May 13 — 1-1, moneyline swings offset for slight profit

Yesterday

1-1 · +0.11u · ROI +11.0% · CLV +0.43%

Split night at 1-1 (+0.11u) with contrasting bullpen narratives. Atlanta's 56-point bullpen advantage over Chicago materialized perfectly as the Braves pulled away late for a 4-1 victory, validating the 55% model edge. Philadelphia's Andrew Painter couldn't overcome Boston despite his 3.73 xFIP advantage over Sonny Gray's 4.44, falling 3-1 as the Red Sox capitalized early. Both picks showed solid market value with 42-45% CLV, suggesting sharp process despite the mixed results.

What hit

Best result. ATL ML (CHC @ ATL) at +122, +0.61u — final 4-1, CLV +0.4%. Graded good bet.

What missed

Worst result. PHI ML (PHI @ BOS) at +115, -0.50u — final 3-1, CLV +0.4%. Graded unlucky (positive CLV, lost).

The model catastrophically misjudged the total, predicting 8.75 runs versus 4 actual in PHI@BOS. Both starters outperformed xFIP predictions - Gray allowed just 1 ER in 6 IP against a 4.44 xFIP, while Painter gave up 1 ER in 5 IP versus 3.73 xFIP. The away bullpen score of 75.2 suggested strong relief pitching that materialized. With 0.45% CLV, this wasn't a bad bet but reveals systematic total overestimation when both starters exceed expectations simultaneously. Retrain priority: starter performance correlation in low-scoring games.

Under the hood

Yesterday's development focused on fixing the daily data pipeline that processes betting markets. The team corrected how the system handles time-based data filtering and added safeguards to prevent analysis of markets where betting value calculations aren't available. These backend improvements ensure more reliable data processing for the research models.