keyboard_arrow_uptop

At this point in the season, most baseball fans are aware that the Phillies‘ Game Five starter this evening, Cole Hamels, has had far more trouble preventing runs in 2009 than he did in 2008. In 2008, Hamels seemed unhittable for much of the season and the post-season, and the Dodgers knew going into Game Five last year that they had their work cut out for them as they faced elimination, down three games to one. After knocking him around in the fifth inning of last Thursday’s series opener, the Dodgers are confident that they are up against a different pitcher than the one that stymied them for the clincher last October, as they face elimination yet again.

Many reasons have been presented for the decline of King Cole. All through last season, prognosticators warned about “the Verducci Effect” and its impact on young Hamels. The Verducci Effect states that pitchers under 25 years old who threw thirty or more innings in the past year than they threw in the year before that are particularly vulnerable to injuries in this year. From the time Hamels hurt his elbow in spring training, many have claimed that he has been playing hurt. However, despite being hit by a line drive and twisting his ankle falling off the mound in early starts, Hamels managed 31 starts during the regular season and threw 193 2/3 innings. His velocity was the same as last year as well, averaging 90.4 mph on his fastball in 2008 and 90.2 in 2009. However, it seems clear that something is going wrong for Hamels on the mound. His ERA climbed from 3.09 last season to 4.32 this year, and his WHIP rose from a league-leading 1.08 in 2008 to a mediocre 1.29 in 2009.

Hamels has maintained all season that this has been the toughest one in his career, and that he has learned the most from it. His detractors claim he is distracted by his celebrity, frequently starring in commercials and advertisements throughout the 2009 season.

However, I will present an alternative view in this article: Cole Hamels has pitched just as well last year as he has this year.

I know that his ERA jump appears ugly, but bear with me for a minute. At this point, most sabermetricians have learned that a pitcher’s peripheral statistics (strikeout rate, walk rate, and ground-ball rate) are far more persistent and indicative of his true skill level than his BABIP and his home runs per outfield fly ball. Both BABIP and HR/OFFB are far more vulnerable bad luck than K/PA, BB/PA and GB/BIP. Cole Hamels appears to be a victim of this as we look at these sets of numbers for 2008 and 2009.


Year  K/PA  UBB/PA  GB%   FB%   LD%  PU%  BABIP  HR/OFFB
2008  .214  .050   41.1  30.3  19.0  9.6  .262   .139
2009  .206  .048   43.6  26.3  19.6 10.5  .321   .153

Looking at this table, we see that there are no remotely significant differences between his strikeout rate, his walk rate, or his batted-ball rates. What has changed is simply his BABIP and his HR/OFFB, two statistics that frequently relate to bad luck. From this chart, it seems pretty clear that Hamels was extraordinarily lucky when he put up his .262 BABIP in 2008, and extraordinarily unlucky when he put up his .321 BABIP in 2009.

Consider several luck-neutral measures of ERA that sabermetricians have developed over the years:


Year    QERA   FIP  xFIP
2008    3.67  3.72  3.75   
2009    3.63  3.72  3.69

Even though pitchers do not exhibit any real persistence with respect to their seasonal line-drive rates, Hamels did not even show any real change in his line-drive rate at all. Instead, he gave up more hits on the same batted balls compared with last year. Consider Hamels’ splits by batted-ball type for 2008 and 2009:


        Ground Balls  OF Fly Balls   Line Drives
Year     OBP/ SLG      OBP/ SLG       OBP/ SLG
2008    .182/.194     .280/.753      .683/ .984
2009    .265/.290     .340/.908      .726/1.034

