Bill James' new take on 'clutch' etc. (Salon)

"It's Copernicus saying, 'Wait a second. Maybe the sun isn't in the middle.'"

... is just a smidge sketchy. ("we revolve around the sun" = "getting on-base is the most important thing in baseball", yes?) But this is me being a TOTAL pissant.

http://sabr.org/sabr.cfm?a=cms,c,1251,40

I think I finally grok the "noise" thing (has Alan Schwarz been hired away from the Times?):

KEEPING SCORE
Baseball's Leading Man of Math Has Some Second Thoughts About the Numbers

By DAVID LEONHARDT

ith the score tied, the bases loaded and a division title at stake, Steve Finley walked to the Dodger Stadium batter's box in the ninth inning last October with a grin. "I knew the game was over," he said afterward. On the second pitch, he sent the ball flying through the late-afternoon California sunshine toward the bleachers, and the Dodgers had beaten the Giants.

The grand slam seemed a proper season's capstone for Finley, a veteran outfielder whom the Dodgers had acquired from Arizona before the trade deadline.

When games were on the line in 2004, he did his best hitting.

What you think this says about Finley, and about what he is likely to do for the Angels this season, offers a good litmus test of your place in baseball's ideological universe.

If you believe his 2004 heroics prove him to be a clutch hitter, you belong to the majority party. Call yourself a traditionalist, and know that most managers and players stand with you.

If you think Finley is no more likely to be a big-game performer this year than Alex Rodriguez - who tended to melt in the clutch last year - consider yourself a Jamesian. You share the view of many statistics lovers who have been inspired by the writings of Bill James. You consider clutch hitters to be the Loch Ness monsters of baseball, because you know that players who outdo themselves in the clutch one season rarely repeat the performances during the next.

The debate had not changed much for years. Then Bill James himself announced that he was thinking of switching parties.

In an essay published this winter - in something called The Baseball Research Journal, a Jamesian outlet if ever there was one - he argued that clutch hitters might indeed walk the earth. Not only that, but he also proceeded to challenge a handful of other numbers-based doctrines, including the belief that hot and cold streaks do not exist.

James, who made his name writing the "Baseball Abstract" books and currently advises the Red Sox, said he no longer had faith in the numbers he and others long used to make their arguments about clutch hitting and the like.

The statistics now seem far too noisy to him - based on too little data - to trump ideas with an inherent ring of truth to them.

"I was wrong about something, wrong about something important, for a long time," James said by e-mail last week. "And since I had contributed heavily to creating the problem, I realized that I had to do what I could to address it."

The central idea behind much baseball analysis is persistence.

Hitting for average, hitting for power, drawing walks, striking out batters and stealing bases are all obvious skills because the players who do them well one year tend to do them well again the next.

But if there is no pattern from one season to another, Jamesians argue, there is probably no underlying skill.

Steve Finley, as it happened, hit worse with runners in scoring position during 2003 than he did when nobody was on base. Did that make him a choker? Or did 2004 make him clutch?

Neither, statistics mavens say. He is just a good hitter who sometimes succeeds in big situations and sometimes fails. You do not get one head and one tail every time you flip a coin twice, and Finley will not replicate his career statistics in every given situation.

There seems to be no persistence, in other words, to clutch hitting. "Here today, gone tomorrow," James wrote in the essay. The same can be said of a handful of other traditionalist chestnuts, like streakiness. The problem, James now says, is that his old argument rested on a shaky stack of statistics.

Imagine if a doctor checked for breathing problems once a week by placing a stethoscope over a patient's winter coat and listening for changes. Every exam would be flawed - filled with noise, as statisticians say - because of the coat. But the comparison of one exam with another would be even worse, with any small differences between them overwhelmed by uncertainty.

In baseball, luck and randomness - weather, ballpark dimensions, the pitcher - play the role of the winter coat. And the search for clutch hitters involves not just one comparison that compounds the statistical noise. It has two: the differential between a player's normal and clutch batting averages and the difference between this differential across seasons.

This messiness could disguise any clutch hitting.

"We ran astray because we have been assuming that random data is proof of nothingness," James wrote, "when in reality random data proves nothing." (The essay, "Underestimating the Fog," is available at sabr.org.)

But the richest thing about James's mea culpa may be the conclusion worth drawing from it. Just as common sense says that some people handle pressure better than others, it seems likely that these differences are not large among major leaguers.

