5 min read

ThunderNumbers: Contracts, performance and the Thunder

ThunderNumbers: Contracts, performance and the Thunder
AP Photo

The reason I proposed (or desperately hawked) Thunder Numbers to Royce is to have a forum to research many of the claims that people make without any real proof. I’ve been hearing a lot of talk about players along the lines of “Player X should have a big season, it’s a contract year”, “Player Y is mad that he didn’t get offered an extension, he’ll be out this season to prove his team wrong” and “It’s a contract year, I (Nick Collison) should totally grow my hair out to look younger.” But does this performance improvement actually happen? And how big is it?  The Thunder have three important players who will be in their contract years in Jeff Green, Nenad Kristic and Nick Collison. Can we expect them to step up more than usual with the added incentive of millions of dollars?

Luckily, I wasn’t the first person to investigate this topic. In 2006, a Vice President of the Federal Reserve Bank of New York wrote an article about NBA contracts and performance, instead of, you know, watching out for any potential housing bubbles. To be fair, I’m sure that this article wasn’t the difference between prosperity and financial collapse (probably); economists often use professional sports to measure measure various economic effects and incentives because it is easier to assess an NBA player’s performance than a random guy. So anyway, he found that there was a positive correlation between contract year and performance, using standard box-score statistics and a combined stat called “composite ranking”. I’ve made a quick table with the paper’s major findings below. Reading the table, it implies that the average contract year effect on points/game is +.847, which is pretty substantial.

 Composite RankingPtsRbdsAstsBksShotsFTsMts
Pre (SD)-0.172-0.244-0.095-0.075-0.020-0.184-0.078-41.468
Post (SD)-0.146-0.215-0.091-0.059-0.018-0.169-0.067-36.636

While the findings are pretty interesting, I had some reservations about the work, including the fact that new contracts can be signed at so many different points in a player’s current contract and in so many different ways: extensions, restricted free agency, and unrestricted free agency, as well as both player and team options exist. It left a lot of room for competing effects, noise, and selection bias to occur. I also wished that another statistic was used for the regressions besides composite ranking, which is out of date and not used very much anymore. Ideally the stat used would be able to demonstrate the real value added to a team, in terms of either wins or money.

So, I decided to try to make these changes to his methodology, and see if the contract effect could be parsed into more specific terms. Pro Sports Transactions has a huge database of transactions to sort through; its great and very extensive. After a lot of data entry, I have 4106 player seasons for 742 different players from 1998-2010, with 1033 contract observations. I decided to use win shares as my main dependent variable for most of my regressions. It’s pretty good as far as aggregate advanced stats go, much better than PER because it is not as affected by usage. For PER, roughly each shot you take at above 29% or 3pt at about 20% will increase your rating, but with win shares there is a much more reasonable account of efficiency. Any action (shot, turnover) operating at less than 92% of the league’s average efficiency is penalized. Also, with the win shares, we can estimate how many wins a player would generate in a contract year and with a little extrapolation, added value in dollars. I’ve posted my spreadsheets and methodology for anyone who is interested on the APBRmetrics forum, so I’ll try to keep it broad and nontechnical here (or at least try to).

Below are the impacts of each contract type I studied, restricted free agency, unrestricted free agency, and post contract effects (after signing a contract 4 years or longer). These three contract types were the easiest to define, and the Thunder happen to have at least one player in each category (Jeff Green, Nenad and Collison, and KD).

 Win SharesMins/GameUsageTS%AST%ORtg
Res. FA.467**0.8270.2490.00345-0.1011.123*
Un. FA-0.018 -0.745***-0.203*0.001250.03970.348
Post.638 **1.095**0.440**0.0014340.696**0.3956
Age2.53 ***7.612***2.687***0.0260***2.518***4.903***

Note asterisks show significance: * p<.1, fairly significant; **p<.01 significant; ***p<.001 very significant

The first thing I noticed is that for some stats there is a very strong  contract effect shown (like win share or defensive rating), and in other cases there is not a clear effect at all (true shooting %, AST%). Looking at win share, you see restricted free agents produce about .5 wins more than they would normally be expected to, which sounds roughly right, maybe a 10-20% increase in this stat, given that the average for the spreadsheet is 3.7.

Strangely, however, free agents show a decrease in win share; this goes against our assumption of contract incentives. Looking further at win shares/ per 48 minutes explains this further, and we see that free agents do seem to play more efficiently per minute, so they must be getting less opportunities. This is definitely the case, as both minutes per game and usage (the percentage of a team’s possessions taken up by that player) are negatively correlated with being a free agent. This could be explained because often if a player is put in the position of a contract year, the team may be exploring options without the player in question around.

Another thing to notice is that there seems to be no decline in performance associated with a long term contract. I wasn’t concerned about Durant’s production dropping off before, and now I’m definitely not. It seems like major disappointments (Eddy Curry) notwithstanding, teams generally offer long term deals to players who actually improve. We also see a jump in  win shares, usage, and minutes the year after a long term contract, which makes sense, too, because if a team has just made a major investment in a player, it will likely want to play him more.

So what can we make of this, especially as it relates to the Thunder? It does seem reasonable to expect that Jeff Green will improve, with his contract status likely to account for added efficiency. His increase could even be expected to be larger than average, given that his win share is about twice the average restricted free agent’s. However, the bigger news to me was in the unrestricted free agent data. If Krstic and Collison hold to the trend, It doesn’t seem likely that the added contract incentives will make a major difference in their play. The data instead fits what we’ve already seen, the team probably won’t wait until this summer to see if they have adequate replacements for them. Drafting Cole was a major step, and they’re going to start gradually plugging him into the frontcourt rotation. As much as people may like to think otherwise, these minutes probably aren’t going to come from Green, they’re likely to come from Krstic and Collison, and the data backs it up.