Has anyone looked to see what all these players' field goal percentage is on game-winning/tying shots in the last x seconds of a game? Those are the shots people seem to remember when talking about "clutch", so why not take a measurement of that? If Kobe is 30-35 in those situations, that's different than if he's 30-100 - but in either case, people may just remember the hits and not the misses.
When I think of a clutch player, I think of an athlete who has mastered time management and can make plays with the same, if not better, focus and mechanics when "timed" as when the clock is not a factor.
Where does he say that Kobe is "definitely" the most clutch player in the league? If anything, he's using Kobe and his late game heroics to demonstrate the flawed methodologies of the various quantitative analyses out there that purport to measure "clutchness." If you have a dataset out there that says Boozer, Maggette, and Marion are better late game options than Kobe or Korver is more "clutch" than Melo, then something is wrong with the variables in your analysis. This is where I think observational analysis is more useful than statistics. Unless you'd rather have those guys take shots at crucial junctures during the game over Bryant and Carmelo. I know I wouldn't.
I agree with this. The same could be said for leadership. Although the concept is a bit more abstract, we always here that so and so is a good leader or he's the leader of the team, but how do you measure that? You don't. Even though there are different ways to lead, people usually recognize leadership when they see it despite the lack of a measurement for it.
It depends on how you measure it and how you interpret it. For example, if those guys did have a much higher shooting percentage than Kobe in the last 15 seconds of games, then based on what do you claim that your observation is more accurate than the measurement? Would you trust a doctor if he said, "Hey, my observation is that you are just fine, even though the thermometer shows that you are having a high fever"?
If his point was that using shooting efficiency at the end of close games as a true measure of clutchness is missing the boat, I don't disagree. He went further -- you can't measure clutch. If that's his position -- "clutch" is an abstract, undefinable, immeasurable quality -- then how can he have such a strong opinion on it? What exactly is "observational analysis" in this case? I take it he believes certain types of events are more relevant than others. Fine, define the criteria by which we should consider certain events and dismiss others, and filter or weight the dataset accordingly.
It's not an exact science, and I think that's the point, but I usually watch most NBA games most of the time and I've seen Kobe make positive things happen over and over and over again in clutch situations (hence the observational analysis). I've seen him do it enough times that I don't need a dataset to tell me that his "clutchness" is at this or that level. Statistics can augment observational analysis, however, but I would never solely rely on them to tell me who to give the ball to at the end. Ultimately, here's the question you have to ask yourself: Who would you rather have shooting the ball in crucial situations if you had a choice - Boozer, Maggette, and Marion or Kobe? Korver or Melo? According to the article, the quantitative analysis tells you that you should go with Boozer, Maggette, and Marion over Kobe and Korver over Melo. I know who I would be giving the ball to.
If you've observed it over and over and over again, it would show up in the data, right? And the data is not biased towards players who get more national tv air time either.
You know, people used to believe that the earth was flat. That was hands down the most reasonable conclusion one could have by "observational analysis." I mean, if you don't believe the earth is flat, just go out to a big piece of land and observe. I can guarantee you that the earth appears flat over and over and over again no matter how many time you look at it. It sounds ridiculous that people on the other side of the world stand upside down.
I don't think your extreme example about the Earth being flat applies here. If a dataset is telling you Kobe is not clutch, but watching him play tells you that he makes positive things happen, which one are you going to go with? Again, here's the question you have to ask yourself: Who would you rather have shooting the ball in crucial situations if you had a choice - Boozer, Maggette, and Marion or Kobe? Korver or Melo? According to the article, the quantitative analysis tells you that you should go with Boozer, Maggette, and Marion over Kobe and Korver over Melo. Not necessarily since there is a dataset that says Maggette and Marion are more clutch than Kobe (hence it's not an exact science), but I'll ask you the same question that I asked Easy: Who would you rather have shooting the ball in crucial situations if you had a choice - Boozer, Maggette, and Marion or Kobe? Korver or Melo? According to the article, the quantitative analysis tells you that you should go with Boozer, Maggette, and Marion over Kobe and Korver over Melo.
I don't know what you're responding to. If the play necessitates a player shooting off his own dribble, I'll take Kobe over anyone. If I have a great pick-and-roll tandem like Williams+Boozer or Nash+Amare and I need to put a spot up shooter at the 3-point line, do you want Kobe in that spot or a more pure shooter? Two problems with that statement. (1) There is no definitive quantitative analysis on this topic. The article refers to one person's interpretation of clutch based on efficiency. I've linked to another analysis by Ryan Parker in NYT Magazine that came up with very different results. (2) The analysis didn't tell the reader that a team should go with Boozer/Maggette/Marion over Kobe. Its defining clutch in a very narrow way (much too narrow, in my view). Its an analysis of which players are most efficient in clutch situations, which doesn't necesarily correspond to players you'd want on the floor.