True chokers will flame out well before reaching the big leagues. Players who discover an edge in clutch situations, meanwhile, will almost certainly try to use it at other times, too.

If it exists, clutchness probably creates only a few extra hits for a batter over the course of a season, despite announcers who claim to see it in every game.

So James's old view might have had the rare distinction of being wrong and still being closer to the truth than the other side's argument.

Copyright 2005 The New York Times Company

Note: this is pretty demonstratably false in other sports (see: tennis, for example) so I see no reason why this would be necessarily true in baseball.

http://www.gladwell.com/2000/2000_08_21_a_choking.htm

(xposts)

Ankiel, Sax and Knoblauch are all examples of guys who had 'yips' THROUGHOUT THE SEASON, not partic in 'clutch' circumstances. (true that Ankiel's problem started in the playoffs, but it persisted -- seems you'd have a start on a case for him if he'd returned to normal the next spring.)

The whole issue is one of sample sizes. Essentially, James is saying that they thought they were using sufficient sample sizes, but in actual fact he now believes they are too small. "There's no such thing as a clutch hitter" = "in a given situation, a hitter's performance will reflect his career numbers". So in that sense, you want Jeter at the plate instead of Womack, because Jeter's career OPS is 850, whereas Womack's is 680.

The more at-bats a guy has, the more sure we are of his abilities. So if his OPS by year goes 850, 760, 890, 1010, 870, 860, 880, it becomes obvious that the 2nd and 4th years were abherrations, and his "true" ability lies in the ~850-880 range.

Now, the random variation in a given measurement goes as the square root of N, where N is the number of trials. The so "relative error", so to speak, is R = sqrt(N)/N = 1/sqrt(N). So, as N grows, the uncertainty in a measured quantity associated with N becomes smaller. However, 1/sqrt(N) is a very slowly increasing function for large N. For instance, we know that over 100 plate appearances, there can be a lot of random fluctuation. For N = 100, R = 0.100.

So if you take a larger sample size of 500 plate appearances, then R = 0.044, a big improvement. But if we take five times *more* PA's, N = 2500, then R = 0.020. We've run into diminishing returns -- there's a much more significant reduction in error when going from 100 to 500 PA's then there is in going from 500 to 2500 PA's.

To summarize:

1. in order to minimize a given random error, one MUST use larger sample sizes

2. it may not be possible to obtain sufficiently large and appropriate sample sizes when applying these methods to some of the problems (numbered 1 through 9) discussed in James' article.

2005:
Category	G	AB	R	H	DO	TR	HR	RBI	BB	SO	SB	CS	OBP	SLG	AVG	TB	SF	SH	HBP	IBB	GDP
RISP W/TWO OUT	13	13	1	2	0	0	0	3	1	3	0	0	.214	.154	.154	2	0	0	0	0	0
RUNNERS ON	19	53	13	13	3	0	2	13	2	10	2	2	.273	.415	.245	22	0	0	0	0	1
BASES EMPTY	17	29	2	10	2	0	2	2	4	7	0	0	.424	.621	.345	18	0	0	0	0	0
BASES LOADED	5	3	1	1	1	0	0	3	1	0	0	0	.500	.667	.333	2	0	0	0	0	0
LATE INNING PRESS	5	5	0	0	0	0	0	0	1	2	0	0	.167	.000	.000	0	0	0	0	0	0
LIP-RUNNERS ON	5	4	0	0	0	0	0	0	1	1	0	0	.200	.000	.000	0	0	0	0	0	0
WITH NO OUTS	16	29	5	7	2	0	2	5	2	8	1	0	.290	.517	.241	15	0	0	0	0	0
2004:
RISP	136	157	79	39	4	1	8	66	24	33	10	2	.346	.439	.248	69	7	0	3	6	8
RISP W/TWO OUT	81	68	22	14	3	1	4	23	9	17	2	0	.308	.456	.206	31	0	0	1	1	0
RUNNERS ON	152	290	94	86	9	1	18	88	38	62	28	4	.378	.521	.297	151	7	0	4	6	18
BASES EMPTY	152	311	18	86	15	1	18	18	42	69	0	0	.373	.505	.277	157	0	0	6	0	0
2003:
RISP	132	145	75	40	8	2	8	63	26	30	2	0	.380	.524	.276	76	6	0	2	10	6
Category	G	AB	R	H	DO	TR	HR	RBI	BB	SO	SB	CS	OBP	SLG	AVG	TB	SF	SH	HBP	IBB	GDP
RUNNERS ON	155	272	96	74	12	4	19	90	42	58	17	3	.376	.555	.272	151	6	0	7	10	16
BASES EMPTY	156	335	28	107	18	2	28	28	45	68	0	0	.412	.636	.319	213	0	0	8	0	0
BASES LOADED	35	7	15	4	1	0	1	13	0	0	0	0	.500	1.143	.571	8	1	0	0	0	0
LATE INNING PRESS	60	63	16	18	2	2	9	22	11	17	2	1	.392	.810	.286	51	0	0	0	1	1
LIP-RUNNERS ON	42	32	12	9	1	2	5	18	3	6	2	1	.343	.906	.281	29	0	0	0	1	1
WITH NO OUTS	130	178	32	52	9	1	14	27	29	31	2	0	.395	.590	.292	105	4	0	4	0	5
WITH ONE OUT	135	191	43	51	7	3	13	42	29	50	4	2	.374	.539	.267	103	2	0	5	3	11

