ESPN's NBA Basketball Power Index (BPI) is a forward-looking measure of team quality. It uses advanced statistical analysis to measure each team's offensive and defensive levels relative to an average team.
For example, a team that has an offensive rating of 3 and a defensive rating of -1 has a total BPI of 2. The team's offensive rating means that they would be projected to score three more points per 100 possessions against an average defense than an average offense, and their defensive rating of -1 means that they would give up one more point per 100 possessions than an average defense would against an average offense. The total BPI score of 2 means that they would be favored against an average team by two points.
The central input to BPI is the score of every game. These scores by themselves do not tell the entire story, however, so they are adjusted for a variety of factors:
Strength of Opponent: There is a difference between beating the Golden State Warriors and beating the Philadelphia 76ers. BPI factors in the strength of opponent played so that these two games are not treated the same in evaluating a team's ability.
Pace: BPI utilizes the final score as its starting point but then adjusts that score for the number of possessions played. A team that scores 100 points on 80 possessions is a more efficient offensive team than one that scores 100 points on 100 possessions against the same opponent. BPI adjusts the final score of each game for the pace of play so that results can be compared fairly.
Game location: BPI adjusts for home-court advantage.
Distance traveled: The home-court advantage tends to grow as the distance the visiting team has to travel to get there grows, so BPI adjusts for the distance traveled as well as home court.
Rest: Teams often face each other with different amounts of rest between games. Beating a team that is on the second half of a back-to-back is not as impressive as beating a team that has had three or four days to recover since their last game, so BPI adjusts for this as well.
Preseason expectations: Teams often are what we thought they were, so to avoid getting caught up in the natural streakiness of the NBA season, BPI incorporates preseason expectations, based on the Vegas expected win totals and the prior year's performance.
Along with BPI, the same factors as those listed above (site, distance, rest, etc.) can be used to calculate a team's average scoring margin and chances of winning in any given game. Each team's chance to win each remaining game on the schedule is used play out the entire NBA season 10,000 times. From those simulations, the chances of each team making the playoffs, advancing through various rounds of the postseason, finishing in certain parts of the draft order and other such projections can all be calculated.
Using all of these inputs helps make BPI one of the most accurate prediction models available: BPI favorites have won over 72 percent of NBA games. For a review of BPI's performance in 2014-15, see: So, how did BPI do with its projections?