[Aran SmithKeymaster]

http://www.gq.com/sports/profiles/201206/muthuball-mit-sloan-sports-analytics-conference

"Muthuball"?  Which 4 teams offered this guy a contract?  OKC, San Antonio, Boston, Houston???

13 NBA positions?

Thoughts?

[DanEboyParticipant]

Check out Aran, mingling with us common folks, creating forum topics. Every time someone comes up with one of these SABER metric type dork stats and new perspectives of looking at things, I often think about what old timer horse racing degenerates say, "Think long, you think wrong."

While these stats and and perspectives serve some purpose, they are often just people over thinking things way too much. Billy Beane had some success, but he never won a world series and he has missed the playoffs the last five seasons, and will make it six this year. I will always take the eye test over these kinds of stats.

They also tend to affect contracts. Players who have high PER ratings can use that at the bargaining table to get more money and so on. When Bill James invented the save in baseball, he instantly made relievers a ton of money.

It’s late, and I don’t even know if any of this makes sense. This is turning into a Mikey V type response.

[mikeyvthedonParticipant]

Yours is way too short. Anyway, when I first heard about the "13 positions", I was kind of skeptical. Reading this was a solid clarification and I think this definitely could be a valuable tool for helping teams improve. I for one would be incredibly interested to see how teams breakdown in generational context, as it seems there is a definite formula for what makes up a championship squad.

Of course not every team is identical, but it seems that defense and rebounding, combined with capable scorers, have been what has built the majority of champions. This may sound obvious, but yet there are still many teams that never seem to get it. In the play-offs especially, mismatches are picked apart and destroyed, it seems speed, width, length and strength are all factors as far as that goes.

It seems that Muthu is definitely interested in moving into the management side of basketball rather than being an overall supplier, though I am sure people will use his formulas. I agree with Dan to an extent that "the eye test" still plays a major part as opposed to raw statistical data building championships. Raw talent could still beat even the most statistically efficient teams, and not every team will have an "All-NBA First Team" or "One-of-a kind".

Still, when one looks at the Detroit Pistons of 2004 (who did not have either, I believe) or last years Mavericks (who had an All-NBA First team player, but one who was classified as something else under Muthu’s formula), these were really well constructed teams that seemed to buck the recent "Big 3" theorem. Stars will always be important, as is surrounding them with the proper talent. This formula seems to be used to do just that, finding the correct role players to surround ones team with. Not every team will HAVE 13 positions, just knowing that they exist might be a great resource.

Talking with someone off the record (do not know if it is cool to say who), he brought to my attention something about "Dork Elvis" Darryl Morey. It seems that while Morey stacks some talent and has produced some under the radar solid teams, he seems to never let them develop the longevity to build the chemistry to make them much more. It is funny that analytics is a way of trying to build team chemistry for wins, however it also takes time to develop that seems to not always be taken into account.

Again, hope I am making sense and that this post is not too long. Thought it was an interesting article and wanted to just drop some things to (possibly) think about. Analytics may never fully replace gut instinct. It can still bring some insight into how these championship or successful teams were constructed and give teams a stronger map towards success.

I believe every NBA takes a lot of interest in this type of analysis and it is becoming the norm to hire brain trust like Muthu. Will be interesting to see if his results or beliefs become more wide spread, would love to see a breakdown of the positions of past teams. The 5 positions will always be common use, just seems that the 13 positions could provide a nice way of breaking down ones roster and developing a winning chemistry. It will show what works (ie. the Dallas Mavericks 2011) and what doesn’t (ie. the Minnesota Timberwolves 2011).

[B-ball fanParticipant]

 I think this looks promising.  I think the word "position" may be a bit of a misnomer, as Muthu seems to be classifying statistical profiles, while positions are typically thought of from an Xs and Os perspective, rather than a statistical one.  

I certainly would be interested in reading a study into the plus-minus numbers of different combinations of the 13 positions.  I wonder how much balance of the 13 positions matter.  Do scoring rebounders fit well with defensive ball handlers?  Do paint protectors fit well with combo ball handlers?  It could be useful for GMs interested in which 2nd rounders to pick or veteran players to sign to cheap contracts.

I am not sure how this could be of great practical use, though.  Teams can already evaluate players from a statistical perspective, and I wonder if the particular positional designation of a player by Muthu might just seem like irrelevant trivia on top of analysis of the efficiency of teams’ roster combinations and other similar tools.  And like MikeyV mentioned, when picking out players, raw talent is apparent from watching players, even without using statistics.  

I wonder if he has done any work to see if players change the "positions" often.  If their statistical profiles remain constant, then this might be useful.  If it is dependent on the situation that the player is in, these "positions" may be of little practical value to teams.

On a second note, he mentioned that last year’s Timberwolves didn’t have a scoring rebounder.  I wonder how he classified Kevin Love?

[omphalosParticipant]

@B-ball fan; it mentions in the article that Griffin and Love are both "scoring paint protectors"… somehow.