also, novotna simply tanks games sometimes, like on purpose.

Having a "bad streak" is not the same as choking. And you can't hypothesize on the reasons for failing absent consistent evidence of persistent situational failure. Is that it, Glinda?

I miss him too, much as he irritated me at times (in a good way), he was a real fan. Maybe he'll come back when he feels better.

Neyer discussed BJ's recent assertion that groundball pitchers are overrated:

http://www.baseballnation.com/2013/3/21/4132110/groundball-pitchers-analysis-projecting-injuries

interesting. i guess i always kind of read the stats-based preference for ground-ball pitchers as almost assuming the comparison was between two marginal pitchers, one of whom had ground-ball tendencies and one of whom had fly-ball tendencies. like if you're looking for a 4th starter there might be a more readily available case for the ground-ball pitcher.

I've got something in common with Neyer: when I quote James from his site, I also get rid of his weird practice of separating sentences with three spaces.

Are 1,000 plate appearances enough of a sample to draw any conclusions about a specific player?

http://www.highheatstats.com/2013/06/robby-canos-rispy-business/#more-13753

I'd like to see these same numbers on Jay Bruce, as that seems to be the biggest local knock on him. Mind you this is a day after he hits a walk off and also hit a walk off dinger to clench the division a couple years ago.

Bruce has 859 plate appearances with RISP, just short of the 1,000 needed to qualify for the study I linked to, but here's where he's at:

Career BA = .258
RISP BA = .250

96.9%--that'd put him tied for 18th on the post-1961 list.

His OBP goes up with RISP (.330-->.359), his SLG goes down (.483--.438).

I'd like to see the overall totals for all players for some context.

I've been scanning some old Radio On stuff; here's a long thing I wrote on Joe Carter in 1994.

http://docs.google.com/file/d/0B7SjjGanCdBAZHNCbHFoNGZ0NHM/edit?usp=sharing

If you can look past the HOF prediction--which, in 1994, seemed reasonable (James thought he'd be elected too)--I think it holds up well. The funny thing is, Carter did go on to two more 100-RBI seasons, one where he was fairly mediocre ('96) and another was he was plain lousy ('97). But by the time he appeared on the HOF ballot, none of that mattered at all.

https://docs.google.com/file/d/0B7SjjGanCdBATV9hUUVwZzRUdjQ/edit?usp=sharing

Sorry--got rid of a name in there.

Free article on postseason experience as a predictive factor:

http://www.baseballprospectus.com/article.php?articleid=22039

We've discussed this before: how much weight (if any) should be given to post-season performance in HOF voting?

http://espn.go.com/blog/sweetspot/post/_/id/41578/should-postseason-matter-in-hof-voting

Some interesting matched pairs--never realized how close Schilling and Brown were in ERA/W-L.

What I said before (in connection to Billy Wagner, I think): should not be counted as a negative, worth giving a little consideration to when talking about a close call (e.g., if you think Beltran will be a close call--some people feel he should go in regardless).

(And only then in a relatively large sample size.)

James has been running a series all week on big-game pitchers--primarily based on the regular season, with a little weight given in the final rankings to postseason play. The whole thing was triggered by, who else, Jack Morris.

The first couple of posts defined a big game according to Big Game Scores, the components of which are fairly straightforward, although the math isn't...I glazed over a bit. Games are either defined as Big or not big--there's no sliding scale.

