The statistical techniques broadly known as sabermetrics that allow managers in baseball to rigorously examine every player’s contribution in every at-bat have changed how the game is played, but developing similar techniques for other sports has proven difficult. Football requires subjective judgments about whether a player executed their responsibility on the play. Basketball is a much more fluid game than both football and baseball. It is harder to separate one moment in the game from the next for analysis or to evaluate two players individually when they are working together (in a screen, for example).
A group of Harvard researchers have taken on the challenge of improving statistical analysis in basketball, encoding 93 gigabytes of data from game tapes and looking for patterns. They estimate the probability of a basket at any moment during a possession based on a player’s scoring percentages, the position of everyone else on the court and other factors. These estimates could allow coaches to make more informed decisions when they call plays and to assess each player’s value to the team more accurately.
This is a promising development for basketball, since it could mean that the players who are crucial to a team but rarely score points get more credit for their hard work. By setting the perfect pick, one player increases the chances that a teammate will drain a three-pointer. With more sophisticated statistics, it is clearer why both players are necessary on the play — and that makes the game more democratic and less consumed by ego and celebrity.
The chart above shows how these probabilities might have shifted at one moment during the game between San Antonio and Oklahoma City last season. The colors of the arrows show whether a play will increase the likelihood of a basket, while the weights show the probability that the play will succeed. A pass from Kawhi Leonard across the court to Tony Parker might get knocked away and wouldn’t do much for the offense. The right move, shown above with a bold red arrow, would be for Leonard to give the ball to Danny Green.
The researchers will present their findings at a conference in Boston this month, but you can also read an essay by one of them about their work by clicking below.
Read more: http://knowmore.washingtonpost.com/2014/02/10/the-data-driven-nba-offense/