[B-ball fanParticipant]

 Thanks, I missed that.  I wonder how that works?  I guess the name of the position doesn’t really matter, but that seems strange.

[xbadgerhustlerParticipant]

I think you guys are kind of missing the point of the article… it’s not necessarily saying that this "13 positions" idea is a fool-proof way to build a championship team… It’s taking one random set of data and saying look at the patterns and trends we can see based on that set of data using this program/software/whatever Ayasdi is… 

Now imagine being able to adjust the data set: 1) to include different metrics (e.g, adding win scores or FG% to his basic example) and 2) having the data evolve as players evolve over time… 

I don’t think that Muthu is trying to say that this system should be used as a decision-maker… No team should say, oh, we need a "Scoring Rebounder,"(let’s take the last year Wolves example) and say, oh we have five "Scoring Rim Protectors," let’s trade one of our scoring rim protectors for a scoring rebounder… Ok, let’s trade KLove for Devin Ebanks… That will balance our team… NO.

It’s just a powerful tool that can be manipulated and used to HELP make decisions.

@bball fan you asked two questions: Do scoring rebounders fit well with defensive ball handlers? Do paint protectors fit well with combo ball handlers? 

I have no doubt the system could be quickly adjusted to answer those questions, as well as every other combination of this arbitrary 13 positions. The beauty of the system is that you can add a couple different metrics as I mentioned before, come up with a new set of "positions" (or probably more accurately described as Roles) and then ask the same questions in probably less than an hour… 

Finally, I just wanna weigh in on my view of how to build a championship team… It’s not about having 5 guys who play the traditional 5 positions… It’s about finding a rotation of players (and the rotation of players can change any given game) that can bring a specific set of skills that will in theory, put you in the best position to score more points than the other team over 48 minutes. Take 3 point shooting as one example of a skill… Traditionally, we would say, that’s a skill that a 2 or a 3 should have.. but really, the only necessity is that there are guys on the court who can stretch the defense… It doesn’t particularly matter if it’s Kevin Love, a traditional 4, or Anthony Morrow, a traditional 2.  

So then the last step is to say, what skills do I have on my team, and what skills do they have on the other team, and how can we optimally deploy our guys to minimize the skills of the other team… I think this all sounds very broad and not concrete, but I think that’s kind of the point. There is no formula or recipe. It’s about having 15 guys who bring a variety of skills and tendencies to the table, and then mixing and matching and EXPLOITING those various skills, based on the opponent.

 

[B-ball fanParticipant]

 My questions about different players fitting together were just rhetorical.  I was just wondering if, like you hypothesize, it could be used to answer questions similar to those, or if the answers were too dependent on the individual player.  

I do see the value in it, but I wonder if using too many of these advanced algorithms just dilute the decision making process.

[xbadgerhustlerParticipant]

@bball fan, i agree, i think the algorithms could dilute the process if you rely too heavily on it for guiding the overall direction of the team, because then the vision for the team can change al the time. But, I think it can definitely help a team look at isolated decisions to try and make the right choice… Like say a team knows it wants to trade a player, like say OKC wants to trade James Harden because they don’t want to pay him… I think it could help them to look at finding a replacement player who can bring a lot of the same skills/production… And then if you can find someone (somehow) who brings a lot of what Harden brings, but is way less valuable, then you can find inefficiencies in the way the league values players.

Take, for example, the Bobcats trade rumors that were floating around with the #2 for Harden. Just assume that OKC wants to unload Harden for a similar replacement talent that costs much less than Harden will command in the offseason… Just for the sake of this example, say that Gerald Henderson has similar production/skills to Harden, then OKC could take a small step down (in this hypothetical) and also acquire the #2 pick and probably more! (Or, to make it more realistic find a player on a 3rd team that plays like Harden, and then bring that team into the trade…) Then OKC ends up with an affordable guy for a couple years who can somewhat replace the Harden’s production/skills AND add another affordable talent like Bradley Beal or Thomas Robinson, all for less than Harden… I think the system would be extremely helpful for something like that.

With that said, I’m sure you could do the same thing without the system, but I think it would be much more difficult, take much more time, and have a lot more room for error.

 

    

[King CaluchaParticipant]

This is not the first time players are grouped in order to be studied. In the following paper, the authors propose 14 clusters and found some of the most winning big 2’s and big 3’s based on that classification.

http://www.sloansportsconference.com/wp-content/uploads/2012/02/Robert-Ayer-Sloan-Sports-Analytics-Conference.pdf

[HaleParticipant]

Not a big fan of this. The dude labels positions poorly (Griffin and Love paint protectors?) and I don’t care what it’s intended for, if I’m a GM I’m not looking at articles on how to build my team. 

[TallmanNYCParticipant]

 I have no idea why Love and Griffin are considered paint protectors unless the ability to defensive rebound counts as protecting the paint. I guess it does. But most people think of shot blocking which neither Love or Griffin do. 

Strange article. 

[Author]

Posts