is he saying that character is a non-factor? that it exists only in the mind of talk radio hosts and sportswriters? well, let's take it a step further Mr. Iconoclast. Let's just reject skill as a factor. Why discuss anything as messy as skill? Let the talk radio and sportswriting vermin stoop that low. We'll just elevate ourselves above such silliness. Skill? We can't let something as vague as that enter into our life as a stat geek.

Conclusion: If you're usually a good hitter, you're usually a good hitter in the clutch.

Knowing that Juan Pierre performs poorly in the clutch doesn't mean a whole lot to me.

Big Papi, Chipper, Pujols come through when it counts

And, whaddaya know, also when it doesn't.

I hate it when Bill gets like this. He was the same way defending bunting. For YEARS, he was writing "We can't just say it's a bad play because the average run expectation goes down. We need to look at the situations in which people actually bunt, and see what exactly happens." Well, then frickin' do it! The data's easily available. It's not at all a complicated study. In the time you spent telling us "somebody needs to do this", you could have been doing it.

Conclusion: If you're usually a good hitter, you're usually a good hitter in the clutch.

Knowing that Juan Pierre performs poorly in the clutch doesn't mean a whole lot to me.

That's simplistic. What we should be looking for here is how much better or worse a player is than his overall average. I think if Juan Pierre (or whomever) is significantly worse in the clutch it is potentially meaningful. And if David Ortiz or Albert Pujols is even better that is also potentially meaningful.

We need to look at the situations in which people actually bunt, and see what exactly happens." Well, then frickin' do it! The data's easily available. It's not at all a complicated study. In the time you spent telling us "somebody needs to do this", you could have been doing it.

Yeah, this James guy has never studied anything. Just keeps making all these remarks and not following up. :-)

Sarcasm aside, that's what he's done in these particular areas.

It's interesting that he cites C. JOnes as a good clutch performer. Jones' BA in clutch situations over the last few years has been .354, .211, .393, .263, and .333. I don't see a lot of consistency.

Doing a quick check Ortiz comes out higher in the clutch than other times in 2002 through 2006 with 07 barely missing out based on OPS 1.042 vs 1.066.

Juan Pierre? Worse in '02,04,06,07 Better in 03,05. However his overall OPS was 607 in clutch over 02 to 07 while his career OPS is 722.

Interesting study. Appears clutch PA's are about 1/2 as often as vs LHP. Regular players will get around 50 AB's a year vs over 450 in 'non-clutch' situations.

If I ran a team I'd probably look at this stuff as useful when debating between two fairly equal players but doubt it would mean anything to me if there was a spread in the first place (ie: wouldn't take a scrub who does well over a star who does poorly). A manager could use it in postseason to determine between bench guys perhaps.

A followup study would be to see if it affects career length or career shape - do clutch guys play longer? Logically if they really get up for those moments they might stick it out longer (can't get that rush anywhere else). What about pitchers? Do clutch pitchers exist and how big a spread is there? Could be very useful for deciding on closer vs setup. Also, is clutch a career thing or a blip? Are spreads big enough to not be random chance? Are certain types of players more likely (outside of superstars) to be clutch?

I suspect a lot of these are questions James has some answers for but cannot release as the Red Sox would be using that data to have an edge on their opposition.

Sarcasm aside, that's what he's done in these particular areas.

I'm not sure I follow you WRT clutch hitting. Isn't this article about the research he's doing on the subject? He's set up an operational definition of "clutch" and run the numbers. I would guess from the end of the article more on this topic is to come next year in that book.

As for the bunting, you may be right. I'm just thankful for all he's done.

I would think that an important aspect of studies examining clutch hitting would be to figure out whether there are different types of pitchers pitching in these situations. Obviously it'll likely be better pitchers, but might it also be power pitchers more than finesse pitchers? It may simply be that good fastball hitters tend to be better clutch hitters because a lot of the pitchers in close and late games tend to have good fastballs.

We made up a system giving weight to each of these seven factors; not saying it's perfect, but you have to start somewhere

Did I miss something, or does James not say what weight is given each factor? (Perhaps you have to buy the book to find out -- I've had that experience before with him :)

Basically, the weighting of the factors is the whole problem. There are about a bajillion permutations in how you can weight such factors (or others you might consider too: like, how many people are in the stands, park factors, splits of various other kinds (platoon, matchup, time of day/night, fatigue, recent grumblings in the press about one's chokiness: "opposition" alone in James's list is a very complicated, multiple factor). The central problem, as I've probably said before, is that clutch is not a yes/no thing; it's a continuum from Tampa Bay playing Texas in April (actually, TB vs. TEX in September is probably the least clutchy; in April there's a lot at stake yet), to Francisco Cabrera looking at a pennant if he gets a hit and no pennant if he makes an out. It isn't just that the former situation isn't clutch and the latter is; it's more like people have no ready way of arguing whether the latter is five, or ten, or a hundred, or a thousand times more clutch than the first. Very few people have ever had the extreme opportunity that Bill Mazeroski had -- so we really don't know if they would come through in the infinitely weighty situation that Mazeroski faced in Game Seven in 1960.

