What is this?
Baseball Game Similarity searches 150 years of baseball history to find the most similar game β or season β to what you're watching right now. Think of it as Shazam for baseball games.
The Three Tools
Now Playing β Picks up a live game and finds the 5 most similar historical games through the same inning. In early innings, the lineup is factored in. By inning 5, the actual scoring pattern takes over.
Today's Matchups β Before first pitch, compares the announced lineup against 394,000 historical lineup matchups to find the most similar games ever played.
Season Pace β Pick any team and find which historical seasons had the same record and run differential at this point in the year. Shows what happened to those teams for the rest of the season.
How Similarity Works
Each game is converted into a 100-dimensional vector β a list of 100 numbers capturing everything measurable about it:
- Score by inning β runs scored each half-inning (30 dims)
- Game events β home runs, strikeouts, walks, stolen bases, errors and more (30 dims)
- Game flow β lead changes, largest lead, comebacks, ties (10 dims)
- Special flags β no-hitter, walk-off, extra innings, shutout (20 dims)
- Park identity β which team was home (1 dim)
Similarity is the Euclidean distance between two vectors. 100% = identical, 0% = maximally different. In practice, 75%+ is a very strong match.
Early Innings: The Lineup Blend
A 0-0 game in the first inning looks identical to every other scoreless first inning. To make early-inning results more meaningful, the live tool blends in a lineup similarity score based on each team's batters and starting pitcher. The blend fades as the game develops:
- Inning 1: 100% lineup, 0% game state
- Inning 2: 75% lineup, 25% game state
- Inning 3: 50% / 50%
- Inning 4: 25% lineup, 75% game state
- Inning 5+: 100% game state
Data Sources
- Retrosheet β MLB box scores and play-by-play, 1871β2014
- Baseball Savant (Statcast) β pitch-level data, 2015βpresent
- Retrosheet Negro Leagues β official Negro League records, 1903β1962
- Lahman Database β season-level team stats for the Season Pace tool
- FanGraphs / pybaseball β player stats and WAR for lineup vectors