May 16, 2026
May 16 — 2-1, moneyline swings delivered profit
Strong 2-1 night banking 0.62 units as San Diego's bullpen advantage carried them past Seattle 7-4, while Pittsburgh disappointed as the day's worst performer.
Yesterday
2-1 · +0.62u · ROI +41.3% · CLV +0.91%
Strong 2-1 night banking 0.62 units as bullpen advantages proved decisive. San Diego's massive 32-point bullpen edge (76 vs 44) carried them past Seattle 7-4, with Buehler's superior 3.77 xFIP holding up against Gilbert's 4.06. Milwaukee's 77-point bullpen advantage over Minnesota sealed a tight 2-1 win behind Henderson's quality start. Pittsburgh disappointed 0-6 despite the defensive edge, as Sánchez dominated with his 3.71 xFIP. Mixed CLV signals suggest solid process with some market timing luck.
What hit
Best result. SDP ML (SDP @ SEA) at +140, +0.70u — final 4-7, CLV -0.9%. Graded got lucky (negative CLV, won).
What missed
Worst result. PIT ML (PHI @ PIT) at +154, -0.50u — final 0-6, CLV +1.6%. Graded unlucky (positive CLV, lost).
The model's systematic weakness emerged in bullpen evaluation: predicted Pittsburgh's pen at 37.9 vs Philadelphia's 75.5, yet Pittsburgh's relief was perfect while Chandler lasted only 3 innings with 4 ER. The away pitcher prediction proved accurate - Sánchez's 9 IP shutout aligned with his superior 3.71 xFIP vs Chandler's 4.06. Model underestimated total runs (7.38 vs 6 actual) but this loss carried positive CLV at 1.61%, indicating sound process despite outcome. Retrain priority: bullpen scoring reliability over starter projections.
Under the hood
Yesterday's development focused on system reliability and data accuracy fixes. The team resolved issues with odds tracking to ensure consistent baseline measurements, improved error handling when external data sources fail or rate-limit requests, and fixed several edge cases in the prediction pipeline. They also strengthened automated testing to catch feature inconsistencies earlier and updated the web interface to refresh more efficiently after new picks are published.