Beyond the record: what can we make of early team performances?
It’s only been five games but, in a sense, these are the games that we learn the most about what to expect from the team. The most pleasant surprise has certainly been Richard Jefferson. After a productive preseason, Dejuan Blair has struggled early and the jury is still out on Tiago Splitter and Tony Parker. Obviously, we shouldn’t make too much of this limited sample, but which details warrant more or less attention?
3-point percentage is one of the most important factors to winning any basketball game. The team that shoots the higher percentage from beyond the arc wins about 70% of the time. However, 3-point percentage is the least reliable box score statistic from game to game.
The typical team shoots 36% from deep, but 10% of the time teams shoot better than 50% and nearly 12% of games teams shoot less than 20% from beyond the arc. This suggests we should be especially wary when we see that the Lakers are shooting 45% from deep or that the Thunder’s 3P% is just 19%.
(As a side note, the high variability of 3-point accuracy has the effect of bringing winning percentages closer to average. Taking more 3-pointers against better teams, or when trailing, increases winning percentage more than attempting 3-pointers against worse teams or when leading.)
Opponent Free Throw Shooting
Besides 3-point shooting, it is easy to see that statistics such as opponent free throw percentage are largely a product of luck (meaning that a team can do little to control how accurate their opponents are from the line.) However, this statistic has a real impact on wins and losses.
Surprisingly, opponent free throw percentage does not appear to be entirely attributable to luck. Teams that commit many technical fouls should be expected to have higher opponent free throw percentages. Perhaps some defensive teams also target poor free throw shooters more than others. In fact, opponent 3-point percentage is no more consistent than opponent free throw percentage from year to year. The following table shows how predictable team statistics are using team statistics from the prior year.
Team stats - year to year predictability
Keep in mind that player minutes are about 65-70% similar from year to year, on average.
The slope is the percentage of last year’s data that should be used to best estimate the respective statistic. (The remaining percent is applied to a figure approximately equal to league average in this instance.) For example, if a team shoots 80% from the line when league average is 75%, we would expect them to shoot about 78% the next year (80%*56% + 75%*44% =77.8%).
Based on the above table, defensive 3-point percentage, defensive free throw percentage and team three point percentage are the most dependent on random fluctuation from year to year.
The variability in game to game statistics tells a similar story. Using credibility assuming a normal distribution, I estimated the slopes of each statistic for each game. If we divide all figures by 63% (not too far from the percentage of common minutes from year to year), the estimates are reasonably close to those in the first table.
Team Statistics - Credibility by game
NBA 2001 early season ratings
Therefore, using the above results and information courtesy of Basketball-Reference.com, we can take a closer look at which teams have played the best so far over the early season.
Team Performance through 11-06-10
|Tm||G||W||L||ORtg||DRtg||SRS||3P%||Opp 3P%||Opp FT%||Luck - Margin||Luck - Schedule||Luck - Shooting||Luck - Total||Prj Wins*|
*Projection assumes teams continue to perform at current standard.
“SRS” – Simple Rating System from basketball-reference.com. This is basically scoring differential adjusting for strength of schedule.
“Luck – Margin” – Wins – Pythagorean Expectation based on scoring margin
“Luck – Schedule” – Pythagorean Expectation – Expected wins using SRS
“Luck – Shooting” – SRS win expectation – Expected wins removing 75% of the differences from average for opponent 3P% and FT% and 33% of the difference between team 3P% and average
In the early season, Atlanta has benefited most from factors that are not likely to continue. As of Sunday morning, The Hawks record is perfect thus far at 6-0 (Note: Yeah, they lost to Phoenix yesterday… this was written before that), but they have picked up approximately 1.4 wins more than their scoring margin dictates (scoring margin is a much better predictor or future wins than win/loss record) and 1.4 additional wins due to a soft early season schedule. As their SRS indicates, the Hawks are much closer to average than perfection.
Houston has suffered the most from poor luck thus far, losing 1.6 games more than their scoring differential suggests, 1.2 due to a difficult early schedule and 0.5 from “unlucky shooting.” At this point in the season, one could argue that Houston at 0-5 has played better than Atlanta at 6-0!
The Spurs have four wins in five games, which is nearly one more than their scoring differential would suggest. Obviously, I doubt that many would expect the Spurs to continue winning 80% of their games and finish with 66 wins. Their scoring margin expects them to finish with around 51 wins, which is more reasonable. However, the Spurs have been particularly hurt by their opponents shooting 42% from long range (compared to 34% from last year). After accounting for team schedule and “lucky shooting” the Spurs are playing like a team projected to win in the neighborhood of 57 games.
One positive factor that this estimate doesn’t account for is the future presences of Matt Bonner and Tiago Splitter (in more meaningful roles, anyway). On the negative side, RJ will certainly regress from his 60% 3-point percentage and his 68% 2-point accuracy. However, his free throw and 3-point rates are encouraging. I like that the Spurs seem more committed to involving him on the offense. George Hill has also managed to get to the line frequently in his limited minutes thus far.
Keep in mind that the 57 “projected wins” for the Spurs is not really my best expectation, but rather a measure of how well they have performed to this point. If, for simplicity, we use the credibility factor for Margin at 5 games in table 2; we can see that 5 games are approximately 37% credible. Therefore, if we assume 41 wins for all teams at the beginning of the year, we now have enough information to project the Spurs to win approximately 47 games. (0.37*57+0.63*41) A better estimate would start with a projection at the beginning of the year, but would require a completely different credibility factor.
It is interesting that the Spurs have seemed to have shifted their focus more to offense thus far this year after being such a defensive force in years past. The Spurs offensive rating has improved from 9th to 5th, while their defensive rating has fallen from 8th all the way down to 17th. Their rank in pace has jumped from 20th to 7th. Most alarmingly, the Spurs, ranked right near the top (5th) in opponent efficient field goal percentage in 2010, now rank 27th. If we assume the same 3-point percentage the Spurs allowed last year, the Spurs would improve to 18th.
Tim Varner mentioned to me in a conversation that Spurs’ team defenses have started very poorly historically, steadily improving throughout the year. He suggested that Coach Pop’s complicated defensive schemes typically take longer for players to learn. Although I’m not sure how significant this learning curve is, it at least seems like a plausible factor to me.
Despite the defensive concerns, there are positives to take from the Spurs early early performances. Expect the shooting rates to even out in the long term, but pay greater attention to changes in team and player approach. And remember, a game or two could still change the perspective rather significantly at this point.