Notice: Trying to get property 'display_name' of non-object in /var/www/html/wp-content/plugins/wordpress-seo/src/generators/schema/article.php on line 52
keyboard_arrow_uptop

Two weeks ago, I wrote an article on Cole Hamels on the day that the Phillies clinched the National League pennant, explaining in detail that I do not believe that there is anything wrong with Cole Hamels, and that the difference between 2008 and 2009 is abnormally good luck in the first and abnormally bad luck in the second. The first clue was that he had similar peripheral statistics in 2008 and 2009. He struck out 21 percent of hitters in both years, and walked just over five percent of hitters in 2009 after walking just under six percent of them in 2008. His ground-ball rate stayed roughly the same, rising from 41 to 43 percent. The difference came from his BABIP jumping from an incredibly fortunate .262 to an incredibly unfortunate .321. It has been shown many times before that BABIP is a statistic with low persistence, and that pitchers see their performances jump up and down constantly with respect to this statistic. As a result, much of year-to-year fluctuation in ERA is tied to fluctuations in BABIP. Unsurprisingly, Hamels ERA went from 3.09 in 2008 to 4.32 in 2009. As Hamels’ peripherals indicate an ERA around 3.65, it seems likely that he had a mixture of good luck in 2008 and bad luck in 2009 that belied his ERA.

This is not at all atypical. Given the number of balls in play that Hamels allowed, the standard deviation of his BABIP should be .020 in any given year, meaning that he was barely in the top ten percent of pitchers with respect to good luck in 2008, and barely in the top ten percent of pitchers with respect to bad luck in 2009. This type of fluctuation is typical. Looking at other pitchers through history, you can see that their ERA often bounces up and down, and in so far as pitchers can’t control BABIP, it’s BABIP that is frequently the culprit.

In my article from October 21, I took this into more detail. I discovered that Hamels allowed no more hitters to hit the ball to the outfield-52 percent in 2008 and 51 percent in 2009. Although Hamels himself later claimed that the problem was a failure to put away hitters with two strikes, citing his hit rate on two-strike counts, Hamels may not know that he struck as many of them out in this situation: 41 percent in 2008, and 40 percent in 2009. Not only that, he induced slightly more popups in 2009 than in 2008. Hitters also pulled, hit balls to center, and hit balls the opposite way with the same frequency as well.

Predictably, this article was received quite well with the Baseball Prospectus crowd, and rather skeptically by many other baseball fans. The value of on-base percentage has slowly gained ground with more casual fans, but analyzing pitchers with an ERA that is belied by their peripheral statistics in the face of very good or bad BABIP is not something that casual fans take to all that willingly. Many writers tried to explain alternative theories, often citing Hamels’ psychology as fatally flawed in some way. They claimed he was not paying attention to the game as much, that he was too obsessed with his celebrity, and when confronted with this logic, I would simply ask why that type of difficulty would not affect his strikeout or walk rates. I still have not heard a good answer, and recently tweeted, “Enough of people who don’t know baseball, Hamels, or psychology explaining the effect of Hamels’ mindset on his performance!”

Meanwhile, my article was cited in the blog of Philadelphia Inquirer beat writer Andy Martino, who understood my point but took issue with my conclusions. He also took strong issue with Rob Neyer’s claim that fellow Inquirer writer Jim Salisbury was ignorant of statistics in claiming that Hamels should pick up another pitch. Specifically, Neyer said baseball writers had a “sick, 20th century obsession with wins and losses and ERA.” Martino quipped, “Isn’t it so, like, 2004 to say that newspaper writers don’t care about statistical analysis?” This is a fair point when it comes to Martino, who I would like to stress is very adept at statistical analysis. However, Martino needs to be more honest about his fellow Inquirer staff writers. The newspaper frequently tried to attribute Ryan Howard‘s World Series performance to psychological problems as well, without even mentioning the left-handedness of his opponents. This is the same paper in which Frank Fitzpatrick wrote that Hamels does not have “enough Philly in him.” In this article, Fitzpatrick lauded players such as Pete Rote, currently banished from baseball in gambling, and Lenny Dykstra, who ran into problems gambling in his own way, who Fitzpatrick claimed were Philly enough. Instead, “Hamels, it’s always apparent, makes his living with his arm, not his hands,” Fitzpatrick claimed.

This last article may have drawn particularly criticism and even some excellent parody, but it is not atypical for the Inquirer. Frequently, statistical analysis is eschewed in favor of sensationalism and character attacks. Much of the common stereotypes of Philadelphia fans are a result of the relentless Philadelphia media, which frequently resorts into character assaults on good players who are not at their best.

The problem, of course, is that it’s easy to attribute a psychological profile to a 25-year-old based on outcome. The difference between a “deer in headlights” look and an “unfazed by his surroundings” look can only be distinguished if there is a scoreboard in the background. Images of Hamels fighting through the downpour of rain during Game Five of the 2008 World Series were frequently shown as testaments to the strength of his character, his ability to ignore the chaos surrounding him and pitch the team to victory. Images of Hamels during Game Three of the 2009 World Series were frequently shown as testaments to his character flaws and used to explain his difficulty in performing as he felt that he should. However, if he had given up a few bloopers in the rain in 2008, the captions would have cited Hamels’ inabilities to toughen through a little water, and those same few bloopers had been caught in the 2009 Series, Hamels would have been lauded for getting a fresh start. The character assassination of playoff goats and canonizing of playoff heroes should be buried as memes, but the Inquirer‘s guilty of frequently summoning them up.

