Three years ago, the Minnesota Timberwolves' p.r. department was looking to make the case for Kevin Garnett's MVP candidacy, something that would make Garnett stand out from his elite peers. They turned to the plus-minus data the team had been keeping, both for its media notes and for then-coach Flip Saunders.
While plus-minus data, which shows how much a team has outscored its opponents or been outscored with a given player on the floor, had been used for years by some coaches, the Timberwolves did something unusual with it -- they subtracted the team's rating (scaled to 48 minutes) with Garnett on the bench from his per-48-minute plus-minus, showing his bottom-line impact on the team: a stunning 22.8 points per 48 minutes.
Even veteran statistical analysts were shocked by the magnitude of Garnett's value to his team. "That would make him the most valuable player this side of Michael Jordan in Space Jam," quipped one APBRmetrician.
Ultimately, Garnett would not win MVP, finishing as runner-up to Tim Duncan for the award he would win the next year (when the Timberwolves were +17.9 with Garnett on the floor). But the statistic invented by the Timberwolves survived and became popularized by 82games.com, rebranded as a player's "Roland Rating" and now known as net plus-minus.
In theory, net plus-minus is the ultimate statistical measure of an MVP, because it speaks directly to a player's value and measures the so-called "intangibles" that aren't accounted for by individual statistics. In practice, it isn't that easy. Net plus-minus is highly dependent on a player's backup and substitution backups. It's also prone to fluctuation -- after back-to-back seasons with net plus-minuses over 20, Garnett was +3.6 last year and is +9.7 so far this year.
Nonetheless, net plus-minus remains a useful measure for evaluating MVP candidates. Take a look at how the net plus-minus leader has faired in MVP voting the last three years and how the MVP has rated:
MVP Plus-Minus Statistics
Given how closely MVP voting has paralleled the net plus-minus rankings, it is enlightening to look at these rankings for this season's most likely winners (through Monday):
Plus-Minus Statistics of 2005-06 MVP Candidates
The leaderboard reflects both some of the strengths and weaknesses of net plus-minus. Carter, Kidd and Prince aren't part of the MVP discussion, and this doesn't necessarily indicate they should be. Both Detroit and New Jersey boast excellent starting lineups with several players who have similarly high net plus-minus ratings. That balance is, in particular, an argument against picking any Detroit players as MVP. Chauncey Billups, enjoying a career year, has emerged as the Pistons' MVP candidate and a serious contender, but Prince, Ben Wallace (+12.2) and Rasheed Wallace (+11.8) have better net plus-minus ratings than Billups (+11.3). (The fifth Detroit starter and fourth All-Star, Richard Hamilton (+5.9) lags behind in this category.)