Print RSS Sign-Up for Insider

Advanced Stats 101: Part One

by Steve Weinman, NBA

For all the words to be written about the value of basketball statistics, pioneering author, statistician, and current Denver Nuggets consultant Dean Oliver cut through all the muck when he said, “Individuals see a game better than numbers. But statistics see all the games.”

That leads to the discussion of which stats are the right stats. Or a part of that discussion, at least. It’s a conversation that remains far from over. But several years of work led by the Association for Professional Basketball Research (the inspiration for the term “APBRmetrics”) have made it evident that, at the absolute minimum, one can do better than per-game measurements of the actions tracked in the traditional box score.

The primary problem with per-game measurements is that not all games are created equal for each team or player. While they may represent equivalent time spans, they do not represent equivalent sets of opportunities. If statistics inform assessments of players and teams with varying numbers of opportunities, the fairest standard of judgment is on how effectively players convert their opportunities.

The word “opportunities” appeared three times in that last paragraph. In basketball, any discussion of opportunities starts with possessions. A possession is simply defined: a chance to score before the defense gets the ball. A game, on the other hand, can be defined by any number of chances to score. In 2009-10, a game played by the aptly named Idaho Stampede featured an average of 102.2 possessions, a D-League season high. A game featuring the Erie BayHawks plodded along at an average of 94.5 possessions. That means that, on average, Idaho had nearly eight more opportunities to score and had to defend its opponents nearly eight more times than Erie did. But when the Stampede and BayHawks – or any other two teams in any game, ever – play against each other, they will each get approximately the same number of possessions. That alone should make it clear how silly it is to evaluate a team’s scoring or rebounding or ability to control the ball or anything else simply based on how many times per game it does certain things. What all this means is that winning is a matter of using possessions most effectively. And if statistics help us evaluate players and teams whose goal is to win, it makes sense to use statistics that best relate to achieving that goal.

The term pace refers to the number of possessions in a game. Rate stats that adjust for pace allow the most accurate evaluations of teams and players. At the highest team level, offensive and defensive efficiency measure points scored and allowed per possession. Because people tend to be more comfortable working with numbers resembling points per game figures, statisticians most commonly use points per 100 possessions. Replacing per-game team offense and defense stats with offensive and defensive efficiency provides a base for measuring teams against each other across the league.

In Part II, we will examine how to use rate stats to measure the components of basketball that determine teams’ overall performance at each end of the floor: shooting from the field, ball control, rebounding and getting to and converting at the free throw line.