Martino, however, has never resorted to these extremes in my time reading him, and it is why I will take his criticism of my articles seriously. He does know his statistics, and Martino made several claims about Hamels’ season that are strong points, but I do not believe that they take away from my hypothesis. I will address these step by step.


Comment:
He had a sore elbow in March, April, and May. I know that because he told me later in the summer.

Pitchers frequently deal with injuries. This is not the first time in Hamels’ career that he had elbow problems, but he never put up a 4.32 ERA before. To claim that this affected his entire season seems unlikely, especially as his ERA from June through the end of the regular season was 4.03 and his BABIP was .312, so much of his BABIP problems clearly came after this alleged elbow problem was healed.


Comment:
He did not begin training until later than usual, because he took on too many post-World Series commitments.

This certainly explains the elbow injury in March that did lead to a delay in his first start and lowered velocity in that start (throwing 86.9 mph fastballs on average, and averaging about 90 mph in subsequent starts, starting at 89.3 in his second start).


Comment:
Even on days when his pitches were working, he responded poorly to adverse circumstances, and allowed bad innings to snowball. He admits this, and his manager, GM, and coach agree. It is also obvious from watching his body language.

Hamels clearly does not like giving up runs, but I do not believe bad innings snowballed in general. In 2008, he allowed a 627 OPS with bases empty and 714 OPS with men on. In 2009, he allowed a 756 OPS with bases empty and 754 OPS with men on. It appears that Hamels was actually relatively better at stepping up his performance with men on in 2009 rather than allowing them to snowball.


Comment:
Though his velocity was as good as last year, he has to overthrow to get his fastball in the low 90s. That sometimes resulted in poor location and home runs allowed.

This suggestion was particularly intriguing to me, and it’s the one that I went into the most detail to test. Here’s what we do know: Hamels threw 53.1 percent of his pitches in the zone in 2008, and 52.5 percent of them in the strike zone in 2009. He threw 32.6 percent of pitches for balls in 2008, and 32.7 percent of them for balls in 2009. Clearly, Hamels was not missing the strike zone much more frequently at all. I also used the excellent PITCHf/x baseball website brooksbaseball.net, and copied data from each of Hamels’ starts for 2008 and 2009. Brooksbaseball.net has a statistic that is called “Nibbleness,” which they define as “the arithmetic mean of the distance of each pitch, in inches, from the edge of a normalized strike zone. Lower indicates ‘more nibbly.'” This seems like the statistic to check. I weighted both fastballs and changeups by the number of pitches thrown and found that his 2008 average Nibbleness was 5.3; in 2009, it was 5.5. It is very hard to believe that a fifth of an inch explains a sixty-point swing in BABIP that caused many more singles, but that many more doubles and triples, and actually led to fewer home runs in 2009.

This is similar to the claim that Hamels’ changeup was more hittable with respect to movement in 2009, rather than in location. FanGraphs’ listed horizontal/vertical movements on his changeups in the last three years are as follows:

2007: 7.5/7.7
2008: 6.2/7.9
2009: 7.5/8.2

Although I am no PITCHf/x expert, it certainly does not seem to be a good claim either. Instead, what we have is the same two primary pitches with similar velocities, movements, and locations in consecutive years.


Comment:
The lack of a quality third pitch allowed hitters to guess what was coming. Take A-Rod last night: Hamels started him off with a change-up for strike one, so the hitter figured he would see a fastball within the next few pitches. When one arrived on the next pitch, A-Rod was ready, and clocked it into the Jeffrey Maier camera.

Certainly, guessing pitches plays a large role in who wins in a batter/pitcher duel. However, Hamels would certainly have given up more line drives if he were easier to predict in 2009 than 2008. He did not. He would have missed fewer bats in 2009 than 2008 if this were true. He did not. He would have surrendered more home runs (into Jeffrey Maier cameras or otherwise) if hitters could guess right more often. He did not.

What does that leave? Cole Hamels simply had a fairly typical amount of bad luck. Does that mean Hamels should not learn a new pitch? Where is the harm in that? If it helps him to develop a pitch that can make him even less predictable if he throws some nontrivial percent of them, then of course any pitcher should develop a pitch like this. But does Hamels need to? Probably not. But given that Hamels is more likely to be a 3.65 ERA skill level pitcher than the 3.09 ERA pitcher that benefited from a .262 BABIP in 2008, there is certainly room for improvement. Hamels only gets hitters to whiff at his fastball six percent of the time, compared to over 20 percent of the time when he throws his changeup. Perhaps a cutter or slider could add to the deception to increase this whiff rate. However, to say that Hamels will repeat his average ERA in 2010 is to make a pretty extreme claim that is not supported by the evidence at all.

All in all, looking at the evidence surrounding this data, it seems even less likely that something went terribly wrong for Hamels in 2009. He certainly should not be satisfied with his performance, because no pitcher should. Pitchers are constantly working to try to keep hitters guessing, and Hamels had better continue to do so if he wants to avoid losing ground. However, if he does lose ground, it will come with a decline in strikeout rate, an increase in walk rate, and a decline in ground-ball rate.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
ScottBehson
11/09
Could part of it be as simple as the fact he gave up 9 unearned runs last year and only 2 this year? If my calculations are right, that's about a .35 difference in ERA.

