May 4 — 1-2, moneyline swings both ways after model retrain

Yesterday

1-2 · +0.00u · ROI +0.0% · CLV -1.02%

Boston delivered the night's biggest win behind exactly what we projected - their elite bullpen (1/100 rating) dominated Detroit's struggling relief corps (57/100) in a 5-4 victory that validated our 20% expected value thesis. The Yankees demolished Baltimore 12-1, making Shane Baz's superior 4.27 xFIP irrelevant as our worst loss of the slate. Cincinnati fell just short 5-4 to Chicago despite the Reds' massive bullpen edge (1/100 vs 35/100). With negative CLV on both losses (-0.5% and -1.55%), the slate leaned more lucky than sharp despite identifying the right edges.

What hit

Best result. BOS ML (BOS @ DET) at +200, +1.00u — final 4-5

What missed

Worst result. CIN ML (CIN @ CHC) at +180, -0.50u — final 5-4, CLV -0.5%. Graded bad bet.

Both losses show systematic model failures despite positive expected value. In CIN@CHC, the model badly underestimated Cincinnati's bullpen collapse - predicted 1.0 score vs Chicago's 34.7, yet Chicago's relief allowed the decisive runs after Cabrera's solid 6 IP/3 ER start. The -0.5% CLV suggests fair pricing.

BAL@NYY reveals more severe miscalibration: Shane Baz's 5.67 IP/5 ER massively exceeded his 4.27 xFIP prediction, while the model's 8.58 total runs badly undershot the 13 actual. The -1.55% CLV indicates this was genuinely mispriced....

Under the hood

Yesterday's development focused on expanding betting market coverage and refining prediction accuracy. The team added support for first-five-inning moneylines and totals, plus alternate game totals, giving bettors more wagering options. They also improved the runline betting system by fixing data leaks that could skew predictions and adjusting overlay calculations to be more selective with recommendations.

This week

Last 7 days: 4-3 · +1.81u · ROI +30.2%