And so when people talk in bars or on BTF about "clutch," the problem is one of weight, and it means that there is no ready way to unify the discussion. You can say, "according to my ad hoc system for weighting clutch events, clutchness exists and these guys show it," and someone else can say, "according to my ad hoc system, no clutchness exists and/or different guys have it to a different degree." You can certainly say that for a certain set of parameters certain players seem to demonstrate clutch ability over and above their already-good abilities, and since that looks like what James is doing here, it would be nice to know the parameters. And of course, if a lot of people's special systems keep converging on the same answers, that's a useful indicator.)

(Also, given the way Win Shares was constructed, it would be interesting to know whether James's parameters were refined by seeing whether or not they pulled up David Ortiz as a good specimen. From what I can tell, Defensive Win Shares in particular seemed to be devised by trying out different hypotheses and seeing if they matched known good fielders, and the balance of offense to defense in Win Shares was arrived at by trying different mixes and seeing which one caused unspecified "problems.")

From the article:

We made up a system giving weight to each of these seven factors; not saying it's perfect, but you have to start somewhere.

Here is James, again, apparently acting as though no one has done any work along these lines before. The first four of these seven factors make up Tango's Leverage Index, and of the other three, I'd be interested in seeing how you could develop a "fair" set of weights for them. I'm sure that Jon Rauch, pitching to Paul LoDuca in the 9th inning with a one-run lead and Endy Chavez on 3rd with two outs in the last week of the season, isn't feeling any less pressure than LoDuca because the Nationals have been eliminated for a long time.

-- MWE

Also, given the way Win Shares was constructed, it would be interesting to know whether James's parameters were refined by seeing whether or not they pulled up David Ortiz as a good specimen.

Based strictly on leverage - which was NOT refined with Ortiz in mind - Ortiz is a very good specimen.

-- MWE

Do clutch pitchers exist and how big a spread is there?

I've always thought yes, surely there could be clutch pitchers. For example starters are known to pace themselves, as evidenced by closers being "better" rate-wise than starters, without actually being better pitchers. It seems obvious to me that pitchers need to pace themseleves, or their arms will fall off even more quickly than they already do.

Thus, the best candidate for a clutch pitcher might be one who has enough talent to cruise along effectively at 90% effort, and save his 100% stuff for clutch situations.

For hitters, there is no such mechanism, there is no reason for a hitter not to give 100% on every at-bat, so I'd expect less evidence for clutch hitters.

As always, I leave the actual study up to someone else :)

is he saying that character is a non-factor? that it exists only in the mind of talk radio hosts and sportswriters?

No, but a post-hoc attribution of a certain character trait because a player got a hit in a clutch situation or failed to do so is not a true character analysis. Judging a player's character because he beats his wife (Julio Lugo), drives drunk (Tony LaRussa), or chases trick-or-treaters with his car (Albert Belle) is fair. Judging a player's character because he gets a hit or fails to do so has to do with a combination of hitting skill and situational luck, not character.

What we should be looking for here is how much better or worse a player is than his overall average.

Right. Am I missing something, or did James not tell us how these players performed in "non-clutch" situations? David Ortiz has good clutch numbers and Juan Pierre has bad ones? That's not very surprising.

James also doesn't define clutch so that we can see how he's breaking the numbers down.

And why does he think that the first "logical path for the discussion" as set out in his column wasn't tried?

Finally, the problem with breaking things down into things like opposition, standings, and the calendar is that not only are these things very slippery in terms of being able to define them, but the further you break the categories down the smaller the samples get and the harder it becomes to pick up the signal with all the noise.

I love James's writing but I think he's contrarian sometimes just for the sake of being so. He's right that if we can't prove that X exists, it doesn't necessarily mean that X doesn't exist -- but if we can't see X, then I don't know why we should care about it even if it does exist.

Here is James, again, apparently acting as though no one has done any work along these lines before.

Am I wrong or didn't he admit somewhere recently (in the Win Shares book?) that he's not really up to speed on the more recent analysis.

Mathematics aside, say you're watching the 2005 or 2006 ALDS (Yankee version). After you've watched A-Rod for the first two games, what would you base your "prediction" for the rest of the series on? What you saw of him in the first two games? Or what he did in the regular season? And why?

