Two nights ago, the Lakers beat the Spurs after three stunning lead changes in
the final few seconds. Last night, the Nets beat the Pistons in triple-overtime.
In any real sense, these games were ties. What is the right way to measure
clutch performance? Is there any stability to clutch performers?
Two nights ago, the Lakers beat the Spurs after three stunning lead changes in the final few seconds. Last night, the Nets beat the Pistons in triple-overtime. In any real sense, these games were ties. What is the right way to measure clutch performance? Is there any stability to clutch performers?
Ties, Damned Ties, and Statistics
Basketball News Services
Two nights ago, the
Lakers beat the Spurs after three stunning lead changes in the final few
seconds. Last night, the Nets beat the Pistons in triple-overtime. In any real
sense, these games were ties. What is the right way to measure clutch
performance? Is there any stability to clutch performers?
I'll tackle the latter question first: is there any stability? In other words, do the same clutch performers continue to come through and make game-winning shots time and time again? There are two ways to answer this question: one empirical and the other logical. Empirically, we can look at what has happened in the past and see if the same people consistently win games in the final moments. Logically, we can try to argue from first principles what should be the result.
As a quick empirical exercise, we can look at the NBA's list of top clutch performances here. The names that come out are: Tim Duncan, Derek Fisher, Robert Horry, Sean Elliott, Allan Houston, Michael Jordan, John Stockton, Rex Chapman, Mario Elie, Alonzo Mourning, Michael Jordan, Magic Johnson, Larry Bird, Ralph Sampson, Gar Heard, Kareem Abdul-Jabbar, and Jerry West. Only Jordan appears twice: once for his shot over Bryon Russell and once for his shot over Craig Ehlo. Only MJ can even be argued to have any stability, at least according to this source.
The NBA's list is obviously flawed. It doesn't list every buzzer-beating shot, it doesn't list any missed potential game-winners, and it clearly is trying to spread names around. That list is intended solely to give fans an appreciation that the magic can happen at any time, from any player, so don't blink. Sometimes it will be by a superstar, and sometimes it will be by a role player. You just never know.
Yet despite the list's deficiencies, its results may still be accurate. It is a daunting task to go through 50+ years of NBA data filtering for all shots at the buzzer, marking those which made it and those which missed, and seeing which names come up most often. And even when all that work is done, the results may be useless.
Here's one reason why: an analysis of buzzer-beating shots by its very nature and definition only looks at close games. Any player that was on a supremely dominant (or supremely terrible) team would never have a near-tie at the end of the game, and thus would never be allowed to display his ability to perform under that particular kind of pressure. Here's another reason why the analysis would be flawed: because such situations happen rarely, it naturally rewards longevity. Jordan's two highlights in the NBA's list are separated by nine years. It is quite possible to be a great clutch performer and not last nine years in the league; the empirical analysis would only display clutch performers who also had long careers.
And now it is time to attempt to answer our question -- is there any stability to clutch performers? -- with logic. At first blush it would seem that logic is the furthest thing from basketball analysis. Analysis should involve regressions and data and graphs. Not necessarily. Those are the tools of a statistician, but they are not the handcuffs of the analyst. Much like a plumber sometimes uses a plunger and sometimes uses a socket wrench, but sometimes, he can just use his finger to jiggle the handle. Tools need to be used in the right circumstance and in the right way. It is our primary question -- what is the right way to measure clutch performance? -- that will address how to use these tools.
For now, let's let logic dictate what it thinks the results should be. Should players consistently be able to hit clutch shots? Imagine repeated trials of the same game, like a Groundhog's Day of basketball. Imagine that both Spurs head coach Gregg Popovich and Lakers head coach Phil Jackson can readjust however they want for that final Lakers possession with 0.4 seconds left on the clock. (In fact, they both did readjust multiple times, as it took three timeouts before the final possession was played out.)
In naively defended possessions, either Shaq or Kobe would get the ball for a quick shot, layup, or dunk. By naively defended, I mean that Popovich makes no adjustments, and tells his players to just play one-on-one or zone or something very standard. Then it would be fairly routine for Jackson to create a quick play, perhaps using multiple screens, of getting someone open in a good position to shoot. Most likely he'll go to one of his best players, either Shaq or Kobe, to hit the final shot.
If this game were repeated indefinitely without adjustments, Shaq or Kobe or both would make some shots and miss some shots and if they made it more often than they missed, they would be considered consistent and stable clutch performers.
Ah, but there's the rub: adjustments. Seeing that he's getting continually burned by one of those two, Popovich would adjust his team's defense to be sure to deny the ball to either of those two. And the inbounder would have to look for a third option. Suppose that third option is Derek Fisher. Then if we repeat these games, if Fisher makes the last second shot more often than he misses, he would become a clutch performer. At least until Popovich adjusts for him, too.
And there's the key: adjustments. The clutch performers are unlikely, by logic, to be consistent because opposing coaches would adjust to them. Even if it means doubling or tripling the superstar, they'll look to do whatever it takes to keep Jordan or Kobe or Shaq away from the ball. That's why so many game-winning shots were hit by Jordan's teammates: the defense was made to crack down on Jordan, leaving John Paxson or B.J. Armstrong open for a shot.
Logically we've concluded then that there ought to be little if any stability in clutch performance, and whatever consistency we do see is more a result of the cross-effect of longevity and clutch performance. Now we can answer our original question: what is the best way to measure clutch performance?
The key insight is realizing the Derek Fisher was the spark plug that got the Lakers the win, but Shaq and Kobe were the engines that made the opportunity to shine possible. They deserve some sort of credit for being on the floor even though they didn't even touch the ball. Why? Because if they weren't there, their replacements would not be so heavily guarded that Fisher would be the open man.
So how do you reward players who are on the court but don't take the last shot? How do you reward Jordan for all the times his mere presence left his teammates wide open? You do it with net team points.
Net team points is the lead your team has over the opponents. It is sometimes called plus-minus, meaning plus all the points you score minus the points you allowed the opponent to score. I don't like the term plus-minus because it's not self-explanatory. Sometimes plus-minus can refer to the net team points scored while a player is in the lineup, and sometimes it can refer to the net team points scored while a player is in the lineup minus the net team points scored while the same player is on the bench. I don't like adjusting net team points for what happened when a player's on the bench, and I've written elsewhere why (the basic reason is that such a contractual incentive would destroy team chemistry).
The problem with simple net team points, with just looking at how the lead changed when you were in the game, is that it rewards two early points the same as two late points, and it rewards meaningless garbage baskets the same as buzzer-beaters.
So what we need to do is adjust net team points with appropriate weights. These weights must reward buzzer-beating shots above all others, because they are the most important. It must weigh garbage time points below all others. And it must applied on a team basis: it doesn't matter who actually scores the point, just who is on the floor at the same time. By its nature, this won't distinguish Michael Jordan from B.J. Armstrong if they are always on the court at the same time. Usually it will be clear who is the star of the lineup, by other, obvious, statistics.
What is the right weight? It is to use probabilities. More specifically, the right approach is to value the offense and defense as options, and the game as a spread option between two assets. Then we can use the results from financial derivatives theory to calculate the appropriate weights, which would merely be the difference in the value of the spread option as time passes and points are scored.
Look for such an analysis from me shortly. And we will soon learn who the greatest clutch performers really are.