SD at ATL MLB Model Forecast Recap: SD LOSS

The result, without the rewrite

The game finished SD 5, ATL 6. The archived model calculation was SD moneyline at +225, and the experimental ledger graded it LOSS for -1.00 hypothetical flat-stake units.

The selected side lost. The recap keeps the original probability and price intact instead of inventing a cleaner story after the final out.

What the original read actually said

Before first pitch, the model assigned SD a 31.1% win probability and compared it with a 30.8% break-even bar at the captured price. The grade judges that exact claim, not a postgame substitute.

The largest supporting model contributions were Elo home-win baseline (+0.14 percentage points), Regressed season form (+0.11 percentage points).

The largest opposing model contributions were Long-run team strength (-0.12 percentage points), Rolling team value (-0.03 percentage points), Rolling run prevention (-0.02 percentage points).

What enters the learning ledger

For this decision, the Brier score was 0.0965 and log loss was 0.3721. Those measurements become useful across a real sample, where calibration can be compared by probability band, checkpoint, and model version.

One final score does not directly rewrite model weights. This receipt joins the rolling error ledger, where repeated patterns can nominate a challenger model for prospective testing instead of triggering a narrative-driven adjustment.

Record boundary

This recap belongs only to the prospective experimental forecast record. It does not claim a proven betting edge, does not represent an official wager, and cannot alter the official ATS Killers archive.