PLOT
Points Left On the Table — chess.com's post-game review, for NBA possessionsWatch a possession with an eval bar — the expected points right now (EPV), read off a ten-player tracking model. Each time a wide-open ball-handler passes up the shot, a chess.com-style badge grades the decision by the points left on the table.
This isn't just a pretty meter — we checked it against reality. Across 208 games, declining a good open look really does cost points (about 0.11 on an average look, ~0.19 on a great one); declining a bad look is free; and the cost grows the better the look was. The signal is invisible to the box score.
The honest limits (the same ones in the paper): the mirror move — shooting over a wide-open teammate — does not hold up, because an “open man” is often open precisely because the defense doesn't respect him; the effect is modest; this scores open pass-ups only; it measures left value conditional on what the cameras can see, not proven causal decision quality; and pinning down which players are best or worst at it is weak. Built on 2015-16 SportVU tracking; the eval trace is leakage-free and recalibrated. Code, paper & gate reports →