A Phoenix Suns cluster analysis

In a business analytics course in my sports MBA program, we learned about a concept called cluster analysis in which a set of objects are grouped so that the objects within each group are more similar than objects in different groups.

I applied that newfound knowledge to an NBA dataset I created that takes into account a team’s regular season offensive rating, defensive rating, offensive rebound rate, defensive rebound rate, true shooting percentage, assist ratio, team turnover percentage and pace to group teams covering 12 seasons of NBA play from 2000-13. The lockout-shortened 2011-12 season was not included because 16 fewer regular season games were played.

I used Ward’s Method to break the data into 10 clusters and found one 31-team cluster that can essentially be called the “Suns” cluster because it features six consecutive Phoenix squads during the heart of the Nash Era from 2004-10. Perhaps it should be no surprise that the three Dallas Mavericks teams from 2001-04 also slot into this cluster since Nash was at the controls of those teams as well.

One of the most interesting aspects of cluster analysis is seeing which other teams the SSOL Suns are grouped with. Since those Phoenix teams put up such consistently incredible offensive efficiency/true shooting numbers, it should be no surprise that they ended up in the same cluster. The Suns’ cluster also includes a host of elite offensive teams such as the 2006-09 Lakers, the ’09-13 Thunder, the ’04-05 Kings, the ’12-13 Clippers and Warriors and the ’06-10 Jazz. The ’07-09 Celtics and ’04-05 Heat are the East’s only representatives in the cluster, which speaks to the stylistic differences between the conferences.

Looking at the dataset as a whole, many clusters possess the same team in consecutive years but few teams had such long runs of belonging to the same cluster. The Atlanta Hawks — who missed the playoffs every season from 2000-07 — held the only seven-year run, whereas the Suns’ perennial Western Conference foes the Dallas Mavericks and San Antonio Spurs each had a six-year streak of being in the same cluster from 2006-13 with the lockout season perhaps preventing a seven-year run. That speaks to the remarkable consistency of the SSOL Suns to maintain such a similar level of play throughout a six-season stretch that could only be matched by their rivals of that era from Dallas and San Antonio. The Hawks’ run of consistency, meanwhile, was not in the way a team hopes to be consistent.

The Suns’ cluster outscored opponents by 5.32 points per 100 possessions, a clear leader among clusters with the next best cluster at 4.40. It is the best offensive cluster with a 108.4 offensive rating and a true shooting percentage of 56 percent. The Suns’ cluster was statistically significantly better in terms of offensive rating than all but one cluster and statistically significantly ahead of all but two in true shooting (one being the Spurs/Mavs cluster).

Defensive rating (surprise, surprise) was a different story, as the Suns’ cluster ranked fifth out of the 10. They were also seventh in offensive rebounding percentage but second in pace, so you can see how this is a very Suns cluster.

This Suns’ cluster averaged 53.7 wins a season, the best average of any cluster and a mark that was statistically significantly better than all but two clusters (yes, one being the Spurs/Mavs one). These teams led all clusters by reaching the playoffs 94 percent of the time (the Suns’ 46-win 2008-09 team was one of two outliers along with the 43-win Rockets from ’10-11), so this offensive-oriented cluster was the best in the regular season.

Thanks in large part to the Suns’ three conference finalists, this cluster reached the NBA’s semifinals 35 percent of the time, good for second behind a cluster that did so 36 percent of the time. However, like the Nash Era Suns, that’s where the cluster hit a roadblock. Even with all the playoff teams, it dropped to fourth in Finals appearances with 10 percent and third in championships with 6 percent. The 2007-08 Celtics and the ’07-08 and ’08-09 Lakers were the only teams to reach the Finals from the Suns’ cluster. The Mavs/Spurs cluster led in Finals appearances with 18 percent and was tied for the lead in titles at 9 percent.

Ultimately, this analysis provides an interesting way to group teams and matches history as well. After all, the six SSOL-era Suns squads certainly belong together in an elite scoring/shooting cluster that led in net rating but did not enjoy as much playoff success as some of the other groupings of top teams.

Tags: Cluster Analysis

  • john

    Nice analysis. I’m liking this methodology. The Nash era Suns will be one of the most talked about teams of this era despite failing to win it all.

  • Bill-in-Tokyo

    Interesting but what were the characteristics of the Spurs/Mavs cluster which made them the best? In an NBA interview with the “Big O” on his 75th birthday, Robinson said the great Celtic teams weren’t great offensively but it was their defense. Perhaps this is the key lesson from your analysis?

  • KD

    Nice use of multivariate analysis Michael. While the sample size is small, and different, what cluster does this season’s Suns fall into, and therefore what does it predict for 2013-14 record?

  • Serek

    My thoughts exactly, KD.

    Michael, maybe you could extrapolate this season’s stats so far and add them to you data set? I wonder where this year’s Suns would go? Maybe we could make some predictions based on that.

    In the meantime, it would be interesting to know what other teams were last years’ Suns with ;)

  • Dave:f32

    Cluster muck?! What does it all mean with regard to the Suns?

    So, you’re telling us there’s a chance (championship)… I luk it ulot (lloyd christmas voice)

    They taught us in college that data can be interpreted in an infinite amount of ways, depending on what ur tryin to convey so…

    What data doesnt do is factor in variables to equations

  • Dave:f32

    And how those variables affect formulas

  • Bill Danforth

    Why would you favor the SSOL Suns in your dataset by including pace? What about their opponents pace? Did they allow their opponents to play at a faster pace thus giving them more possessions and an opportunity to score more points against the SSOL Suns.

    From what I’m reading you have simply skewed statistics to support what everybody already knows, SSOL scored a lot of points but failed in the playoffs because the pace slows down into a not so typical SSOL Suns style which your analysis correctly supports as the Spurs cluster has the highest number of titles and Finals appearances.

  • http://www.valleyofthesuns.com Michael Schwartz

    @Bill The Spurs/Mavs cluster was actually “only” fourth in defensive efficiency and not statistically different than the Suns’. It actually ranked better offensively (third) than defensively. However, they were elite in defensive rebounding percentage, leading all clusters at 75 percent. I think for championships in particular the sample sizes are too small to prove anything with only 12 clusters and really Finals as well with 24. Too much luck and random variation, which is why I tried to stay away from any kind of a hard conclusion statement.

    We learned in class that even adding in one new record can really change around the clusters, and that’s what I noticed when trying to add this year’s Suns. They would fall into what can be considered a “good” cluster that makes the playoffs the majority of the time, but I think we all know if they keep outscoring the opposition by 3.2 per 100 they will probably make the playoffs.

    Last year’s Suns were in the biggest cluster with 65 teams that averaged a -2.24 net rating (third worst) and actually made the playoffs 32 percent of the time. The cluster averaged 36 wins, so the Suns were obviously on the very low side of that. There were a handful of mediocre-bad Clippers, Knicks and Bucks teams in that cluster.

    The inclusion of pace is probably a big reason the six SSOL Suns teams were in the same cluster since they all had similar offensive efficiency, true shooting and pace numbers.

    Another “flaw” is that these were all regular season numbers since I was grouping teams by regular season success. Some teams, like the Spurs, always seem to kick it up a notch in the postseason, but then you are dealing with tiny samples with postseason data. Bottom line, I don’t think we can use this to reach conclusions about the kind of teams that win in the playoffs because the samples are too small.