Celts' Close Game Results Indicate Early Season Fortune

A Daily Babble Production

While the faith here remains strong that the best of the 2008-09 Boston Celtics is yet to come, it is clear that through the season to date, the Celtics played more successful basketball prior to Christmas Day than they have since.  Observers have offered plenty of reasons for the drop-off, including injuries, scheduling that included a couple of West Coast road trips and the natural lull of the middle of the NBA's marathon season.  Here's one more for the pile: chance.

I spent some time recently catching up on the writings of Basketball Prospectus' brilliant Kevin Pelton, who produces some of the most intriguing and comprehensible statistical work on the Web.  In early March, Pelton took an insightful look at team performances in close games.  Using data from the last eight seasons, he mapped out team success in games decided by five points or fewer alongside a variety of factors, including age, offensive and defensive ratings and non-close game win percentage. 

It's that last one that interests me, especially because the other factors exhibited correlations bordering on negligible at best.  But non-close game success maps to close game success in a logical manner.  While teams that are good in general tend to win close games more often than those that aren't, Pelton notes that the difference between them is mitigated with everyone being pulled closer to the .500 mark.

This makes sense.  Just in the way that short playoff series or single-elimination tournaments (see college basketball or the NFL playoffs) tend to provide an edge to inferior teams, so does a close game.  The longer a competition goes, the more likely that both teams will regress to their mean levels of performance.  As Pelton writes, "Say you were coaching (or cheering on, if you prefer) an underdog team playing a powerful opponent. If I gave you the opportunity to advance directly to the final five minutes of a close game, even if you were trailing by a couple of points, you would take that scenario because anything could happen over the shorter period."

Based on the data and Pelton's regression line of best fit, a team that wins only 20 percent of its non-close games projects to win 40 percent of its close games. A team that wins 80 percent of its non-close games is expected to win 60 percent of its close games.

We return now to the Celtics.  In 2007-08, the Celtics went 55-7 in non-close games and 11-9 in close games.  A 55-7 team has an expected winning percentage a bit better than .600 in close games.  While the Celts only played .550 basketball in tight contests last season, the difference between expected and actual performance is nowhere near enough to merit consideration for Pelton's lists of biggest outliers. 

On the other hand, this season's Celtics began the season 27-2 overall, but that breaks down into 20-2 in non-close games and 7-0 in close games.  An non-close winning percentage of .909 predicts a close game success rate around .630.  That leaves a difference of .370 between the actual and predicted performance of the pre-Christmas Celtics in close games.  The 2002 Sacramento Kings hold the single greatest full-season over-achievement in close games over the years studied by Pelton with a difference of .304. 

The Celtics' performance through the first two months of the season was absurd, but it was over a small sample size.  Four teams from this year make Pelton's top five over- and under-achievers lists (compiled March 4), and he intimates that this can be explained by the fact that those figures will moderate as the season continues and sample size increases.

That's exactly what happened with the Celtics.  Starting with the visit to Los Angeles on Christmas Day, the Celtics have won 27 of 44 games overall but have gone only 5-6 in games decided by five points or fewer.  While that represents a bit of under-performance based on Pelton's calculations, it's nowhere near the magnitude to which the team overachieved earlier in the season.

For the season to date, the Celtics are now 42-13 (.764) in non-close games and 12-6 (.667) in close games.  That's still a shade less than .100 better than their predicted performance in close games, but it's a lot more plausible than the early-season results.

Pelton suggests that, as indicated by the .454 correlation between performance in non-close and close games, close games are neither decided completely by luck nor by the so-called true ability of the teams involved.  Superior teams should still perform better in close games, but not to the extent that they do in general.  Since there is no relationship between close game success and such factors as age, pace and offensive and defensive ratings, it seems reasonable that chance is simply part of the equation here in addition.

The Celtics played better basketball earlier in the season, but they benefitted from better fortune as well.