As the NBA season begins tonight in Cleveland, I thought I should breakdown two advanced statistical measures that you may read about it over the course of the year, either via this site or any others.
Stats can be particularly useful in basketball because players tend to be consistent from year to year. Large improvements or failings, statistically speaking, mostly come when a player is either young or aging. Obviously, young players can improve while old players often decline.
Both PER (Player Efficiency Rating) and WP48 (Wins Produced Per 48 Minutes) are what I refer to as “aggregate stats.” The purpose of each is to create a single all-in number that defines a player’s value. As opposed to saying, “In 2008-9, Anderson Varejao averaged 8.6 points, 7.2 rebounds, 1.0 assist, 1.0 turnover, shot 53.6% etc. per game,” we can say, “In 2008-9, Anderson Varejao had a PER of 14.6 and a WP48 of .168” (more on those specific numbers in a second). This also simplifies the conversation when it comes to comparing one player against another using stats.
PER was invented by John Hollinger, who now writes for ESPN. Economics professor Dave Berri is the primary voice behind WP48. He’s co-author of the book The Wages of Wins.
PER has become the most popular advanced statistical measure in basketball, largely because of the machine that pays for it - ESPN. In Hollinger’s own words, PER is a per minute rating of a player’s statistical production. A PER of 15.0 is average. Wilt Chamberlain had the highest one-season PER in NBA history: 31.84 in 1962-3. LeBron’s PER last year - 31.67 - is 4th all-time.
The idea behind WP48 isn’t all that different, although there is one important distinction: WP48 measures the impact a player’s statistical production has on the outcome of his team’s game. WP48 takes the box score statistics, weights them based on how important each is to winning, and then stuffs them all together into a single number. The result is a stat like the one I mentioned up above for Varejao with .100 being average. To determine how many wins a player produces over the course of the season, you would multiply their WP48 times the number of minutes played. WP48 isn’t listed on Basketball-Reference, so I don’t think I can pull up all-time numbers, but for last season Chris Paul rated highest in WP48 with .470 and LeBron was second at .426.
Not surprisingly, the WP48 camp and the PER camp like to argue about which statistic is a more accurate representation of a player’s value. There are obviously discrepancies between the two - Varejao, for instance, rates below average in PER and above average in WP48. Another example would be Michael Beasley, who had a PER last season of 17.29 (above average) and a WP48 of .049 (below average).
Where does this discrepancy come from, you wonder? The clearest difference between how both PER and WP48 are calculated is that PER emphasizes shot creation while WP48 emphasizes shooting efficiency. In other words, a player’s PER goes up with every shot he takes even if he doesn’t make it; his WP48 only increases if his shooting percentage goes up.
This is a key difference in philosophies. What Hollinger is arguing with PER is that a player’s value increases simply because he’s capable of taking a shot.
I don’t totally agree with this argument. Sure, I believe an offense needs players who create their own shots in order to be effective, but I don’t think that every shot taken should be given the same value. Just to cite one example, I saw Ben Gordon take a lot of terrible shots in the post-season last year - shots that, when they didn’t go in, hurt his team. I don’t believe he should be rewarded for that.
Ultimately, I trust in WP48 more than I trust in PER for two reasons:
1) I think that shooting the ball well is more important than simply shooting the ball.
2) WP48 correlates to wins and PER correlates to nothing.
It’s important to note that neither WP48 nor PER factors in defense outside of rebounding, blocks, and steals. This is why Wayne Winston came up with his model for Adjusted +/- and began working for the Dallas Mavericks (through last season).
There’s more to be said about all of this, but I’ll spare you at the moment. If you’re interested, you can read more about both PER and WP48 by searching the web, reading Hollinger’s work on ESPN.com, or what Berri has written on his blog.
I’ll write about Adjusted +/- and Statistical +/- in the near future. The latter was originated by Dan Rosenbaum, who has been on the Cavs payroll since 2004.
TNT be the spot tonight…
(view comments)
