--- slug: "win-probability" title: Win Probability depth_tier: entry side: UNIVERSAL category: Analytics source_id: "ANA-0017" created: 2026-05-13 updated: 2026-05-13 canonical: https://learningfootball.xyz/terms/win-probability aliases: - win probability model - WP - WPA - win probability added - "real-time win probability" --- # Win Probability ## Short Definition A real-time analytical model expressing the likelihood of winning at any game state—based on score, time, down, distance, and field position. ## Definition Win probability is an analytical model that expresses, at any moment during a game, the historical likelihood that a team will win based on the current game state. The inputs typically include score differential, time remaining, down and distance, field position, and which team has possession—variables that collectively describe a game state that can be compared against thousands of historical games in similar situations to produce a probability estimate. The model's value is in converting abstract game situations into comparable, quantifiable assessments. A team leading by seven points with two minutes remaining and the opponent driving is a specific situation that has occurred many times; win probability assigns a number to it based on how often the leading team won from equivalent positions. This gives analysts, broadcasters, and coaches a framework for discussing not just what happened but how likely it was. Win probability added (WPA) is the derived metric that measures the impact of individual plays on a team's win probability. A field goal that shifts win probability from 55% to 72% has a WPA of +0.17 for the kicking team. A fumble that shifts probability from 75% to 40% has a WPA of -0.35. Summing WPA across all plays in a game produces a picture of which players and decisions most influenced the outcome in terms of winning probability, not just raw statistics. The model is most useful for evaluating decision-making in high-leverage situations. Going for it on fourth down, attempting an onside kick, choosing to take a knee or push for more yards—all of these decisions can be evaluated against win probability curves that show the expected value of each choice. Coaches who consistently make decisions that improve their team's win probability, even when those decisions occasionally produce bad outcomes, are making analytically sound choices that compound over a season. ## Example "Win probability at 98% with two minutes left and the ball. They punted instead of running out the clock. Win probability dropped to 91%. A seven-point swing in likelihood from a decision that looked routine. That's what WPA captures that box scores miss." ## Common Confusion Win probability is a model output, not a guarantee. A team with 85% win probability still loses 15% of the time—the model reflects historical rates across similar game states, not certainty about the outcome of this specific game. A team that was favored at 90% and lost did not suffer a "model failure"; they experienced an outcome that happens roughly 10% of the time in equivalent situations. Win probability describes likelihood, not destiny. ## Related - epa - dvoa - fourth-down - analytics - qbr - pff-grade --- Source: https://learningfootball.xyz/terms/win-probability