My point isn't that "clutchiness" (or the lack of it) is inalterable. It's simply that it's not a heresy to note some fairly obvious cases of athletes tightening up, and producing poorly, under pressure. And in the short run, often there's even a fair amount of predictibility in this---again, see my above example. To think otherwise is to think that athletes are oblivious to both memory and negative reinforcement.

None of this has anything to do with "character", BTW. Character is demonstrated by voting against any candidate who plays demagogic games against illegal immigrants.

Mathematics aside, say you're watching the 2005 or 2006 ALDS (Yankee version). After you've watched A-Rod for the first two games, what would you base your "prediction" for the rest of the series on? What you saw of him in the first two games? Or what he did in the regular season? And why?

On the regular season, of course. The "why" is simple. What Pedroia said.

What we should be looking for here is how much better or worse a player is than his overall average.

Not exactly. What we're interested in is how his performance changes, compared to how the typical player's performance changes in similar situations. If a typical player performs, say, 20 points worse overall, then a player whose performance doesn't change at all has done better than the norm.

-- MWE

What Pedroia said.

Do I find this in The Quotations of Chairman Dustin, or do I just ask Kevin?

Do I find this in The Quotations of Chairman Dustin, or do I just ask Kevin?

Well, I couldn't find his quote after a brief search, but it was after he opened the playoffs 2 for 13 or whatever. He said something along the lines of "I just hit .317 in over 500 at bats in the major leagues in the regular season, and so 15 poor at bats in the playoffs isn't going to tell me I suck."

Pedroia evidently knew how to ak-shen-chu-ate the positive. That's a pretty good talent in itself in clutch situations.

Character is demonstrated by voting against any candidate who plays demagogic games against illegal immigrants.

Since that would rule out voting for almost all known politicians, I guess what you are telling us here is that you are an anarchist.

In 2006, the last year for which I have data, there were 288 players who had at least 30 PAs in high leverage (LI>=2.0) situations. As a group, those players hit .280/.349/.455 overall, and .281/.362/.442 in the high leverage PAs - virtually equivalent production, shaped differently.

Of that group of 288 players, 141 showed a gain in OPS in high-leverage PAs, while 146 declined and 1 (Ray Durham) held steady. The largest gain was posted by Kevin Mench, who hit .284/.338/.459 overall but .423/.500/.769 in high-leverage PAs. The largest decline was by Shea Hillenbrand in his SF incarnation. Hillenbrand hit .248/.275/.415 overall for the Giants, but .071/.121/.071 in his high-leverage PAs.

Ortiz ranked #30 on the list, hitting .385/.507/.750 in high-leverage PAs against .287/.413/.636 overall. Pujols was 4th overall, .415/.542/.943 in high leverage, .331/.431/.671 overall. ARod was near the bottom, hitting .225/.333/.380 in his high leverage PAs vs .290/.392/.523 overall. Jeter, on the other hand, was just ahead of Ortiz at #28, .382/.514/.600 in high-leverage vs .343/.417/.483 overall.

There was virtually no correlation between OPS, or OPS+, and the difference between the hitter's performance in high-leverage situations and his overall performance. The R^2 for OPS was nearly zero; OPS+, around 0.08. You had some relatively bad hitters, like John Buck and Johnny Estrada, performing very well in high leverage situations, and some very good ones, like ARod and Manny Ramirez, performing poorly.

This was, of course, only one year's worth of data, and I'm sure there is a lot of Y-T-Y variance for individuals.

-- MWE

Well, I couldn't find his quote after a brief search, but it was after he opened the playoffs 2 for 13 or whatever. He said something along the lines of "I just hit .317 in over 500 at bats in the major leagues in the regular season, and so 15 poor at bats in the playoffs isn't going to tell me I suck."

Yeah, I remember that. As an aside, I was at a SABR regional meeting last Saturday and some octogenarian described Pedroia as Eddie Stanky, if Stanky could hit. I don't think that Pedroia will ever walk as much as him, but they seem to be attitude twins.

I am extremely reluctant to impugn the character of any player based on what could be a random data outcome...

except when I feel like doing it. Then it's okay.

When James first went to work for the Red Sox I remember that he disappeared from public view for a while. Now he has a book coming out and according to the SI story a new website. Does this mean that his Red Sox commitments are less, or less proprietary now? I know Voros has separated from the Sox, anyone know what James' status is?

Given that a batter is usually facing better relief pitchers in the clutch, you'd expect virtually all batters to do somewhat worse in the clutch than they do in non-clutch situations. In other words, if a batter performs the same, or is only slightly worse, in the clutch than in his regular at bats, you could argue that it's evidence of "clutchiness".

15:
He's right that if we can't prove that X exists, it doesn't necessarily mean that X doesn't exist -- but if we can't see X, then I don't know why we should care about it even if it does exist.