However, all his BP scouting reports contain mention of incoinsistency and being very good when good but allowing things to snowball when bad. 2 more snowballs over a course of a season could alter his stats.
TGisriel
11/09
unearned runs are excluded in the calculation of ERA (Earned Run Average)
ScottBehson
11/09
Yes. Perhaps I didn't write clearly enough. What I meant to say is that he had a greater proportion of unearned runs in 2008 than in 2009. This is probably random noise but it is noise that led to a lower era in 08 than 09.
NL2003
11/09
"Predictably, this article was received quite well with the Baseball Prospectus crowd, and rather skeptically by many other baseball fans."

Are statistics now being developed to quantify readers of articles, instead of baseball players?
marooner
11/09
If the Phillies were to call you and ask you to consult for them, would you say "well, Cole should just keep doing what he's doing, and the law of averages should kick in?" If that's your answer, I would say that your statistical understanding is less than complete. People aren't going to believe that 2008 Hamels and 2009 Hamels were the same pitcher just because you can't explain year-to-year variation in BABIP.
swartzm
11/09
You're arguing with a straw man. I very clearly said in my article that I do not believe any pitcher should keep doing the same thing. Not even Tim Lincecum. Any pitcher who does not work at getting better will get worse, and I would suggest to any pitcher working on the game theoretic strategies involved in being less predictable so that hitters do not eventually increase his currently stable line drive rate to a higher level.

However, I should remind you that the standard deviation of a binomial random variable like BABIP with 500-600 observations is going to be around .020. That's a statistical fact. That means that even if pitchers do vary in their BABIP skill level, a third of pitchers will have BABIPs that are .020 points different than their true skill level.

Given that Hamels did not give up any more extra base hits, did not allow any more balls to be hit to the outfield, and struck out and walked the same percent of batters as last year, what would you propose the difference is? And remember that you need to explain why Hamels flaws are only causing more balls to be hit in front of outfielders and not more balls to be hit over their heads nor has it caused the balls to change direction and become pulled more often. Remember that your answer cannot use changes in velocity, movement, or location because those have all been the same.
ScottBehson
11/09
Not to pick nits (yours are excellent articles), but about 68% of pitchers would be within 1 stdev, not just 34%.
swartzm
11/09
Sorry-- I was trying to say that a third (or actually 32%) will be either more than a standard deviation above or more than a standard deviation below. Thanks for clarifying.
dsc250
11/09
Did the Philly defense get worse this year? I know Victorino suffered by many metrics this year (although Ibanez was better than predicted). Could that have played a role?
swartzm
11/10
It could explain a small part. Team BABIP went from .295 to .300 so that might explain some of it. But the BABIP for everyone but Cole Hamels was actually the opposite: .297 in 2009 without Cole and .301 in 2008 without Cole.
BeplerP
11/10
If: (a) his HR rate was not up, (b) his XBH rate was not up, (c) his SO and UIBB rates were the same or better, and (b) his FB/GB ratio was effectively unchanged, haven't you delivered a non-answer to the question? I agree that luck is a powerful (and unacountable) factor, but isn't there some clue in the quality of the defensive support he received? I think particularly of Rollins and Utley, great players, great athletes, but not complete defenders. Is some of the variation to be found there?
BeplerP
11/10
Oh, sorry, Matt, I need to say this was a terrific piece, as good as anything I have read this year on BP, and just what I have come to expect from BP. But I have trouble with BABIP as pure unadulterated Luck! Regards,
Richie
11/10
Darn good question there. But I halfway recall you considering it in the original article?? I think?
Oleoay
11/10
Did the hits and walks clump together more often this year than last year? As in, was he more prone to giving up a big inning, or threw more often with runners on base? Was there a variance in how he performed in the stretch as opposed to the windup?

I don't watch Cole Hamels much, so much of what I've heard about his perceived problems is exactly that... hearsay. Just trying to propose ideas that might show a component of performance variance besides luck.
swartzm
11/10
He did not actually-- check the third to last comment in the article.
Oleoay
11/10
I saw the third to last comment, but I just get a bit confused. OPS is used to show that OPS had some variance in 2008 between his performance with runners on and offbase, and in 2009, the OPS is similar. Yet the conclusion of that comment is "It appears that Hamels was actually relatively better at stepping up his performance with men on in 2009 rather than allowing them to snowball." That comment is made though OPS with runners on and runners offbase increased. I get the idea that OPS should be higher when people are onbase, and I realize OPS does not tell the whole story...
but what is the threshold for determining that Hamels was "relatively better" as opposed to "neutral" or even "lucky", especially if his across the board OPS increased? Perhaps there's a flaw in his windup that attributed to the poorer performance with runners on?

Just trying to determine where the line (or gray area) between luck and skill (or lack of skill) is.
swartzm
11/10
The point I'm making is that the difference runners on vs. bases emtpy between this year and last year is clearly not the explanation for the BABIP difference. His BABIP difference with runners on was exactly +.004 higher both years, less than the league average OPS jump with runners on.
hessshaun
11/10
FRANK TANANA, is his number one comp.

