Home > Commentary > ThunderNumbers: Visualizing Success – Team Similarity Diagrams

ThunderNumbers: Visualizing Success – Team Similarity Diagrams

(Click here to embiggen)

Please note: the comparisons were generated using stats through Friday’s games. I don’t think they would have changed much over the weekend, but keep it in mind.

In my last article, I used statistical similarities to look at how this season’s Thunder and other NBA teams compared to past NBA teams, and used those comparisons to look at how the 2011 teams may fair in the second half and playoffs. A lot has changed in the NBA since I posted that article, with the Thunder going on a bit of a hot streak before cooling off recently, so I wanted to take a renewed look at the comparisons, using network diagrams to help illustrate my points.

To make the network graphs, the closest three comparisons for each team from 1981-present were made into “connections”, and an algorithm was used to arrange the connected teams as close together as possible. You will see that, in general, good teams cluster with other good teams and bad teams group together, too, which is a positive sign that the method is working well. Additionally, teams with unique systems, like Utah under Coach Sloan and the flex offense, group together remarkably (above).

It should be noted, however, that this graph is attempting to put the 50+ dimensions of team statistics (all used for the similarity distances) onto a 2 dimensional plot, so there is an oversimplification that naturally occurs. While it’s highly likely that teams that share common comparable “neighbors” are also similar, it’s not guaranteed. A note on how the graphs were formatted: size is proportional to winning percentage, and color is associated with playoff performance, NBA champions are pictured in gold, runners-up in green, playoff teams in blue, and current 2011 teams in red.

The first thing to notice from the network graphs is that over the past month the Thunder’s comparisons have grown more favorable. Despite the Thunder being 9-6 over the last month (about on the same pace as before, including the last two OT wins vs. poor teams), the teams that are most similar to the Thunder have become more flattering. This can be seen by the network diagram close-ups shown above. The Thunder is no longer surrounded by many non-playoff teams, which they were in December. The Thunder have moved away from the cluster of Wizard’s mid-00s teams that I was so worried about before, and toward OKC’s 2010 team.

A lot of other decent playoff teams show up nearby, like the mid-00s Denver teams with Melo and AI, and two conference champions (although they are probably two of the weakest conference champs ever). The ’10 Thunder team still isn’t within view, which is surprising considering the lack of roster turnover, but not surprising considering the differences in playing style from last season to this year. The ’10 Thunder had more promising teams comparisons, with the surrounding teams having a greater winning percentage (size) and more success in the playoffs (two teams nearby were runners-up and two won the championship). This highlights just how good of a team the Thunder was last year, and lends more statistical credence to the fact that the Thunder has slipped in performance from last year, even if that is not seen in the wins column.

The network graph can also give us some insight into which teams might be the most legitimate contenders for the NBA championship this spring. The plot shows a few areas with a high concentration of championship and runner-up teams and with other great teams surrounding them. So who shows up in those key areas this year? Miami, L.A., Orlando, San Antonio, and Dallas appear in the “championship region” with the Lakers seeming to be surrounded by the most championship comps.

The network graphs are really interesting, and show a lot of compelling data/trends. I encourage you to look around in the network graph files for yourself, and see if you notice anything I missed. Let me know in the comments if you’ve found anything of note!

NBA team network diagram 2011 (you can zoom): recommended: (.png) also: (.pdf)

If you’re interested in checking out individual player comparisons, check out an excellent post at the old Arbitrarian blog. It hasn’t been updated in two years, but still one of the most innovative NBA stats blogs on the web.

Share:
  • Print
  • Digg
  • Facebook
  • Twitter
  • del.icio.us
  • Fark
  • StumbleUpon

Commentary

Post comment as twitter logo facebook logo
Sort: Newest | Oldest

very cromulent use of the word 'embiggen.'

I would take a look at Dalembert.

Here are a few rumors from ESPN on some upcoming centers that may get traded before trade deadline.

1. Dalembert's current market
We've previously rounded up chatter for possible Samuel Dalembert destinations, as it's been rumored he could be on the move by the trade deadline.

And as ESPN's Marc Stein reported in the Weekend Dime, the Kings are still open to moving the center, but it comes with some caveats.

Stein explains: "The Kings, sources say, are open to moving big man Dalembert during the next 27 days before the deadline, but they remain unwilling to take back anything other than expiring contracts to preserve maximum flexibility on the other side of a new collective bargaining agreement ... unless someone wants to donate a star. With Dalembert possessing an expiring contract valued at $13.4 million, it's not easy to find a match that keeps the Kings in the same cap position as they are now."

2. Knicks showcasing Mozgov?
Knicks rookie center Timofey Mozgov had received 15 DNP-CDs in the last 17 games, but on Sunday he played 40 minutes and scored 23 points and grabbed 14 rebounds against Detroit.

The Knicks are hoping they can put together a package that the Nuggets will accept for Carmelo Anthony and ESPN.com's Chris Sheridan wonders if Mozgov could be part of a proposed deal for the small forward.

Sheridan writes: "Just a few days ago, the debate in New York was whether it would be worth it to include Landry Fields in a proposed package for Anthony. This week, you can argue the same point about whether Mozgov is a keeper or a spare part. Perhaps Anthony himself, who has become talkative again lately, will weigh in on the merits of each of those players Monday night when he pays his one and only visit to Mikhail Prokhorov's rental unit on the other side of the Hudson. And if Melo says he's a bigger fan of Mozgov than he is of Mikhail, we'll have another clue to decipher as we count down to the Feb. 24 trade deadline wondering whether the trade that has been slow-brewing since the start of the season will finally go down -- and whether either of the two prized Knicks rookies (yes, there are now two thanks to Mozgov's big night) will be staying in New York for the long term."

@f5alcon

@gr8ball83
Haha. Exactly! I'm taking no less than 30%!

@Mark!
It doesn't necessarily indicate anything, although it implies similarity it is not always the case unless the lines are pointing to each other. The '11 Thunder are actually pulled in a lot of different directions due to their connections to SAS'81, ATL'08, and POR'86. Those are very different teams in terms of success, and the Thunder end up not very close to any one of them, but in the middle of them all, if that makes sense.

The 2010 Thunder are much more centralized, as they are right next to their second best comp, DET'04.

@TempBoy Brandon
i can offer up a lot of processing power for number crunching

@gr8ball83

lol, yeah good idea, dont want to spread profits to thin, unless somebody wants to invest a lot of money into our fund.

@Andy Collard

What does it indicate if two teams are close together but do not share a connection? Our team appears near Milwaukee '10, but I don't recognize any similarities on the surface.

@TempBoy Brandon
sounds good, enrollment is capped, sorry everybody!

@f5alcon
Count me in as well. We'll give Andy 10% and then cut the profit 3 ways.

@Crow
Great idea. I can do it when I get home from work. That would be an easy thing to do.

@f5alcon
I'm in for that project for sure

@TempBoy Brandon
apply formula to stock market
profit???

Would you have the ability to write some sort of code to calculate average win % and other stats of teams within x or y distance of a given team? It could of course be done manually but it would be cool to have a quick automated way to do that and report it for teams.

Pretty cool stuff. Although, it makes me think of basketball as some big science experiment and the only dude who really knows who is gonna win it all is that guy from A Beautiful Mind.

FYI: files at the bottom are the ones to use. The top one is shrunk to load quicker.

I miss last year's Thunder.

embiggen is an awesome word

Trackbacks

  1. [...] This post was mentioned on Twitter by Royce , kweisenthal. kweisenthal said: ThunderNumbers: Visualizing Success– NBA Team Similarity Diagrams, by Andy Collard http://bit.ly/h90jfi [...]