DingerStats Is Back — Rebuilt From the Ground Up
We took two years off to build something worth coming back for. Here it is.
The Engine
Every MLB game prediction starts with 10,000 Monte Carlo simulations — not a formula, not a gut feeling, but ten thousand full 9-inning games played out computationally.
25-State Markov Chain Model
Every batter and pitcher has their own transition probability matrix built from 10 seasons of Retrosheet play-by-play data (2016–2025). That's 1.7 million at-bats distilled into individual player models that capture exactly how each player changes game state.
Elo Rating System
Dynamic team strength ratings that update after every game. Momentum, strength of schedule, and the broader context that player-level simulation alone can't see.
Ensemble Blend
The Markov engine handles player matchups. Elo handles team context. Together they're more accurate than either alone.
Park Factors
Every simulation adjusts for the run-scoring environment of the home ballpark. Coors Field and Dodger Stadium produce very different games.
The Results
Validated against 7,289 MLB games across three full seasons (2023–2025):
Fully transparent. Every prediction logged, every result tracked, seasonal accuracy updated in real-time at dingerstats.com/accuracy.
What's Ahead
- Spring training predictions starting this week — lineups are experimental in spring, so treat these as warmups
- Full regular season coverage from Opening Day
- Deep dives into park factors, pitcher matchups, and team trends
- Real-time accuracy tracking at dingerstats.com/accuracy
Math, Not Gut Feelings
10,000 simulations per game. 1.7 million at-bats of training data. Every prediction posted publicly before first pitch. No cherry-picking. No hiding bad days.
Follow along.
—DingerStats