Its amazing how you can ignore complete and raw data taken then throw in some subjective thoughts, a few names then say those numbers mean nothing because they dont quantify your bias. That said last second heroics and clutch are two different categories and Kobe isnt #1 in either. Then where do you just come to the conclusion of throwing out everything, assists, rebounds, steals, blocks, etc... as if they all dont matter. This article is a complete joke. How about the teams and players stomping the competition where they dont end up in these situations? Do they get credit?
Dangit, this subject is making me review my notes in statistics as well as research methods. Part of me feels that the article writer just doesn't like stats, maybe because he doesn't understand the nuances of statistics. So there's something statistical looking thats goes way against observation, and wants to trumpet it as "SEE STATS IS WRONG EAT IT HOLLINGER" Don't blame statistics for this. Blame the (wrong) interpretation that media claims to be true, because there's no way Morey calculates clutchness the way the article did. The key concepts here are properly defining our variable clutchness, and then properly operationalizing this variable. http://www.experiment-resources.com/operationalization.html Let's say we define clutchness as a high level of production when production significantly affects the outcome of the game. Extending a lead from 2 points to 4 points in the first quarter increases your probability to win by a small amount, but extending a lead from 2 points to 4 points with 24 seconds left? Your probability of winning went from 60% to perhaps 90% (just a guess). I would find this a reasonable enough working definition of clutch that we all can agree on, yes? Problem #1: clutchness, as referenced in the article, isn't even defined. its just a shooting % under certain particular conditions. its just a formula. And since we only have a formula to work off of, we run into our primary problem with the clutch formula as referenced in the article: it extends to the last 5 minutes. The data being gathered (last 5 minutes of game) does not correspond to our mental definition of clutchness (production significantly affects outcome of the game). The variable wasn't operationalized correctly. In other words, the data did not measure what we were trying to measure. We're trying to measure clutchness, but instead we end up measuring shooting accuracy. The whole ruckus the writer is talking about is this improper association. Don't blame statistics for the improper association. Blame NBA.com for touting as truth "data from shooting % is indicative of how clutch a player is". Shooting % is just a result. Statistics did what it was supposed to do, measure shooting %. Its not statistics fault that nba.com came up with a whacky conclusion from shooting % and declare that it is a snapshot of clutchness. Stats don't lie. People just use stats to lie, and for whoever came up with that formula, was probably ignorant of "true" statistics. [end statistcs rant]
I was responding to your most recent post in this thread at the time. But I just thought your only issue with the article was that he supposedly said Kobe was "definitively" the most clutch player even though that statement is nowhere to be found in the article. You're saying you'd take Kobe no matter in any situation that calls for a player shooting off his own dribble. I don't know whether you realize it or not, but you just tacitly acknowledged the limitations of the quantitative approach you're so laudatory of with that statement, especially considering the fact that datasets out there ostensibly declare that others are more clutch. I would go one step further...ask most fans, people in the media, NBA personnel (coaches, players, front office guys etc..) who they believe is the most clutch. I'm pretty sure the majority will say Kobe. Are they all ignorant or are the methodologies used to measure clutch play and the datasets flawed? Also, the fact that statisticians who try and quantify clutch play come to different conclusions demonstrates the arbitrary nature of the quantitative approach. While on the surface it may seem "objective," it essentially boils down to which independent and dependent variables are included in a model (hence the subjectivity or arbitrary methodology). That could lead to a whole host of issues such as data manipulation, selection bias, omitted-variable bias, measurement error, etc... We all use stats in one way or another, whether we look at the box score, a players #s during the course of the season, and/or duration of his career. I know it's the new "in" thing, but I just hope you realize the limitations of the more hardcore quantitative approach to basketball. Don't get me wrong, it has its uses, but it's secondary to the more traditional observational approach in my opinion. It all boils down to what you do when there is a conflict between the two. If the quantitative analysis is saying so and so is more clutch, but your eyes are telling that this other player is better down the stretch - what do you do? I would trust what I see as opposed to abstract and sometimes obscure statistics.
The measure, for example, doesn't take into account if the shot was created for the player, or the player created himself. Players like Korver, is almost always single-covered, and barely shoots if he's not open. But you can't get open in every possession, particularly down the stretch when the defense is tight, but someone has to shoot or it's a TO. Korver can't get as many shots off as Kobe in crunch time, but those shots are more open. Hence you can't simply use per minute stats to inflate Korver's numbers.
Take into account the percentage of shots that are assisted or use the play by plays to determine the type of shots being attempted (it distinguishes jump shots, runners, drives, etc.). I didn't say it was easy, but that doesn't mean its not possible to use stats to get a better grasp of "who's clutch".
