John Hollinger's Power Rankings are at it again. Only this time, instead of being angry their team doesn't rate as high as it should, fans in the Pacific Northwest are baffled as to why the Portland Trail Blazers still rank fifth in the NBA, per Hollinger, despite placing no better than 14th in any subjective power ranking.
In part, this reflects the extra emphasis Hollinger's system places on the last 10 games. When a 38-point win over the Phoenix Suns and the 44-point win over the Charlotte Bobcats drop out of this period, the Blazers' Hollinger Ranking will look somewhat different. Still, the issue is hardly unique to Hollinger's system. My schedule-adjusted differential, which makes no adjustment whatsoever for recency, has Portland ranked sixth in the league. Pretty much any system that assumes point differential is more important than wins and losses will have the Blazers as an elite team.
A glance at a graph showing differential against winning percentage (through Saturday) makes the discrepancy clear:
In terms of differential--looking horizontally on the graph--the Blazers rank with Atlanta, Oklahoma City and San Antonio in the NBA's second tier of contenders. Switch to the vertical perspective and winning percentage and all of a sudden Portland is down with Western teams like the L.A. Lakers, Memphis and even Utah who are battling just to make the playoffs.
Generally speaking, there are two reasons why a team's differential will not match its record. By frequently blowing out opponents, a team earns the equivalent of multiple wins in terms of point differential while only improving its record by one win each time. By losing close games, a team will have essentially played even with its opponent without getting any credit for it in the standings.
Usually, when a team outplays its differential, the issue can be traced to one reason or the other. What makes the Blazers unique is they qualify under both criteria.
The combination of the two factors produces a remarkable pair of stats. The Blazers' average margin in wins (+15.7 points per game) ranks third in the league. Again, Portland ranks behind Philadelphia (+17.4) and Chicago (+15.9). At the same time, the average size of Blazers losses (-6.6) is smaller than any other team in the league, just ahead of Minnesota (-6.7). On their own, these numbers aren't particularly meaningful--New Orleans (+14.8) improbably had one of the strongest margins of victory, compiled in just four wins, before beating Utah Monday--but they serve to reinforce what we already know. Portland wins blowouts and loses the close ones.
The more important and interesting question is where the Blazers go from here, and again the two different aspects of their differential must be considered separately. From an intuitive standpoint, blowing out badly overmatched teams isn't a particularly meaningful feat. There is ample statistical evidence that winning blowouts is meaningful, but do a couple of extra points really matter in a 20-point game?
My past research showed that capping differential did not improve its predictive ability. We must be careful when applying history to this most unusual post-lockout season. Because blowouts are so common--and possibly a product of the schedule rather than the teams involved--it's unclear whether they hold the same meaning. However, the graph above shows a fairly typical relationship between differential and winning percentage.
There's no real reason to think this year's schedule should affect performance in close games. There, it is clear Portland's performance will get better. The relationship between record in games decided by five points or fewer and all other games is much smaller so far than in years past, which is to be expected because the sample sizes are still so limited.
I now have data on close games for a full decade, and just one team in that period has fared as poorly in games decided by five points or fewer as this year's Blazers have thus far. The 2009-10 New Jersey Nets went 1-13 (.071). But those Nets were a dreadful team that briefly threatened the record for fewest total wins in a season. My research has found that better teams do tend to win more close games, albeit less decisively than in lopsided games. The other squad to win less than 20 percent of its close games (the 2008-09 Sacramento Kings) was similarly poor. Among 500 teams, the worst record in games decided by five points or fewer belongs to the 2006-07 Indiana Pacers, who won 41 games and reached the playoffs despite losing 22 games by five points or fewer (going 8-22, .267).
So it's clear that Portland will end up doing better in close games. The key matter is how much the Blazers will improve. Portland fans will point out that the team's woes in close games can be connected to the poor play the team has gotten from its point guards. There are two problems with this argument. The first is that decision making, while more stable than outcomes, is affected by sample size as well. A handful of plays loom large in the indictment of Jamal Crawford and Raymond Felton in the clutch.
The larger issue is that similar causal arguments can be used to explain virtually any team's performance in close games. The Lakers are 9-3 this year in games decided by five points or fewer; that's a testament to Kobe Bryant's late-game heroics, right? Yet over the past seven years, the Lakers have been no better and no worse than expected in these situations. The causal stories fail to hold up over larger samples.
As we saw in applying Thinking, Fast and Slow to the NBA, human reasoning doesn't like outcomes without explanations, so we construct stories to explain what is nothing more than randomness. This conclusion isn't as interesting or as fun as pointing fingers, but at this point the most likely culprit in the Blazers' close losses remains statistical noise.
Kevin Pelton is an author of Basketball Prospectus.
You can contact Kevin by clicking here or click here to see Kevin's other articles.