In basketball, though it isn’t discussed particularly often (or wasn’t until talents were taken to south beach) there are diminishing returns on offense skill. There are a limited number of minutes and shots per team, and while more good offensive players will tend to increase the general efficiency of their teammates, there is a theoretical limit to how much each additional offensive upgrade will make the team better.[1]
The same is true in hockey and football, though probably to a lesser extent in the hockey than in basketball. In baseball, nine players have to hit (whereas in the extreme basketball case, a Wilt/AI/Kobe/Carmelo can shoot the ball 90% of time while his team watches for a quarter or more), so, while you will see some diminishing returns if you hit an amazing hitter ninth in the order, the fact that the ninth player is generating fewer outs will actually improve the total offensive output (a team with nine .400 OBP guys will get 2-3 more baserunners and trips to the plate per game than a team with nine .300 OBP hitters, which gives them a better chance to score more runs).[2] In hockey, I would assume it would take much longer to see diminishing returns since so there are so many more players involved. If you have three A+ centers, while the other team has only one, you will have a distinct advantage because every team has to play at least three centers in every game.
I’ll admit that following thought was not inspired by an ad for the Madden game, but it would seem that diminishing returns vanish (at least until we reach playing time maximums) for defenses.[3] In football, if you have three receivers who are 30% more likely to gaining yards than an average player when each plays with other average players, you will not be 90% more likely to gain yards, because the quarterback can only pass the ball to one of them. Much more likely, you’ll be something like 40-60% more likely to gain yards thanks to having multiple options and being able to pick the best one. On defense, however, if you have three players who are 30% more likely to sack the QB than average, each of those players will require extra attention from the offensive line; while only one can actually sack the QB, you would likely see increasing returns (i.e., more than 90% more likely to get a sack) because you can take full advantage of the extra ability to get around the offensive line on each play, and the added threat from each above average sacker will make it harder to double cover or block and single one of them.
Let’s put it in basketball terms: if you have five really good shooters (we’ll say 45%), each of them can create space for the guy with the ball, giving him more time and vision and allowing him to pass to each one. Let’s say that in this situation your shooter gets the ball off and into the basket 52% because he’s more open and the team can get one of the best shots possible (which is about what you get from teams with good spacing and good offensive efficiency). Still, at any one time, only one guy is shooting the ball, so the skill of the other shooters only really matters in how it forces the defense to react to them. Now let’s say you have five really good defenders. There’s no good way to express that, but each defender is being used at the same time, guarding either their man or their space on the floor. Additionally, since no defenders have to “pick up the slack” off anyone else, each can be more effective against their particular opponent than they would be with a bunch of scrubs. In other words: increasing returns!
There are, in my mind, a couple implications for the fact that you get increasing returns on defense and decreasing returns on offense. First, it helps explain why, in basketball and baseball, it seems that defenses can improve dramatically with just one or two switches, a phenomenon most recently documented with the 2008 Rays and Celtics and last year’s Mavs team: the Rays made a couple crucial upgrades and jumped from last to first in defensive efficiency; the Celtics and the Mavs each credit one key defensive addition for changing the way the whole team reacted when they did not have the ball. A massive upgrade at one position defensively helps even when that player is nowhere near the ball and the player they’re covering has no chance of getting it. In contrast, upgrading the Bull’s shooting guard to Ray Allen would do nothing for them offensively if Allen stood there with his back turned to the ball.[4]
Second, it should mean that in baseball, your really really good shortstop or centerfielder is going to make the defense better around him as well. This is not to say that he’ll make the actual defenders better, skill-wise or by metrics, but that the way they can adjust themselves because he plays so well at his position should mean that fewer balls get through their zones as well as well as through his.[5] The reverse is also true: witness how Jacoby Ellsbury is suddenly a defensive metric darling. Now, it’s likely that his defense improved. It’s also likely that playing next to someone other than Jason Bay
Which brings me to my final thought: right before last spring break, a friend and I were talking about our all-defensive teams for basketball. The qualification was that the player had to be good enough to get everyday playing time, and we went back only to 1990. Some of the picks might be replaceable with others, but we settled on the Gary Payton, Joe Dumars (MJ would have been cheating, though he really is the best answer), Pippin, 2004 KG, and early 90s Hakeem. It’s sortof cheating that most of those guys were good offensively as well. Still, what if you went with the primarily offensive guys: AI, Kobe , Carmelo, C-Webb, and I don’t know, Pau
[1] Think about it this way, if you will. Assume that your team has six guys who are capable of creating/scoring 15 points in forty minutes of play. 6*40 = 240 minutes (or the total number of minutes your team can play in a game) and, if they use strategy to perfectly maximize their output, 15*6 = 90 points. Adding a seventh player of equal ability will not increase your scoring at all, since adding him will take minutes away from other players who would score the same amount in that time. This is, of course, the physical maximum of diminishing returns. In all likelihood, a team would experience diminishing returns long before it reached this situation, in part because their shot-sharing strategy would not be perfect.
[2] I would assume that the difference would compound, as well. A team with nine .300 OBP guys would generally have a pretty low chance of scoring a run in any given inning without hitting a lot of home runs (which are, generally speaking, rare). Even if one of their players does get on base, the chance of another player batting them in is rather low. For the team with the .400 OBP, the chance is likely quite a bit higher, both that someone will get on and that someone will bring that batter home. You would expect the .400 team to not only have more baserunners but to score a larger percentage of them.
[3] Another thought about this commercial: EA and Gamestop, probably the two most anti-consumer video game-related companies on the planet, are teaming up to make it so that you only get the All-Star team in Madden if you purchase it from Gamestop, which means we’ve officially reached the point where purchasing bonuses are things that would have been included in the game a decade ago.