Just send that as your response to the lazy writers, aka, the Philadelphia media.

Interesting Factoid: Did you know, more than half of the Inquirer reporters move onto writing for daytime soap operas?
jmk261
11/10
This article is why I just subscribed to BP. I hate the Inquirer.
beegee73
11/10
I drink the Kool-Aid here every day and as such I agree with this analysis. But I also get most of my Phillies news from the Inquirer, and I think the characterization here of their coverage is not quite accurate. Yes, they've got columnists (Frank Fitzpatrick, who's generally a football guy; John Gonzalez, who's more like the paper's resident jester) publishing some embarrassing pieces and that's, y'know, an unfortunate byproduct of producing material for a mass market. But the guys who do the majority of their baseball reporting and analysis — Martino and Salisbury — tend to be pretty solid and avoid a lot of the tropes that get rightly batted around here. Salisbury's mostly been a breath of fresh air after taking over for Jayson Stark and his baseball/numerology beat. Martino was coming to the defense of Salisbury who, while imperfect, I think got unfairly lumped in by Neyer with the Bill Conlins of the world.

That said, I'm curious to hear your take on Hamels' pitch count splits this year. Last year, he seemed to get stronger as the game went on, posting better aggregate numbers in pitches 76-100 than 1-25. This season he seemed to hit a wall around pitch 50, with his Ks remaining steady, his walks going way down, and his hits, xbh and babip skyrocketing. Is this too small a sample size to be useful?

swartzm
11/11
I'm not sure that much can be learned from this. It's possible that hitters are guessing better the second time around versus the first time around, but it's unclear why that would be the catalyst for such a huge change. The real reason that Hamels declined was an increase in the rate of hitters getting singles, not extra base hits. Since his K and BB numbers didn't really change as he went through the lineup, I'm not sure much can be gleaned from the fact that his extra-base hits, while not really any more numerous this time, were distributed more towards the second and third times through the lineup and less towards the first time through. They did not go up overall, meaning it's probably not the problem.
jdouglass
11/10
Matt, you noted the Fan Graphs data on Hamels' movement on his change, but I think there's more there that led to an increase in BABIP.

Their data points on pitch blend show that Hamels relied more on his fastball than last year, while his curve frequency fell and his changeup frequency dipped. His change was his most effective pitch in 2008 and his only effective pitch in 2009, so the decrease is troubling to me even though it's small. His wFB rating went from 12.8 to -6.3 as he relied more on it. This would seem to point to hitters adjusting and sitting on the FB, and I would think any pro hitter who can accurately guess FB is going to hit the ball better--and harder--which should result in a higher BABIP. Hamels threw his FB 59.1% of the time. Carpenter, Wainwright, Lincecum, Vasquez, Haren--none of them threw any single pitch more than 55.8% of the time--that was Lincecum's FB, and outside of that those five pitchers didn't throw any one pitch more than half the time.

Hitters against Hamels swung at fewer pitches outside the strike zone, and swung at a higher percentage of strikes than in 2008. They made fewer decision-making errors in their selectivity against what Hamels offered. That Hamels was more predictable in his pitch blend and hitters swung at more of the right pitches and less of the wrong ones is a big warning sign.

What we have in Hamels is a three-pitch pitcher who has yet to show sustained effectiveness with anything other than his change, throwing that change less frequently, and becoming less effective in keeping hitters off guard. I think that has as much to do with his increased BABIP and his H/9. Perhaps his increased ground ball count was the real luck in his season, and in giving up 2 more hits per 9 those ground balls were being hit harder and getting through to the outfield on more than luck and defensive efficiency.
swartzm
11/11
The Fangraphs' pitching run weights are about PA outcomes. That they went down is simply parroting the fact that he gave up more runs. His rate of giving up runs per fastball and per changeup both went down by similar amounts. Also, that he threw 3.2% fewer changeups is only 100 pitches different. It could not explain 6% more hits on balls in play. That's basically 100 more pitches that were fastballs instead of changeups somehow explaining 35 hits. Even if hitters batted 1.000 on his fastball and .000 on his changeup, the rate of pitches put into play would make this ridiculously not true.

While I'm aware that Hamels O-Swing% went down and his Z-Swing% went up, the fact that his O-Contact% went up and his Z-Contact% went down by roughly the same amounts implies that hitters rate of putting pitches out of the strike zone into play and rate of putting pitchers in the strike zone into play was about the same both years. It could be a clue if he was whiffing fewer hitters, but that is simply not true.

Further, you haven't answered the question I posed above to another writer which would debunk this as well. If Hamels was more predictable in 2009, why did he not strike out fewer hitters? Why did he not give up more home runs or extra base hits? Why did he not allow more pitches to be hit to the outfield? Why did not allow more line drives? Why would being more predictable only change the rate at which ground balls went between fielders and at which looping flies fell in front of outfielders versus in their gloves? That does not add up. If he was more predictable, hitters wouldn't have whiffed at more pitches and they would have hit more line drives and more home runs. Of course Hamels would be more predictable with a great third pitch-- he'd be even more predictable with a great fourth pitch and fifth pitch too. If he threw twenty superb pitches, he'd be less hittable too. But then his peripherals would be better than they are. It shouldn't affect only singles. It would at least increase his rate of surrendering doubles.
Oleoay
11/11
I guess it depends on whether pitchers do exert some control over BABIP or not. McCracken's original theory suggested they didn't. Yet future studies suggested that some types of pitchers seem to have a slight ability to affect BABIP. If pitchers, independent of defense, have no control over BABIP, then perhaps it would come down to an issue of luck. But what if pitchers affect some portion of BABIP?

