In our constant quest to keep up with cool things that are already out there for other sports, we're going to be beta-testing a new in-game win probability feature. This already exists for other sports -- such as college basketball, pro football, and the NBA -- but we're giving it a shot ourselves. We'll delve into the details later on, but for now the high-level view is that we base the probabilities on a combination of predicted efficiencies and the efficiencies observed in the current game. As the game progresses we slowly phase out the prediction and replace it with the current game's data.
We're labeling this as "beta" because there are certain things we don't take into account, including possession, current pace, and field position. We only use the score and time remaining. We still need to do some more rigorous analysis to see how valid our approach is, but (a) they pass the "eyeball test" of plausibility, and (b) they look pretty cool.
Below we provide a snapshot of all games from this past Saturday. The graphs are specifically chosen during in-game moments with slightly more than five minutes left in each game.
In this game, Temple started out as heavy favorites over Army and jumped out to a quick lead. The vertical grid marks indicate the end of each quarter, while the horizontal ones show how much the game is leaning towards each team. As Temple continues to rack up scores the game slowly swings in their favor. Army grabs two quick touchdowns in the second half and manages to bump the game slightly in their favor, but with five minutes left and a 27-point lead this one is pretty much a lock.
UCF starts out as 80% favorites in this one and score a quick TD to maintain their expected win probability. ECU rings up 17 quick points to make it a game, but UCF ties it up going into halftime. Since ECU is doing better than expected, even though the teams are tied after two quarters, the probability has swung slightly towards the Pirates. ECU scores a TD coming out of the half, making them the favorites for the first time in the game. UCF ties it up just before the end of the third quarter, and as time passes the odds begin regressing towards 50%: overtime. ECU gets another score to go up by a TD with five minutes left. With time winding down and ECU in the lead, the odds begin to dwindle for UCF ...
All remaining charts after the jump.