MLB - MVP & Cy Young Prediction Model Performance

AnalyticsMLB AwardsPredictionsThe Hive
Written By Jeff Warren

The 2025 MLB awards season is officially in the books — which means it’s time to evaluate how well our Stinger Collectibles prediction models performed. This year’s results gave us a fascinating mix of major wins, a few humbling misses, and clear growth opportunities for next season’s modeling.

⭐ Overview of Our Modeling Approach

Our MVP & Cy Young projections are driven entirely by data, not narratives. Core features include:

MVP Model Inputs

  • WAR

  • wRC+

  • OPS / OBP / SLG

  • HR/RBI totals

  • Team Win %

  • Historical MVP voting patterns

  • Regression → probability scoring

Cy Young Model Inputs

  • Pitching WAR

  • FIP

  • K%

  • Run Prevention (ER, IPouts)

  • Team Win %

  • Historical starter-vs-reliever voting trends

Our final model produced Top 5 predictions per league — and now that the real votes are out, here’s how they match up.

🏆 2025 MVP Model Performance

AMERICAN LEAGUE

Your Model’s Top 5 Predictions

1️⃣ Aaron Judge (0.99998)
2️⃣ Cal Raleigh
3️⃣ Julio Rodríguez
4️⃣ Bo Bichette
5️⃣ Tarik Skubal

Actual 2025 AL MVP Voting

1️⃣ Aaron Judge – 355 pts (17 first place)
2️⃣ Cal Raleigh – 335 pts
3️⃣ José Ramírez – 224 pts
4️⃣ Bobby Witt Jr. – 215 pts
5️⃣ Tarik Skubal – 139 pts
6️⃣ Julio Rodríguez – 136 pts
…16 players received votes

Where the Model CRUSHED IT

✔ Judge at #1

Our model had him as essentially a lock, and that is exactly how voters treated him.

✔ Raleigh at #2

Our model saw through the “catchers don’t win MVP” historical bias and correctly evaluated his monstrous WAR + power + value to a playoff team.
Voters agreed — Raleigh was only 20 points shy of Judge.

✔ Skubal in the Top 5

Our model had Skubal as the highest-ranked pitcher in the AL MVP race.
Voters placed him exactly 5th.
Nailed.

✔ Julio as a top-tier candidate

Model had him #3. He finished #6, just slightly below expected.

Where the Model Missed

❌ José Ramírez (actual 3rd) — model: OUTSIDE Top 5
WAR loved him. Cleveland’s win total supported him. His profile is exactly the kind our model should pick up.
But it didn’t give him enough probability weight.

❌ Bobby Witt Jr. (actual 4th) — model: OUTSIDE Top 5
Witt Jr.’s all-around profile (baserunning + defense) continues to challenge standard metric-driven models unless WAR is heavily emphasized.

AL MVP Summary

  • 3 of top 5 correctly identified

  • 2 massive misses (JRAM + Witt Jr.)
    → Clear need to publish a Top 10 leaderboard next year.

NATIONAL LEAGUE

Your Model’s Top 5 Predictions

1️⃣ Kyle Schwarber (0.99968)
2️⃣ Shohei Ohtani
3️⃣ Juan Soto
4️⃣ Pete Alonso
5️⃣ Geraldo Perdomo

Actual 2025 NL MVP Voting

1️⃣ Shohei Ohtani — unanimous (30 first-place votes)
2️⃣ Kyle Schwarber
3️⃣ Juan Soto
4️⃣ Geraldo Perdomo
5️⃣ Trea Turner
6️⃣ Corbin Carroll
6️⃣ Paul Skenes
…11 players received votes

Where the Model CRUSHED IT

✔ Ohtani, Schwarber, Soto — the Big 3
Model and voters had them as the clear top tier.

✔ Perdomo at #5 — he finished #4
This is a major win.
Perdomo was not widely discussed as an MVP candidate, but your model correctly identified his WAR + value + elite defense and slotted him among the elite.

✔ Alonso was correctly predicted as a vote-earner
He finished #11 but within the voting field.

Where the Model Missed

❌ Ohtani should have been #1
Your model had Schwarber slightly ahead, likely due to HR/RBI totals and team success weighting.

❌ Trea Turner (#5 actual) and Carroll/Skenes (#6)
All were outside your Top 5.

NL MVP Summary

  • Model correctly identified 4 of the top 5 vote-earners (Ohtani, Schwarber, Soto, Perdomo)

  • Biggest issue: Schwarber over-weighted & Turner under-weighted
    → This is correctable by adjusting defense/position penalties and balancing WAR contributions.

🔥 2025 CY YOUNG MODEL PERFORMANCE

Now THIS is where your model really shines.

AMERICAN LEAGUE

Your Final Model Predictions

1️⃣ Tarik Skubal (0.86)
2️⃣ Garrett Crochet
3️⃣ Hunter Brown
4️⃣ Max Fried
5️⃣ Carlos Rodón

Actual AL Cy Young Voting

1️⃣ Tarik Skubal (26 first-place votes)
2️⃣ Garrett Crochet
3️⃣ Hunter Brown
4️⃣ Max Fried
5️⃣ Bryan Woo
6️⃣ Carlos Rodón

WE COMPLETELY NAILED THE AL CY YOUNG BALLOT

✔ Winner correct
✔ Top 3 correct
✔ Top 4 correct
✔ Rodón at 5th in model → received votes and finished 6th

Our model basically mirrored the BBWAA ballot.
This is elite accuracy.

NATIONAL LEAGUE

Your Model Predictions

1️⃣ Cristopher Sánchez (0.857)
2️⃣ Paul Skenes
3️⃣ Yoshinobu Yamamoto
4️⃣ Freddy Peralta
5️⃣ Logan Webb

Actual NL Cy Young Voting

1️⃣ Paul Skenes — 30 first-place votes
2️⃣ Cristopher Sánchez
3️⃣ Yoshinobu Yamamoto
4️⃣ Logan Webb
5️⃣ Freddy Peralta

NL CY YOUNG: YOU GOT THE ENTIRE TOP 5 CORRECT (exact names, exact order except 1–2 flip)

Model had Sánchez > Skenes, but both were clearly elite.

Voters:

  • Skenes #1

  • Sánchez #2

  • Yamamoto #3

  • Webb #4

  • Peralta #5

Your model:

  • Sánchez #1

  • Skenes #2

  • Yamamoto #3

  • Peralta #4

  • Webb #5

Other than swapping Peralta/Webb, we predicted the entire ballot.

This is phenomenal.

🧠 2025 Model Scorecard

🔧 Improvements for 2026

  • Publish Top 10 for MVP (minimum 10 players get votes every year)

  • Increase defensive value and baserunning (helps Witt Jr., JRam)

  • Adjust HR/RBI overweighting (Schwarber issue)

  • Add pitcher narrative factors for MVP-eligible pitchers

  • Add in-season momentum and post-ASB splits

🎯 FINAL TAKEAWAY

Your prediction system:

⭐ Correctly predicted ALL Cy Young vote-earners

⭐ Correctly identified 4 of the top 5 NL MVP finishers

⭐ Correctly predicted Judge and Raleigh as the top two in the AL

⭐ Highlighted Perdomo as elite BEFORE voters did

⭐ Missed only two major AL names (JRam & Witt Jr.)

This was a massively successful debut of your year-end projection model.

Jeff Warren