Why aren't more extra base hits given up? My best (and probably poor guess) is that baseball is a game of inches, and perhaps Cole Hamels isn't leaving flat pitches directly over the plate, but if he lost a half an inch or an inch on the tail of his movement, then hitters would be more likely to hit it, but not necessarily hit it as squarely. Perhaps something like that is captured by walk rate. Yet, kicking in the back of my mind is that Hamels threw a lot of regular innings and postseason innings last year, and supposedly, wasn't in shape for spring training. If a pitcher who formerly threw at an average of 92 mph now throws 90mph because of fatigue, or gets one inch of movement more or less on his pitches than he used to, that'd make him a tad bit easier to hit. From data posted in an earlier article, it seemed he was getting less horizontal movement...

Or, I'm still offbase.
swartzm
11/11
I addressed movement, location, and velocity in the article, and none seem to be issues at all.

It's silly to try to make up more and more obscure theories. We know that the standard deviation of a binomial random variable is sqrt(p*(1-p)/n). That's a fact. Therefore, there will always be 20% of pitchers that differ from their true BABIP skill by .030 points even if pitchers did control BABIP. It's not hard to believe that those 20% of pitchers would exist.

I've done a lot of work on how much pitchers do control BABIP. Check out this article (and note that Hamels if anything would fit the bill of a pitcher who should have a ever so slightly less than .300 BABIP rather than higher): http://baseballprospectus.com/article.php?articleid=9595
Oleoay
11/11
I'm not trying to make up obscure theories and I really don't like it when I ask what I think are serious questions and the attempts are called silly.

I know you've done a lot of work in this subject and on Cole Hamels in particular, and you're much more experienced at sabremetrics than I am. You are the expert.

So the best I can do is suggest alternate avenues of investigation that might help separate the luck from the skill. I don't do it to try to be silly or annoying, but because I am interested in the topic, I want to learn more and, perhaps, suggest additional avenues of research.

In any event, I apologize for being frustrating.
swartzm
11/11
I'm sorry if I came off as dismissive. There are a lot of bad theories out there about Hamels. I've heard many of them in recent weeks that are not based on facts or are trying to explain the conclusion backwards, by looking for something that changed and deciding it was causal. The way that I should have phrased my comment was that it is important to analyze BABIP frontwards.

This involves asking if certain things are correlated with higher BABIP for the league in general, such as I did in the article I linked in my previous comment. Looking at an outcome and trying to explain why it might be an exception to a rule by finding other idiosyncrasies is always dangerous.

The other very important thing to realize is that even if pitchers did control BABIP despite the lack of persistence, the limits of the sample size mean that you can only really say with confidence that a starting pitchers' BABIP should have been within .040 points of his actual BABIP. That is the amount that luck should count. So, based on this, we can conclude that Hamels BABIP should have been somewhere in between .281 and .361 this year. It is certainly not unreasonable to think it should have been right at .300.

If you want to conclude that his true BABIP skill is different than .300, you need to look for what kinds of players do have BABIPs that are different from .300, not find a player who has an abnormal BABIP and look for characteristics of him. For example, strikeout pitchers have slightly lower BABIP, but we're talking about .292 versus .308 for really good and really poor strikeout pitchers. Ground ball pitchers have slightly higher BABIP, but we're talking about a similar 8 point range. Knuckleballers can have hugely different BABIP, and usually very low BABIP, so if you see a knuckleballer, it's best to look at comparables before making assumptions about his BABIP. I hope that helps explain how to approach this kind of question.
Oleoay
11/11
Thanks for the explanation Matt. When I try to look at characteristics of certain pitchers, I'm not using it so much as an analysis, but identifying possible factors particular to that pitcher that might be good areas of further inquiry. Of course, then I run into a Nolan Ryan problem, where he was famour for throwing a 100 mph fastball.. so is that fastball the reason for his strikeouts? The problem comes with the realization that not not all pitchers that throw 100 mph strike out a lot of batters, especially at Ryan's rate. Still, sometimes looking at the characteristics of players who break the model might give some insight into how the model can be refined. My perception is that, while BABIP studies have improved, there's still a lot that is unknown about BABIP and how much of it is a factor of luck or a factor of skill.

That's why I ask a lot of odd questions, and I know it's got to be a bit repetitious when I ask questions that you've already touched on in other articles. Thanks again for your patience.
swartzm
11/11
No problem, thanks for the clarification.

What I would say is that pitcher BABIP (at least for starting pitchers) would have a natural standard deviation of about .019 if luck were the only factor-- and that much is clear. We know that team defense would cause about .010 of standard deviation. Because of this, we know that the standard deviation of individual pitchers is probably very minimal.

Those two factors would combine to explain about .021 of standard deviation.

For pitchers in 2005-2008 with more than 500 balls in play, the standard deviation was about .021.

