Posted on: March 3, 2012 8:43 am
Edited on: March 3, 2012 9:13 am

Can we get smarter at building teams?

Research suggests the Magic may not have built around Dwight Howard the right way. (Getty Images)
By Matt Moore

So there's this big sports analytics conference called the Sloan Sports Analytics Conference. If you're an NBA hoophead/blognut/basketball freakazoid you likely have already heard about it, since most of the writers who like to stretch beyond the tired cliches tend to spend a lot of effort talking and writing about this thing. 

The event's held at MIT with a bunch of "wicked smaht" people talking about a number of things that would likely bore you to tears if you're not a fan of sports geekery. It's not athletes talking about swagger (though NBA Deputy Commissioner Adam Silver dropped a reference to that Friday which was hilarious), or about clutch (though there was a panel on how pressure impacts players). It's people that work in front offices and behind the scenes (for the most part) talking about regression analysis, paradigms of alternative thought, and correlation (not so much causation). 

It's easy for a lot of people to toss this stuff to the wind. It doesn't fit with how we usually view sports, and how we usually view sports is fun. It makes it complicated, it hones in on stuff that's too specific (a common complaint from players and coaches who love the forest, hate the trees), and it sounds like stuff that's too dense to take anything actionable from.

That's the trick. It's like what blogging really is about (besides funny videos, photoshops, and poor syntax). You have to search through the mess, take out the interesting components, and try and put them into a bigger perspective. What follows is an attempt to guide you through two compelling research papers presented at the conference, and why you should care about their results. 

Big 2’s and Big 3’s: Analyzing How a Team’s Best Players Complement Each Other

One major takeaway from the Big 3 results is that the data shows that, cluster 8, the multi- faceted small- forwards who are very good 3 - point shooters, are great players to build a team around, as long as there aren’t any similar players among the most talented players on the team. Very good results occur when these small- forwards are surrounded with a variety of player - types; the Big 3’s with the highest coefficients (7 - 8 - 12, and 8 - 10 - 12) both include players from cluster 8. This was true with the Big 2’s as well.  

Robert Ayer presented this study which had its methodological quirks. (I would have liked to have seen a better efficiency model than the one provided and even accounting for minutes, we should never be using per game numbers for anything more than a highlight clip for toddlers at this point.) But overall the thought process was really ineresting. Essentially, he classified players, factored their efficiency, and then ran analysis to discover what worked and what didn't work together. It's like using all of the data from NBA history to create models of the players we describe in broad terms and then using advanced metrics to figure out which of those archetypes should be used together to build a team. 

The Rub: Putting a pure point with a dominant center may not be as effective as pairing a versatile wing with a little shooting ability with the same dominant center.

The explanation: This calls into question the idea of the point-guard-big-man fit. For example, Dwight Howard, it has long been thought, needs to play with a great pure point guard. But his greatest success has been with a versatile three who could run the pick and roll and had some three-point shooting ability, in Hedo Turkoglu during the Magic's 2009 run. The analysis suggests that Howard would do better with, say, Andre Iguodala, who can defend, distribute, rebound, and score when called on, versus say Deron Williams. This doesn't mean that the two aren't a good fit. It just says it's possible that if you consider Deron Williams and Andre Iguodala equal talents in terms of their relative skillsets, that Iguodala and Howard might find more success from a production standpoint. 

It also speaks to how Otis Smith's move for Vince Carter in 2009 may have been the right move. If you improve upon Hedo Turkoglu's three-point shooting with Carter while keeping the same versatility, it's a win. The flaw may have been over-estimating Carter's diminished ability as a passer due to age. 

The fact that so much success was gleaned from wings in the study, be they versatile passers or high-volume scorers, suggests a radical shift in traditional thought about the strength of players. Wings are most often criticized regarding their tweener status while classic big men and point guards are idolized, outside of the exceptions like Kobe Bryant and Michael Jordan. Turns out your small forward can have a huge impact on your winning percentage simply due to his position. 

In another real world application, high-scoring small forwards fit well with high-scoring 2-guard, but high-scoring power forwards and those same small forwards did not make the list of good matches. So there's that, Knicks fans.

The Rub: Having two players that do the same thing on your team isn't just bad, it's really bad. 