But I think James's point here is that we haven't successfully measured/observed clutch hitting yet, not that we are unable to observe/measure clutch hitting ever.

Maybe we shouldn't care about things we cannot possibly observe. But we shouldn't automatically dismiss things that we haven't observed when we haven't exhausted the means of looking for them.

Also, I agree with nycfan and celadus: the quality of pitching a batter faces in "clutch" situations (however one defines those) should be taken into account, since it's probably significantly different than the overal quality of pitching faced.

<blockquote>you can say, "according to my ad hoc system for weighting clutch events, clutchness exists and these guys show it<.blockquote>

Which is why I always invite the other person to define clutch and then run the study with their definition.

Roughly 99% of the time the person responds to the request for deinition with some version of Potter Stewart. (As Tom Lehrer said, "I know what it is and I like it")

However, every time somebody's come up with a definition and I've run a study:

a) I've always found a few players who've met any given definition

b) Always about the number you'd expect if it was strictly random (generally one or two fewer than you'd expect who'd qualify as chokers. From what I can tell poor results in the clutch by any definition can kill a marginal career)

c) You could never identify clutch hitters in any kind of timely fashion.

d) I've never found it to be terribly important. Paul Molitor's record over the last 15 years with RISP is unlikely to have been random, but it also produced ~2 runs a year for his team as best I can tell. (Over and above what his counter stats told you)

On the last point, I've been meaning to look for players whose teams consistently beat their pythags. Chris Dial dug up one -- old Pete Rose. Seems to me that if there's an important clutch ability it pretty much has to manifest itself in the team's record. I can't bring myself to care if the impact is hard to find and/or minor.

Oh yeah, James has written any number of times that character is important. Just doesn't like it being invoked as an explanation for carrying a stiff. (See for instance his comment about Sparky Anderson on the importance of Enos Cabell's attitude. And yes, I'm aware that Cabell turned in two OK years directly after James' comment. They were a long time coming)

I've raced competiviely for over 30 years, first running and now triathlons, and I have also been a fairly consistent purveyer of all kinds of televised sporting events. Lately I have anticdotally observed that the essence of sport is a test of wills, one person or team exerting their will on another. This is probably less true in pure power, or one dimentional contests, but more so in events that require a more sutble mix of physical skills and mental focus, such as batting or pitching. It seems that the basic game-on-the-line hitter vs. pitcher mathcup is a classic battle of wills, and honestly I would fairly suprised if there wasn't a large cluch component associated with either performance.

This was, of course, only one year's worth of data, and I'm sure there is a lot of Y-T-Y variance for individuals.

Year to year correlations would give us some insight into how much variance there is.

Isn't year to year correlation the entire point? The fact that a player did well compared to his peers in year X says nothing about whether he has an ability to hit well in the clutch.

Isn't year to year correlation the entire point?

Yes.

Maybe we shouldn't care about things we cannot possibly observe.

If you want to follow up on this line, the religious discussion is taking place in the Gotham Baseball/Milledge thread

It seems that the basic game-on-the-line hitter vs. pitcher mathcup is a classic battle of wills, and honestly I would fairly suprised if there wasn't a large cluch component associated with either performance.

I completely agree. Intuitively it's obvious that clutch performance -- the ability to perform better than one's opponent when both are under extraordinary pressure -- is a real and important component of every baseball player's skillset.

The problem is in finding empirical evidence to support that intuition. And it is the case that after a couple of decades of pretty serious effort, no significant supporting evidence has been located, at least at the major league level. One explanation for this absence of evidence is the myriad sample size obstacles that have been noted in this thread, but nevertheless it is the case that something that would seem to be easy to find has in practice proven remarkably elusive. The paucity of y-t-y correlation is just one illustration.

Intuitively it's obvious that clutch performance -- the ability to perform better than one's opponent when both are under extraordinary pressure -- is a real and important component of every baseball player's skillset.

The problem is in finding empirical evidence to support that intuition.

Absolutely. The difference between a great pitcher (Greg Maddux) and a disappointment (Bobby Witt) might be entirely due to willpower and tactical intelligence. (It certainly wasn't due to arm strength.) But what does that tell you? That you'd rather start Maddux than Witt in a key game? You knew that anyway.

A more interesting comparison is between two guys like Josh Beckett and Jake Peavy. They have almost identical regular-season W-L records, are very close in career *ERA+. But Beckett has earned a rep as a postseason hero, and Peavy keeps blowing his chances at one. Cases like these make observers say, "well, it's obvious you'd rather have Beckett in a big game." But can you actually predict anything about what they'll do next October with any confidence? Peavy's career regular-season record after August 1st is 36-17. Beckett's is 28-22. If you slice up the data a little differently in some arbitrary way, it suddenly doesn't give you much confidence at all.