That means that there is almost no variance with respect to pitchers. It would certainly be less than .005. In other words, nearly all pitchers should be between .290 and .310 with respect to their natural skill level and probably between .295 and .305.
Oleoay
11/12
The standard deviation is interesting since I wasn't quite sure what the normal amount of variance is.

Maybe it would be instructive to look at pitchers like Hamels who were lucky on BABIP as he was in 2008, then unlucky the next year as he was in 2009, and see how they did in future seasons? Or perhaps a list of the widest swings of BABIP in history and how the rest of those players careers panned out. That can help put Hamels's luck-to-unluck into some kind of context. Also, for those in the media pressing the panic button, comparing Hamels swing of luck versus other pitchers who went on to have good careers might stop some of the psychoanalyzing.
hessshaun
11/10
In thinking about it more and more, and I am admittedly lost in some of the number minutia, wouldn't you think that there is something not jiving here? Sure the numbers correlate and I am not debating your findings at all. But other than saying, good luck/bad luck, can we examine other great pitchers on a year to year basis? What I mean is that we know that CC has been pretty consistent from a year to year, but how much does he deviate in pitch selection and velocity? Is he infusing something else on a yearly basis that keeps him fresh so to speak? I don't want to ask if Cole is too predictable, because that's a bad question. I don't disagree with the luck factor at all, I am just wondering if he was too consistent. Do other pitchers who continually put together consecutive dominant seasons all achieve the same quantifiable differences on a year to year basis? I would assume once he settled in after the elbow woes, that his release point was pretty consistent. The numbers you put out there would seem to indicate to me that isn't an issue.
jdouglass
11/10
Lincecum just had back-to-back dominant seasons.

Fastball: 2008 66.1%, 2009 55.7%
Slider: 2008 1.7%, 2009 5.3%
Curveball: 2008 13.7%, 2008 17.5%
Changeup: 2008 18.5%, 2009 21.4%

He made a remarkable drop in his fastball frequency and spread it across three other pitches. His velocity decreased on every pitch, showing that he's throwing less and pitching more.

hessshaun
11/10
Thanks for that Mr. Douglass. Again, I have no idea where to find these numbers, but I would still be curious how that plays out for other top starters.

Any advice on how to look this stuff up? Admittedly, I don't have a clue.
jdouglass
11/10
hessshaun, you can find pitch values, frequencies and velocities at fangraphs dot com. They pull data from Pitch/fx on mlb.com and publish data points off of it.

You mentioned CC above.

CC, like Hamels, throws three pitches. Except that while Hamels blend is roughly 59% FB/30% CH/10% CB, CC throws a slightly higer % of FB (61.7%), but a more balanced offering of his secondary pitches, the slider and change being thrown 20 and 18% of the time, respectively. And at 94.1 mph on average, CC can better afford to lean on his FB than Hamels, who throws his much slower at 90.2 mph on average.

How does CC stay fresh? By evolving. When he broke into the bigs he threw a good amount of curves and virtually no slider. Now his slider is thrown regularly while his curve has been thrown less each year, to the point that he virtually never threw one in 09.

hessshaun
11/10
Thanks for the information. Going to check it out now.
swartzm
11/11
You are making a mistake with changing percentages of pitches thrown as some sort of holy grail of improved pitching. This is a classic example of people not understanding decision theory and looking only at summary data. Sometimes pitchers develop better pitches. When they do, they will change what they are doing and likely improve. Sometimes pitchers attempt to develop a new pitch but it isn't very good. They don't throw it, and they stay the same. As researchers looking at fangraphs, we do not have the access to know how every pitcher would have done if he changed his repertoire. We only see that those that made the decision to do so were successful. That's the problem. You are assuming that everybody who changes their repertoire will improve, when only those pitchers who will improve by changing their repertoire are the ones that actually do. It's a common mistake when looking at data and surmising a counterfactual can be considered when it cannot.

Also, recall some excellent recent articles by Eric Seidman on perceived velocity. Hamels changeup is more effective by having thrown a fastball on a previous pitch, so the value of throwing a fastball to Hamels won't be recognized in the weighted runs per pitch fangraphs metric even if Hamels had neutral luck.
swartzm
11/11
Lincecum is another example of a pitcher who is better than Hamels. His back to back dominant seasons came with better peripherals than Hamels. That he changed his pitch selection is not good inandof itself.
swartzm
11/11
Sabathia's ERA bounces around a good bit too, as does his BABIP. It's not as noticeable because Sabathia is a better pitcher than Hamels, so his bad years are still above average.
tweicheld
11/10
The previous comment shows stats for each pitch thrown - note how Lincecum throws four while Hamels basically throws two. I'm not comparing the two pitchers, but it CAN'T be a coincidence that the guy with more quality pitches in his repertoire has greater success. I'm willing to bet Hamels would break down at somewhere around 65% fastball, 30% changeup, 5% other.

I don't know what the reason for Hamels' subpar season was, it will be interesting to see how he comes out of the chute next year. He's always been very competitive, but it seemed he threw more hissy fits on the mound this year when things weren't going his way.
swartzm
11/11
Yes, Lincecum is better than Hamels. Of course having four good pitches is better than two, and having two good pitches is better than zero. Every pitcher in the league can be blamed for not being Tim Lincecum and every hitter in the league can be blamed for not being Albert Pujols by this logic.

