WAR For Cricket

In baseball, WAR (Wins Above Replacement) is a comprehensive metric to quantify a player’s contribution to team success, relative to a replacement-level player. The idea has slowly migrated to other sports. In cricket, however, WAR remains underdeveloped. Most public evaluations still rely on averages, strike rates, or wickets — useful, but context-blind.

In this post, we take a first step toward building WAR for cricket, using publicly available ball-by-ball data from Cricsheet. We begin with One Day Internationals (ODIs), and separately estimate WAR for batters and bowlers. The goal: capture how many more wins a player contributed than a replacement-level player might have in the same context.

Batting WAR

Batters face different conditions: powerplays, death overs, batting positions, and shifting run environments over time. We use a Generalized Additive Model (GAM) to predict expected runs per ball as a function of:

• Over number
• Batting position
• Year

We then downscale predicted performance to define a replacement level and compute:

WAR_i^{context} = (Actual Runs - α × Expected Runs) / 10

Bowling WAR

For bowling, we flip the logic: better bowlers concede fewer runs per ball.

We compute each bowler’s runs conceded per ball, define replacement level at the 20th percentile (i.e., worst regulars), and calculate:

WAR_i^{bowl} = (R_{rep,i} - R_i) / 10

Where:

• R_i: Runs conceded
• R_{rep,i}: Expected runs at replacement run rate
• Again, using 10 runs ≈ 1 win

We don’t (yet) model bowling context (phase, pitch, matchups), but doing so is a natural next step.

Initial Results

Extensions & Improvements

• Strike Rate Modeling: Batting WAR can better reflect impact if we model not just run totals but how quickly runs were scored.
• Context-Aware Bowling: Just as in batting, not all overs are equal. Modeling bowling context (death overs, match state, opposition strength) would refine estimates.
• Runs per Win: We use 10 runs ≈ 1 win as a constant, but this could be estimated more carefully across time, format, and gender.
• T20s and Tests: WAR can be adapted to other formats — but Tests will require a different frame (e.g., wicket impact or win probabilities).

More at: https://github.com/outside-edge/war