When it comes to controversial topics for an APBRmetrician, it's hard to beat point differential. Sure, per-minute statistics have more than their share of detractors, but few statistical methods have drawn more criticism than John Hollinger's power rankings. Introduced last year, they rely on teams' point differential instead of their win-loss records.
I used the controversy as an excuse to write about the Chicago Bulls' impressive differential and compare various statistical methods for rating teams in terms of their ability to predict performance the following season. There's another, related topic that hasn't really been touched on: how well does point differential predict a team's performance within the course of the season?
The 41-game midway point is an easy mark to use for such a study, and with most teams having already played their 41st game or about to do so within the next few days, now seems as good a time as ever to study the issue. Dougstats.com offers performance broken down by half for past seasons (except, curiously, the 2004-05 campaign). I used the site's numbers to go back through the 2001-02 season and look for teams whose expected record (with one point per game in differential equaling 2.7 wins over the course of an 82-game season) was at least three wins better or worse than their actual record at midseason.
Let's take a look at how those teams performed in the second half, starting with teams who won fewer games than expected based on their differential.
Team Year 1HW 1HE Diff 2HW
BOS 0607 12 17.2 +5.2 12
CHI 0607 23 27.3 +4.3 26
ORL 0102 20 24.0 +4.0 24
MEM 0607 10 14.0 +4.0 12
CHA 0506 11 14.9 +3.9 15
SAS 0607 28 31.4 +3.4 30
NJN 0203 28 31.0 +3.0 21
Average 18.9 22.8 20.0
Sometimes this whole differential thing does work out; not only did San Antonio have a better second half, but they turned it into a championship last spring. The Spurs and the aforementioned Bulls were able to improve their seeding, as predicted by differential, while the Magic was able to work its way into the playoffs.
The weird team here is the 2002-03 Nets, who slumped during the second half of the season despite a strong point differential. New Jersey was just 3-5 in the month of April and only one game above .500 in March, and there weren't any injuries that explain the difference. That didn't stop them from sweeping Boston and Detroit en route to the NBA Finals before San Antonio ended their run in six games.
The issue that creates some problems for this analysis is that below-average teams who are unlucky in the first half of the season tend to see their motivation dissipate. Last year's Celtics and Grizzlies, notably, were clearly in the lottery not long after the midway point of the season, meaning more minutes for young players and a drop-off from the level of play reflected by their differential.
Team Year 1HW 1HE Diff 2HW
MIA 0607 19 14.5 -4.5 25
MIL 0506 21 16.6 -4.4 19
POR 0506 15 11.3 -3.7 6
UTA 0506 20 16.6 -3.4 21
MIL 0102 26 22.7 -3.3 15
LAL 0607 26 22.7 -3.3 16
DAL 0102 29 25.8 -3.2 28
TOR 0304 20 16.9 -3.1 13
UTA 0607 27 23.9 -3.1 24
Average 23.9 20.4 19.5
Based on this small sample size, a weak point differential is a much worse sign at midseason than a strong point differential is a good one. Four of these teams not only dropped off from their first-half pace, but also underachieved relative to their first-half differential, which is why the average second-half win total for the group as a whole is lower than their first-half expected win total.
Last year's Miami Heat team stands out as the biggest counterexample in this study. Instead of sliding backwards, the Heat improved by six games in the second half of the season, with a point differential to match. Clearly, despite the shoulder injury suffered by Dwyane Wade, the Heat was not the same team down the stretch, getting more inspired play from its veteran players.
Of the 16 teams identified in this mini-study, 12 of them saw their second-half performance trend as we'd expect based on their point differential. That's not overwhelming evidence, but it does seem to indicate there is value to looking at point differential at the midway point.
Using the same criteria, six teams qualify this season--three with better point differentials than records, and three who have won more frequently than their point differential would indicate. Let's take a look at them. (Note that I checked all six teams, and none of them are comparable to the Bulls in terms of building their differential with a few lopsided wins, making it less impressive than it appears.)
Utah Jazz (+3.9, 23-18, 26.8 Expected Wins). The Jazz's point differential stands to have the most impact, by far, on the postseason race in the Western Conference. If the season ended today, Utah would be on the outside of the playoffs looking in. However, the Jazz has a better differential than three teams that would make the West playoffs, including the two teams who lead Utah in the Northwest Division (Portland and Denver). Given that the Jazz has moved within a half-game of eighth-seeded Golden State, I'm expecting to see Utah bump one of the current top eight out of the playoffs and potentially even rally to win the Northwest for a second straight year.
Memphis Grizzlies (+3.6, 12-29, 15.6 Expected Wins). Their inability to win a close game has hidden a lot of the progress made by the Grizzlies this season. In games decided by three points or less, Memphis is 1-9. Young team or not, there's no way that trend can continue; expect a few of those close games to go the Grizzlies' way over the second half of the season. Barring a trade involving Pau Gasol, that should translate into a better second-half record and a lot more optimism in Memphis heading into the 2008-09 season with some strong young talent.
Miami Heat (+3.3, 8-32, 11.5 Expected Wins). By no measure has the Heat had a successful first half of the 2007-08 season (except, perhaps, as measured by winning protests to the league office). Miami's point differential hasn't been quite as bad as the team's record, however. Nine of the games in the team's current 14-game losing streak have been decided by seven points or fewer...then again, three have come by 20 or more.
Milwaukee Bucks (-3.6, 16-25, 12.4 Expected Wins). The Bucks got off to a quick 7-4 start to the season. Early point differential can be unreliable, but in this case the fact that six of the seven wins had come by single digits, with three losses by 15-plus points, told an accurate story. Milwaukee was going nowhere. Don't expect the Bucks to hang around the fringe of the playoff race in the East much longer.
Cleveland Cavaliers (-4.5, 22-18, 18.0 Expected Wins). If anyone figures to be a counterexample from this year's group, it's the Cavaliers, who clearly have underachieved during the first half of the season despite beating their point differential so far. As the supporting cast around LeBron James solidifies with Anderson Varejao in the lineup, Cleveland has picked it up, going 10-2 since beating Miami on Christmas Day. Then again, just two of the 10 wins have come by double figures, so the Cavaliers still aren't playing great basketball. Still, last year's Eastern Conference winners have to be better than a .500 team, right?
New Jersey Nets (-5.3, 18-22, 13.1 Expected Wins). The Nets are currently percentage points back of the 19-23 Indiana Pacers for the eighth and final playoff spot in the East. However, New Jersey has been outscored by 5.5 points per game this season, a number which screams "Lottery!" One of the year's most remarkable stats is that the Nets have only managed two double-figure wins all season; only Miami (one) has fewer. New Jersey has gone 6-1 in games decided by three points or fewer. The youngsters up front might mean the Nets are a better team in the second half than in the first half, but I don't see any way that will translate into wins and losses. Look for New Jersey, currently riding a five-game losing streak, to slip out of postseason contention.
Kevin Pelton is an author of Basketball Prospectus.
You can contact Kevin by clicking here or click here to see Kevin's other articles.