The Explanation:  Sacramento brought in John Salmons, Isiaiah Thomas and Jimme Fredette to a back court that already featured  Tyreke Evans and (presumably, in restricted free agency) Marcus Thornton. The idea was shots on shots on shots on shots. But instead, you have several players essentially with redundant skill sets, and the paper points out this stuff kills teams' production. 

A practical application of this is an assault on the best player available concept towards the draft. It's fine to draft a player like the one you have now, as long as you move one or the other, or do not play them together. The negative impact the study reveals in redundant players suggests that there's no point in stockpiling at a position if the two players are essentially the same.

A counter to this though lies in a confounding wins vs. production element from the paper:

Most observers would think that a Big 2 from the same group would not fit as well; this is partially contradicted by this analysis. While multi- faceted small forwards who shoot 3’s don’t fit well together (8 - 8, - 4.046), teams with two high scoring 2 - guards (2 - 2, 3.97) have historically over - performed their expected win total, given the team’s overall talent level and coachi ng skill. Digging a bit further into the data, nearly all of the teams with multiple high - scoring 2 guards played at a higher than median pace; although further analysis would be required to state conclusively, this is perhaps instructive on  the style of play that teams with two high - scoring 2 - guards should employ.
So pretty much if you want to stick two gunners on the floor together, that's allright. Some real world examples of this might include the 2011 Hawks which employed lineups featuring Joe Johnson and Jamal Crawford, and to a certain degree the Nuggets with J.R. Smith and Carmelo Anthony. The best example featuring a fast team might be the Seven Seconds or Less Suns with Leandro Barbosa and Johnson.

Takeaways: When you're building a team, you need to avoid big men stepping over each other. But you can duplicate shots, if you run in a fast-pace offense. However, you should look before you leap because it turns out small-forwards are pretty important by all accounts.

NBA Chemistry: Positive and Negative Synergies in Basketball

Why is Chris Paul for Deron Williams a mutually beneficial trade? Overall, our SPM ratings rate Chris Paul and Deron Williams nearly the same, but with differences in skills. Paul is a better ballhandler, Williams a slightly better rebounder, and Williams is better at offense and defense.

The SPM framework predicts that Chris Paul is a better fit for Utah because he creates a lot of steals (3.1 steals per 48 minutes (“SP48M”)), while no one else in the New Orleans lineup does (West 1.0 SP48M, Stojakovic 1.1, Chandler 0.7, Butler 0.9). Utah, on the other hand, has many players who create steals (Kirilenko 2.0, Boozer 1.5,  Millsap 1.7, Okur 0.9, Williams 1.4). Because defensive steals has positive synergies in our system, Chris Paul's  ballhawking skills fit better in Utah, where he can team up with others and wreak havoc to opponents' ballhandlers.

Conversely, why would New Orleans trade for Deron Williams? Our framework predicts that Williams is a better offensive fit with New Orleans. There are negative synergies between two good offensive players since they must  share only one ball, and the New Orleans starters take fewer shots than Utah’s. At New Orleans, Deron Williams  would not need to share the ball with so many players.  

Allan Maymin, Philip Maymin, and Eugene Shen presented a doozy of a numbers-fest which took a non-traditional spin on advanced plus-minus. In short, how well does a team do in a specific area like rebounding or turnovers versus their opponent when a player is on the floor versus off. There were some methodology issues in this one as well, but the concept was intriguing. 

The Rub: The get-at here is that player skills are irrelevant if they don't mesh with the team. Their kicker was the Paul-Williams trade concept, which says that both teams would benefit if they made a trade for each other's guard because of who the rest of their teams were. 

The explanation: This goes back to building around a star. In short, you can build good players around a great one but it doesn't matter if those other players' skills aren't complimented by the strengths of your star. We focus a lot on bringing in talent around a player. But bringing in offensive weaponry when your star's biggest impact on other players is defensively is missing the point. It's not about trading the best player, it's about finding the best players to surround them with. 

This seems obvious, but look at how many teams create logjams with their decision-making. For years the Warriors have been a defensive nightmare despite having two guards who both need the ball in the backcourt. The paper also touches on ball-handlers being redundant with one another because there's only one ball to share. The success of dual-point-guard lineups seem to contradict this measure, but in those situations, the players do thrive because one player takes on a scoring role. Understanding role play is crucial to this and it would be great to get coaches' thoughts on these ideas. 
The views expressed in this blog are solely those of the author and do not reflect the views of CBS Sports or CBSSports.com