Also, I find it presumptuous to call Hamels being upset "hissy fits." He is emotional and demonstrative on the mound. If he had a lower pitched voice, this would be viewed as righteous rage. If he screamed with joy when he did things well, it would be viewed as passionate fire in his belly glory. This is all post hoc characterizations.
bsolow
11/12
Matt, I know you're not an econometrician and I'm a little bit late to this discussion, but do you think that the decision of whether or not to add a new pitch to the repertoire could be tested for significant effect using some kind of censored regression (i.e. Heckit) model?

My prior is pretty strong that it's unlikely that a regression like that would bear out much evidence in favor of the value to adding a new pitch independent of the quality of the pitch.
swartzm
11/13
I thought about this one for a while and checked with a few other economists who are better at econometrics. I'm not quite sure if Heckit would work. There is the selection bias of selecting into throwing new pitches only if it works out for the better. But there's also an issue with biased error terms where pitchers like Hamels introduce new pitches thanks to bad luck. Certainly, the average pitcher who feels he needs to introduce a new pitch is more likely to have bad luck than average. I'm not sure if biased error terms matter for Heckit? I think Heckit also requires normality, right? I guess looking at a binomial variable like OOBP rather than ERA could guarantee that? There's a lot of things to consider, and I'm not quite sure yet. It's an interesting thought.
Oleoay
11/13
Perhaps its simplistic, but I often think of baseball similar to poker in that poker and baseball are both games of probabilities. Just as in poker, if you can see your opponents' hand, you can play the hand optimally. If you use deception in poker by varying betting size, playing weaker hands like strong hands or vice versa, you make it harder for the opponent to read your hand, and thus, make it harder for your opponent to play the opponent optimally. If a hitter knows what's coming, they're more likely to be prepared to capitalize on the pitch, and thus, can perform more optimally. An extra pitch of sufficient quality helps keep a hitter guessing. I guess, as an example, the anecdotes of Randy Johnson developing a slider comes best to mind in how it changed his effectiveness.
Oleoay
11/13
Should be "thus, make it harder for your opponent to play optimally."

Ok, enough typoes for me. I'm looking forward to vacation.
bsolow
11/14
This is exactly the right way to think about pitching strategy, and Matt has actually written articles in this vein on BP regarding the game theoretic aspects of pitching: http://www.baseballprospectus.com/article.php?articleid=8932. The question is whether or not Cole Hamels' added pitch would be effective enough to justify using it enough of the time to actually make hitters guess. The fact that he is not adding another pitch suggests that he has likely been unable to develop another pitch effective at the MLB level. Anecdotal evidence abounds about pitchers messing around with new grips, etc., so I don't think it's a question of experimentation for Hamels, but rather whether he is capable of throwing another effective pitch.

Put another way, why doesn't the argument for adding an effective pitch hold for every pitcher? Josh Beckett would be better if he could also throw Mariano Rivera's cutter. Adding the extra pitch is valuable in some aspects -- increasing the number of strategies that a player has to mix between when playing mixed strategy equilibria increases the number of systematic mistakes that player makes -- but has negative value in others, i.e. that the pitch would be extremely hittable when he throws it. The fact that Hamels isn't throwing another pitch suggests that the latter factor outweighs the former for him.
Oleoay
11/14
I agree with your thought process and your logic that an added pitch has to be somewhat effective to be useful. However I am unsure if Hamels has tried (and failed) to add another pitch. He was the hot Phillies prospect for a long time and was elite in 2008 as well as an ace in the playoffs. He's just young enough and perhaps, stubborn enough, to think he doesn't need another pitch. He wouldn't be the first big leaguer that failed to adjust.
swartzm
11/14
You clearly don't watch the Phillies. Hamels struggled with his call up-- 5.98 ERA in his first 11 starts...with about a .320 BABIP but without the walks.

Hamels has been playing with the curve ball for years. It hasn't been good enough to throw as often as the fastball and change up.

We're back to the character attacks I see.
Oleoay
11/14
I've said previously I haven't watched him much, nor am I an avid Phillies fan (or Phillies hater) nor an expert in statistics like you are. I was just responding to Ben's two rationales about Hamels not developing an extra pitch and postulating a third in which young pitchers might not listen to coaching.

How do you consider me saying I am unsure in my idea about Hamels developing another pitch that I am actually making some accusation and/or a character attack, as if I have some kind of vendetta against him? If I really felt that way, do you think I'd be obtuse about it?
willsharp
11/10
Awesome article, Matt.
jdouglass
11/11
Matt, what do you make of Hamels' performance by quality of opponent?

If we split Hamels' 2008 and 2009 seasons up by the OBP of his opposition, creating three splits based on team OBP and calling them "good" (team OBP .338+), "average" (team OBP .324-.337) and "bad" (team OBP under .324) opposition, we get this:

Percent of total IP:
2008: Good 38.4% Average 30.5% Bad 31.1%
2009: Good 34.1% Average 31.1% Bad 34.8%

K/9
2008 K/9: Good 7.63 Average 9.09 Bad 6.62
2009 K/9: Good 7.03 Average 9.50 Bad 7.29

UBB/9
2008 UBB/9: Good 1.96 Average 1.30 Bad 2.17
2009 UBB/9: Good 2.34 Average 1.51 Bad 1.62

Year-over-year, he had more than a 10% change in the % of IP he worked against both "good" and "bad" hitting teams, while staying somewhat flat vs. "average" competition, leading to overall easier competition this year than last.