Today, in part seven, he listed his Top 11 since 1952:

1. Roy Oswalt
2. Bob Gibson
3. Randy Johnson
4. Don Sutton
5. John Smoltz
6. Johan Santana
7. Andy Pettitte
8. Ron Guidry
9. Whitey Ford
10. Bruce Kison
11. Mike Mussina

Jim Kaat got a whole post to himself--he just missed the top of the list, contrary to a common perception that he wasn't good in big games. (James seems to be advocating for Kaat's HOF case being taken up again by the Veteran's Committee.)

Morris is being saved for the 10th post.

But what happens in Big Games is important whether or not it is indicative of an underlying skill. Bill Mazeroski’s home run in the 1960 World Series is a big deal, whether or not it had anything to do with Mazeroski’s ability as a hitter. Madison Bumgarner pitching 8 shutout innings in the 2010 World Series and 7 shutout innings in the 2012 World Series is important, whether or not it has anything to do with Bumgarner’s character, his underlying skills, or the allegation that he has a girl’s first name and is a bad gardener.

At risk of offending 70% of you to illustrate the point...Lee Harvey Oswald is historically important, whether or not he was a good shot with a rifle. Underlying skills are not always the issue; sometimes the issue is simply what he did. Underlying skills are important in the winter, when you are putting together next year’s team--but when you are looking back at last year’s team, what matters is performance, not ability.

I did not expect Oswalt at #1. Can you explain his methodology a bit more? His postseason record is nothing special, so I guess he gets a lot of credit for winning games for the '04-'05 Astros and the '10 Phillies?

I got lost on the methodology:

I intended for this to work out so that we could use the SQUARE of the Virtual Elimination Percentage, rather than raising it to the power 1.80, but...raising it to the square turned out to be just too damned tolerant, too lenient. (Thanks, Obama.) Using the square, rather than the power 1.80, a team was not virtually eliminated if they were 77-63 when another team was 90-50. My judgment is that, in that situation, you ARE virtually eliminated. 77-63 against 90-50...you’re absolutely not winning.

Anyway, the list is based almost exclusively on regular-season big games, with a small adjustment in the final rankings for postseason. In regular-season big games, as determined by James, Oswalt was 46-12. Not sure of his ERA...he usually provides that, but for some reason he doesn't for Oswalt. But Gibson, the #2, was 36-14/2.26, so I assume Oswalt's ERA was impressive too.

Bruce Kison!

http://www2.readingeagle.com/REnetImages/2010/05/09/9059071/Article_9059136.jpg

http://www.tradingcarddb.com/Images/Cards/Baseball/72/72-72Fr.jpg

Was actually watching when he pitched his great relief game in the '71 Series: 6.1 IP, 1 hit, 3 K, 0 BB.

The culmination of the whole series; it's long, but I'll excerpt the beginning and end:

OK, we have come, at last, back to the can opener which opened this particular can of worms, which is the question: Was Jack Morris, in fact, a Big Game Pitcher?

He was not.

He had the one brilliant post-season, of course, but other than that one three-week period he absolutely was not; it is not questionable, it is not debatable, it is not unclear. It does not seem likely that the conclusion could be altered by studying the question in a different way. Jack Morris did not have a great or even good record in Big Games, and the people who believe that he did believe that because they believe that, but not because there is any actual evidence for it.

In the games that our system has designated as regular season Big Games, Jack Morris made 46 starts, won 18 games, lost 19, 3.79 ERA. His teams were 24-22.

If you use a higher standard for what is a "Big Game", his record gets worse; it goes down to 10-14, although his ERA improves to 3.51. His teams won 14 of his Biggest Game Starts, lost 15...

If you want to advocate for a pitcher being in the Hall of Fame based on his performance in Big Games, advocate for Ron Guidry, or Jim Kaat, or Mickey Lolich, or Mike Mussina.

Yeah, but if your standard is "Game 7 of the WS" then he was the best Big Game pitcher of all time. Where's that HOF plaque?

Sure, he got destroyed in the '92 postseason (losing every game he pitched) but that was only because he knew the Jays had three other really good starters so he didn't have to be as great. Just like he pitched to the score, he pitched to the team's abilities! He sucked because was saving his best stuff for the '91 WS, when he knew the team needed him most (he traveled back in time for it).

he had his time before that time, is what you're saying?!