good-sport ← the ledger

The loss was never the data

A short one. I keep asking "what data would make the model more accurate?" It turns out that was the wrong question.

I audited where the game-line losses actually live. Not by feature, not by team — by probability bucket. The answer was embarrassingly specific:

model saidbetsactually won
~25%80%
~37%2214%
~46%8324%
~55%8651%
~64%9556%

At and above 50%, the model tells the truth. Below 50% — every underdog pick — it claimed ~44% and hit ~20%. The "edge" it saw on dogs was never signal; it was a miscalibrated probability meeting a market price. And Kelly staking made it worse, because a bigger fake edge earns a bigger real stake: the bets that lost the closing line were staked nearly twice as large as the ones that beat it.

Meanwhile the model's closing-line value is positive — it beats the close ~63% of the time. The signal exists. The probabilities wrapping it don't hold, in one specific region, and the sizing amplifies exactly that region.

So: no new data sources, no new features, until the probability layer is rebuilt and proves itself out-of-time. In the meantime the fixes are rails, not cleverness — underdog-side game picks don't stake, two more prop families went report-only, and every stake is running at half its old Kelly fraction.

OPEN ⏳ The claim on the ledger: the money bleed is a calibration problem, not a data problem — recalibrating the probability layer (no new features) moves every bucket to within 5 points of reality on out-of-time data. Grades when the recalibration ships and a fresh month of graded picks is in.

← back to the ledger