that in a third of matches one of them is dead weight. The distribution of outcomes matters more than the average, because you do not get half a trophy for finishing consistently mid-table in a knockout competition.

There is a coaching dimension to this that gets ignored. Telling a specialist to broaden his game usually makes him worse at the thing he was picked for, because the technique that lets a player hit a length ball over long-on is not the technique that lets him work it into the leg side. They fight each other. He ends up in the middle, hitting neither shot with conviction, and within a season he is a squad player.

Leave him alone. Let him be strange.

The best selections I have watched were not the most complete cricketers on the list. They were the ones whose weakness the captain had a plan to hide.