top of page

Model Results

This statistical model produces MLB game predictions using player-level performance forecasts from the proprietary Miñoso Projection System, combined with game-specific projections for plate appearances (batters) and innings pitched (pitchers). First implemented for the 2024 MLB season, the model blends key elements of Joe Peta’s WAR-based approach from Trading Bases with enhancements drawn from leading public projection systems (e.g., ZiPS, Steamer, THE BAT). The result is a customized, data-driven framework for estimating win probabilities for each MLB matchup.

Minoso Model V2 (2024-Present)

2024

Selections:

654

Record:

354-300

Win %:

54.1%

P/L (U):

+22.87u

ROI:

2.74%

Avg Odds:

-112

Post Mortem
  • Positives

    • Avoidance of team bias (all 30 teams selected at least 6 games)

    • Avoidance of HFA bias (49.69% of selections were Home teams)

    • 20u+ return

    • ROI holding at close to 3.00% despite 654 samples

  • Negatives/Items to Address

    • Off-season testing showed too low an edge threshold was used (2025 will raise)​

    • Final month performance fall off (-14.43u in September 2024)

      • Model did not deal well with constant lineup shifts (2025 will end in August)​

Minoso Model V1 (2023)

2023

Selections:

856

Record:

408-448

Win %:

P/L (U):

47.7%

+8.57u

ROI:

1.00%

Avg Odds:

+112

Post Mortem
  • Positives

    • Avoidance of team bias (all 30 teams selected at least 7 games)

    • The model consistently identified mispriced teams—such as ARI, STL, CIN, and DET—while effectively avoiding market-driven sentiment trends. It generated +10.56 units selecting LAA and +8.09 units fading them.

  • Negatives/Items to Address

    • ​Off-season testing showed the edge threshold used was too low, raised for 2024

    • Plate appearance & pitching rotation projections alone were not enough to capture impact of player absences on lineups​ (Proprietary WAR esque valuation metrics used for 2024)

    • Model underweighted LIVE Baseruns and consistently miss-valued teams that deviated far from pre-season projections (optimal weighting tested and implemented for 2024)

WAR Model (2019 - 2022)

2022

Selections:

687

Record:

338-349

Win %:

P/L (U):

49.2%

-36.06u

ROI:

-4.15%

Avg Odds:

-105

2021

Selections:

753

Record:

417-336

Win %:

55.4%

P/L (U):

+30.54u

ROI:

3.00%

Avg Odds:

-115

2020 (COVID)

Selections:

126

Record:

Win %:

75-51

59.5%

P/L (U):

+22.48u

ROI:

11.86%

Avg Odds:

-112

2019

Selections:

371

Record:

219-152

Win %:

59.0%

P/L (U):

+78.74u

ROI:

11.88%

Avg Odds:

-113

bottom of page