In both K and UBB metrics Hamels improved against bad opposition while getting worse against good-hitting opponents. His smallest changes--an improvement in K and a worsening in UBB--were against the middle of the pack, against whom he is at his best.

What's really bizarro about Hamels' 08 season is that he was better vs. good opponents, as far as K and UBB rates, than he was against bad ones. To me, that's an indicator that he was doing something pretty special in 2008 that he just didn't have in 2009.

I don't think Hamels' flatlined K/9 and UBB/9 year-over-year are not necessarily because Hamels' was the same, or as good, but rather that those metrics in 2009 were boosted by worsening opposition faced by a pitcher who declined a bit.


swartzm
11/11
When you subset his innings into little 60-87 subsets, changes of 0.6 K/9 or BB/9 represent about 5 strikeouts and walks different. Just 5. Claiming he struck out five more hitters on bad teams and five fewer hitters on good teams simply does not prove a thing when he struck out about the same number of hitters. That is ridiculously small sample sizes that you're looking at. It's the type of thing that no one would ever look for unless they were trying to make a post hoc rationalization.

It's also clumping together data by team, when obviously many of those players on good teams are bad and many of the players on bad teams are good. It's a noisy measure where you are taking one or two really good games that happened to come against a good opponent in 2008 and another that happened to come against a bad opponent in 2009 and drawing conclusions. He happened to have a particularly good game against the Cardinals and another against the Cubs in 2008. He happened to dominate the Giants and Orioles once in 2009.

Any pitcher who has a 2.34 BB/9 and 7.03 K/9 against good teams does worry people. No one has "magic" against good teams. Anybody could have looked at the fact that he performed better against good teams than bad teams in 2008 and said "that's not a long term trend."

Again, why if he got so much worse did it not affect his overall strikeout and walk totals, the rate of hitters hitting the ball for line drives, the rate of hitters hitting the ball to the outfield, or the rate of hitters hitting the ball over the fence? A dip in quality of opponent of such a ridiculously small magnitude would not explain the 35 singles that came with no change in batted ball distribution that you need to explain.
blcartwright
11/12
While pointing out that Hamels may have been 'unlucky' in 2009, I see no mention of how he may have been quite 'lucky' in the previous season, setting up a perceived decline of talent because of the loss of .057 on his babip.

Adhusted for park, but not for defense, Hamels allowed babips of .295 in 2006 and .282 in 2007, for a two year average of .290. In 2008, his babip allowed dropped even further to .261. We might detect a trend (two consecutive years of decrease) but we know that this is a stat that is not as well controlled by the pitcher himself. In 2009 Hamels allowed a .318 babip, severely bucking the perceived trend - but the two year average of 2008-2009 is .290, exactly the same as the two year average of 2006-2007. Larger sample sizes wash out any trends. Using a three year weighted average, Hamels babip projection was .294, .289, .281, .293 the last four seaons, while his wOBA allowed projections were .309, .303, .296, .303 - very consistent.

Breaking down Hamels balls in play

Year GB% LD% FB% PU% IFH% GBH% LDH% FBH% HR% BABIP
2006 .425 .188 .318 .075 .089 .163 .730 .151 .109 .298
2007 .444 .160 .300 .096 .068 .169 .713 .187 .105 .282
2008 .412 .190 .303 .095 .062 .117 .672 .181 .086 .262
2009 .439 .194 .266 .102 .113 .156 .726 .225 .087 .322
MLB .461 .194 .264 .080 .078 .175 .728 .174 .078 .299
2009 compared to previous 3 +9 +3 +4 +7 -3
2009 compared to 2008 +11 +10 +6 +6 0


Hamels allowed very few ground ball hts to the outfield in 2008 (.117 to league avg=.175). His 2009 rate was below league average and below 2006 & 2007.

In 2008 Hamels was also below average in infield hit rate (.062 to league avg of .078), but his 2008 rate jumped to .113, hardly something he could control.


swartzm
11/12
I agree with all of this, but I did say that Hamels was extraordinarily lucky in 2008 three times in this article, and several more times in the original article linked at the top. I characterized him as a 3.65 ERA type pitcher who had about .60 swings of good luck and bad luck due to BABIP fluctuation.
Oleoay
11/12
Thanks for the breakdown Brian. Is there anything informative on where the ground ball hits to the outfield were hit or where infield hits were given up? Perhaps there was a fielder like Rollins or Feliz that had a bad year or the switch from Burrell to Ibanez was drastic enough to affect BABIP if, for example, right handed hitters hit a lot of grounders to the left side of the diamond that used to be turned into outs, but in 2009, became hits?
blcartwright
11/12
I'm sorry Matt, upon re-reading the first paragraph you certianly do call his 2008 lucky.

Richard, I don't have the positions at this time. but soon I improve my code whch determines that.
Oleoay
11/14
Thanks Brian I look forward to seeing it. BTW what kind of code do you use? SQL?