Ranked picks with source labels.

Model has BAL 58.6% vs market's 48.3% (CLE -112), a major disagreement. Rogers' 0.00 ERA in 7.0 IP is tiny sample noise. Cantillo's 3.2 IP sample e...

NYY -147 and model 58.1% align well. KC's .518 OPS over 21 games is brutal; NYY's 4.9 RPG offense should handle Ragans (9.00 ERA, 2.50 WHIP in 4 IP...

Model likes HOU 58% but STL has superior offense (4.5 RPG, .812 OPS vs 3.8 RPG, .695 OPS) and better record. Burrows' 7.94 ERA in 5.2 IP is alarmin...

Assad's pitcher data is completely missing (fallback mode), making this a structural blind spot. NYM's 1.8 RPG over 21 games is historically extrem...

Model at 58.1% vs market 57.0% is near-consensus on SEA, but offense_defense_mismatch flag fires (SEA 9-13, worse record). T-Mobile's 0.855 park fa...

Model picks PHI (57.5%) over market (52.6%) despite ATL's superior offense (6.7 vs 3.1 RPG, .789 vs .57 OPS). Painter has 2 starts with no usable s...

Model likes CIN at 57.4% but market has MIN at 53.1%. Both pitchers show 6.75 ERA in just 4.0 IP—extreme small sample noise. MIN offense at 5.4 RPG...

Market has LAD at 73% implied vs model's 56%—market is right here. Coors Field (PF 1.47) is driving the 11.5 total; model's 9.34 is a structural mi...

Model projects 9.14 runs vs market 7.0, a 2+ run gap flagging structural miss. Ace matchup (Crochet 0.00 ERA, Valdez 1.50 ERA in 6 IP each) support...

Model has MIA 53.5% vs market 47.4%, disagrees by 6pp on a coin flip. Misiorowski's 19.8 K/9 in 5 IP is tiny sample noise but he's clearly dominant...

McClanahan's stats are completely missing despite 10 starts; Keller shows 6.0 IP over 10 starts (0.6 IP/GS) which is impossible. Data integrity is...

Model favors LAA at 57.5% but market leans SDP (51.4% implied). SDP has better record (14-7 vs 11-11) and King's 0.00 ERA in 5.0 IP (small sample)...