Jake Lawrence · Legibility / Decision Analysis / Sport Statistics · Field Data theme
Nine hundred and two of my own games say I have been making the same unexamined decision since 2010, and losing about half a point a time for it. Play the position and the record answers back.
A decision audit of one player's 902 games as Black in the Italian Game, 2010 to 2026, turned on the author rather than an institution. The archive scores 46.3 percent, the worst of any opening he plays often, and 95.1 percent of those games contain the same third move, a move that is not theory and was never chosen so much as absorbed. Aggregating the whole archive to ply 16 finds four decision points where the most-played move measurably trails another move from the same archive, ranked by a Wilson lower bound so a short lucky run cannot outrank a long record, and four lines that lose on their own, the sharpest being fifteen games reaching the same queen check for zero points across twelve years. The score falls as the opposition rises (52.1 percent under 1000, 28.0 percent at 1400 to 1600), which is the shape of a system that works until it is tested. The signature interaction is a playable trainer: the position appears, you move, and a mistake is not marked wrong but invoiced, with the games it already cost you. Aggregates only, no opponent rosters. The three confounds are published beside the findings, and the engine evaluations are labeled as a frozen measurement rather than a computation, because Stockfish is not bit-reproducible and pretending otherwise would be the failure this program exists to name.
The position paper argues that classification systems become infrastructure by going unexamined. This is the same claim tested on its own author: a move played 775 times without ever being chosen, and the cost of that invisibility measured in half-points.
The Legibility Gap measures what a state cannot see about a recovery. This measures what a player cannot see about his own decisions. Both find the gap is not in the data but in whether anyone ever aggregated it.
Back to the release table · See it on the network · Machine-readable catalog
I design and ship AI tools, full-stack apps, and data pipelines — end to end, to production. Tell me the problem in a sentence; I'll give you an honest read on fit within a day.
Work with me →