[4] I don’t know. Maybe he’s having a staring contest with Denzel Washington or something.
[5] I believe it was an ESPN article that recently argued that 1. Curtis Granderson should be an MVP candidate (another discussion, for another time) and that 2. his defense wasn’t as bad as it looks because he plays next to Brett Gardner and so adjusts in ways that UZR doesn’t correct for. I’m skeptical, in part because I find the argument questionable generally, but also because UZR hasn’t much liked his defense since 2007 and he played next to Brett Gardner last year and recorded a slightly positive score.
A very interesting thesis. I think, though, that offensive skills may not be as zero-sum as this would suggest.
ReplyDeleteI'm thinking particularly in terms of basketball (and I'll admit shamelessly that I began this thinking about Steve Nash). As you describe for a theoretical all-defense team, "There’s no good way to express that, but each defender is being used at the same time, guarding either their man or their space on the floor...no defenders have to “pick up the slack” off anyone else, each can be more effective". I think, though, that the same works in offense. Each offensive player occupies a man/space on the floor, which impacts how the defensive players align against them, which could lead to better/more efficient shots for better scorers.
Going to the Ray Allen/Bulls example: Yes, adding Ray Allen at SG would help offensively even without Allen having the ball because defenses would have to adjust their approach to account for him. This would create more advantageous spacing (and fewer double teams) for a guard like Derrick Rose to operate, opening up driving and passing lanes, and creating easier shots for teammates. (And, look at Steve Nash or cleveland-Lebron for great passers/playmakers able to make their teammates score more efficiently than they would otherwise, increasing offensive returns.)
The other problem, I think, with seeing limited returns in is tying offensive skills so directly to "creating/scoring" ability, (which, yes, is the point, so to speak) when there are ways that players can impact offensive outcomes without having the ball.*
Clearly, a team with an AI-Kobe-Melo backcourt would suffer from diminishing returns, as all are, as you say, "'go-to' scorers". They create offense by holding the ball and scoring it themselves- and that's their primary offensive contribution. But what if, starting with a PG with great court vision and a traditional go-to scorer, you surrounded them with: (instead of more pure scorers) a big body that excels at offensive rebounding/put-backs, a set/pull-up shooter on the wing, and a post-up, low-block threat? This "all-offense" combination would probably lead to increasing returns by capitalizing on complementary skills. Or for instance, think about the significance of having great screen-setters on your offense' shooting ability.
The points offensive players are capable of scoring - and how efficient they can be in scoring them - are, to a degree, dependent on the court situation they are in.** Having a combination of players like the one in my all-offense team, could allow those players to maximize their team-scoring ability beyond what their individual average-outputs would predict (re: footnote 1 math)
So while it may be true that only the player with the ball can score, I'm not sure it follows that only that player's offense contributes at a given time. How players work together matters too, as much on offense as defense.
*Instead of scoring, let's try another statistic. If you have a team full of great rebounders, would you really reach a point of diminishing returns because there are a limit to the available rebounds to be secured? If you put Dwight Howard and Kevin Love on the same team, I'm not sure they'd each pull down 14RPG, but they'd get a lot of rebounds...
**Ray Allen is again a great example of this. With Seattle, Ray was basically a 20FGA/24ppg player, as he was the best (and probably only) offensive player on his team. With Boston, he scores 17ppg on 13fga. He doesn't score as many points as he can, but he uses fewer possessions to do so, because of the balance of offensive skills his teammates possess. His efficiency and presence then creates opportunities for his teammates as well.
Unquestionably, a player like Steve Nash increases his teammates offensive abilities and efficiency. And if you were to create a team with Nash, Allen, Danny Granger, Dirk, and Pau, say, you would probably get a very very nice offensive output. But let's say you replace Danny Granger with Melo (a clear upgrade). Even assuming that Melo somehow manages to not royally screw up the offense (cough New York 2011 cough), you still will get a "decreasing" return on that investment, because upgrading from Granger to Melo will net you less of an increase than if you had a team of scrubs and added Melo to it. In contrast, I believe--I'm not totally sure--that if you added Dwight Howard to a team that already had Rondo, Tony Allen, Iguodala, and KG, their defense would improve more than if you added him to a team of scrubs.
ReplyDeleteTo put it this way: there's simply very little you can do, no matter how skilled and well-built your team is, to push their combined efficiency up much above 110-115 points per 100 possessions in the modern NBA; there's just always a limited chance that a player will make a shot once they decide to shoot the ball. Adding other very highly skilled players doesn't negate that unless you can turn every possession into an uncontested layup. It allows you to maybe maximize your skill at shooting by pulling the defense off and getting the right person the ball at the right time, but that person still has a limited ability to make any one basket.
The same is true on defense, to a degree, but imagine the two teams above playing against each other. The offensive team will have nightmares scoring on the defensive team, because the combined ability of the second team to choke off ball movement and protect the paint will force so many outside shots that are outside and contested (now, this is what Dirk does for a living, but still). You trade Granger for Melo, and it will still be a nightmare, because Melo will still taking tough contested shots without good outlet options to the rest of the floor. If you replace KG with Amar'e, it still won't be easy for the other team to score, but they'll abuse Amar'e, which will open up the floor for everyone because the defensive scheme won't be as tight. The gain from a mid-level offensive option to an elite one for the offense makes less difference here--at least conceptually--than the gain for the defense from bad to good.
There might be a problem here with example characters. But even if we substitute Tony Allen, for example, for Danny Granger, the offensive gain that team offense will get will only be expressed in fewer double-teams. It isn't like Melo has a preternatural skill for contested jumpers: if you can contest every shot, no matter who is taking them, they are unlikely to go in (unless that person is 1. Dirk, 2. Lebron this year against Boston).