In MLB the model stays less certain than the market for most of the game
On the segment-clean corpus the model carries MORE uncertainty than the market through innings 1 to 8 (0.976 against 0.9553 bits at inning 1) and only becomes the sharper of the two at inning 9 (0.4181 against 0.4589); their disagreement runs 0.0839 early to 0.2067 late.
Data integrity
This timing measurement is revision 2, rebuilt on the segment-clean corpus (2026-09-17). The stored corpus contains 178 MLB and 27 international soccer games. Individual measurements can use fewer games after eligibility checks; read their own denominators. Revision 1 values are withdrawn and kept in the timing regeneration receipt.
- market_convergence
How to read this figure
Higher entropy means more uncertainty (1.0 bit at a coin-flip, 0.0 at a certainty); watch whether the model line drops below the market line as the game runs.
Continue with the data
Open the interactive analysis: Model-market convergence by game checkpointWhat it means
This outcome-free view tracks two things over game time: how far apart the forecasters sit (gap) and how much uncertainty each still carries (entropy, in bits). On the segment-clean MLB corpus the market resolves uncertainty faster than the model for eight of the ten innings, and the two only cross at inning 9. In soccer they converge instead (gap shrinks by -0.0732) as entropy decays toward zero.
Caveats and confounds
This is a labeled STANDARD_INSTRUMENT with no outcomes attached; entropy decay late in a game is expected, so it describes behavior rather than scoring accuracy. The inning-10 checkpoint rests on 46 ticks, just above the 30-row floor, so read the last point as thin.