FiveThirtyEight and Inpredictable teamed up to make a nice interactive visualization that shows every NBA team’s chances of winning at every minute of every game, according to Inpredictable’s basketball win probability calculator.
I was curious what the FiveThirtyEight graph would look like if every team didn’t begin every game at 50% win probability, so I took all of the NBA in-game gambling odds from Gambletron 2000 for the 2014-15 regular season, and produced a similar interactive visual. Gamblers of course take into account all available information when determining a team’s win probability, including team quality, injuries, motivation, and anything else they think is relevant:
The x-axis in the above graph is time in the game, and the y-axis is the average win probability for each team at that point of the game, according to in-game gambling odds
Most series in this gambling data graph are much flatter compared to the FiveThirtyEight graph, where every team starts at 50% before fanning out to its final winning percentage. The flatter gambling graph makes sense because gamblers do a pretty good job of figuring out pregame win probabilities – if they didn’t, it’d be easy to make a lot of money gambling, which, spoiler, it isn’t!
Nevertheless, there are some teams that deviate substantially from the expected winning percentages implied by gambling odds. Here’s a scatterplot that shows each team’s expected pregame winning percentage on the x-axis, and actual winning percentage on the y-axis. The teams that are above the diagonal line are the ones that are outperforming gamblers’ expectations, whereas the ones below the diagnoal line are performing worse than gamblers expected (hover over the points to see the data):
The Atlanta Hawks led the league in “wins above gamblers’ expectations”, with an actual winning percentage of 73.2% compared to an expected winning rate of 61.8%. The Houston Rockets and Golden State Warriors have also both performed significantly above gamblers’ expectations. The lowly Minnesota Timberwolves, in addition to having the worst absolute record in the league, are performing the worst relative to gamblers’ expectations. The Timberwolves were expected to win 26.7% of their games, and yet have only managed to win 19.5%.
NFL, MLB data
Since Gambletron 2000 tracks more than just the NBA, I generated the same graphs based on in-game gambling data from the 2014-15 NFL season and the 2014 MLB season. Here are the NFL graphs:
The Arizona Cardinals, Dallas Cowboys, and New England Patriots performed the best relative to gamblers’ expectations, while Tennessee Titans, Tampa Bay Buccaneers, and New Orleans Saints all won significantly fewer games than expected.
The MLB graphs are notable for how much closer together the teams are in expected win probability. In both the NBA and NFL, expected pregame win probabilities range from roughly 25% to 75%, but in baseball all of the teams fall between 40% and 60% pregame win probability:
The Kansas City Royals and Baltimore Orioles outperformed the most, while the Colorado Rockies, Arizona Diamondbacks, and Oakland Athletics fell shortest of gamblers’ expectations. The A’s are also interesting because gamblers gave them the highest expected win probability of any team, and yet they fell well short of expectations.
N.B. the MLB data includes only about 65% of all games because gambling markets are declared invalid if the previously announced pitchers don’t start as expected. Accordingly, the actual winning percentages might not match up to the full 162-game records.