So, Cole surrendered more hits on each type of batted ball in 2009 than he did in 2008. His rate of infield hits was roughly the same (5.4 percent in 2008 vs. 6.1 percent in 2009), but more ground balls reached the outfield in 2009: 20.4 percent vs. just 12.8 percent. League average is about 18 percent, and the Phillies’ infield defense is above average, so it seems like he was just very lucky in 2008 and very unlucky in 2009. Hamels actually got more hitters to pop up in 2009 than in 2008 (though only by a small margin), but the fly balls that did go to the outfield landed more frequently. He allowed 14.1 percent of outfield fly balls that were not home runs to go for extra-base hits in 2009, up from 11.4 percent in 2008 (league average is 11.2 percent). He also allowed more fly-ball singles-presumably these were mostly bloopers that landed in no-man’s land-up from 6.8 to 9.6 percent of his non-homer outfield fly balls (league average was 7.5 percent in 2008 and 6.9 percent in 2009). He also saw more line drives sneak between or in front of players, erasing his lucky .683 line-drive BABIP in 2008 with a more typical .726 BABIP in 2009; league average was .718 in 2008 and .724 in 2009. Hamels also allowed more extra-base hits on line drives in play as well, jumping from 19.7 percent to 21.9 percent in the past season; here again, league average was 19.5 percent in 2008, and 19.0 percent in 2009.

With all these extra hits on balls in play, it seems that just because Hamels does not appear to be exhibiting any different trends in his more persistent statistics, he is not necessarily pitching the same. Many have claimed that Hamels has left more pitches over the plate, in hittable locations. The lack of a notable increase in line-drive rate makes that theory seem doubtful, but let’s look at a little more data.

If Hamels was throwing the ball in more hittable positions, we would expect that he might be allowing more balls to be hit to the outfield. This is not true:


       Percentage
Year    Hit to OF
2008     51.6%
2009     51.3%

One might expect that Hamels was not inducing poor contact if he was allowing more hittable pitches. Are hitters fouling off more pitches? No, they are not-they have gone from 4.1 percent to 3.7 percent, roughly the same amount of fouls. Is he allowing hitters to get around more often? The answer certainly appears to be no, as he has allowed similar rates of balls pulled, hit to center, and hit the other way:


Year   Pull  Center Opposite
2008  28.7%   54.3%   17.0%
2009  29.9%   54.0%   16.1%

Those rates are all very close to league average, although closer to 26 percent of pitches are pulled, and 57 percent of pitches are hit up the middle.

OK, so is he getting more swings-and-misses? Throwing more pitches out of the strike zone? Consider his pitch-by-pitch results from StatCorner.com:


Year  Ball% ClStr%  SwStr%  Foul% InPlay%  IntBall%
2008  32.6   16.6    11.2   19.7   19.3     0.6
2009  32.7   17.5    11.8   18.3   19.2     0.4

Ball%: Percentage of pitches thrown that were balls.
ClStr%: Percentage of pitches thrown that were called strikes.
SwStr%: Percentage of pitches thrown that were swinging strikes.
Foul%: Percentage of pitches thrown that were fouled off.
InPlay%: Percentage of pitches thrown that were put into play by the hitter.
IntBall%: Percentage of pitches thrown that were intentional balls.

It certainly seems like Hamels is missing more bats without missing the strike zone any more. FanGraphs keeps data on pitch-by-pitch results as well, so let’s take a look at Hamels’ data from 2008 and 2009:


Year  OSwing ZSwing Swing  OContact  ZContact  Contact
2008   30.8   67.7   50.4   60.8      83.4      76.9
2009   26.8   69.8   49.4   63.2      79.3      52.5

O-Swing%: Percentage of pitches swung at out of the strike zone.
Z-Swing%: Percentage of pitches swung at in the strike zone.
Swing%: Percentage of pitches swung at overall.
O-Contact%: Percentage of swings that made contact among pitches out of the zone.
Z-Contact%: Percentage of swings that made contact among pitches in the strike zone.
Contact%: Percentage of pitches swung at overall.

Although there has been a slight tendency for hitters to better recognize pitches in the strike zone in 2009, they have made contact on fewer of their swings at pitches in the zone, and more on their swings at pitches out of the zone, which would typically lead to a lower BABIP as, on average, hitters do not do well hitting balls out of the strike zone.

It appears that looking at his pitch-by-pitch results, and his plate appearance by plate appearance results, the only differences that are showing up are hit rates on each type of batted ball. I explained a couple weeks ago that pitcher BABIP varied significantly by count, so perhaps Hamels is letting hitters make contact on less favorable counts. This does not appear to be true. I showed in that article that hitters tend to have particularly good BABIP on three-ball counts and particularly poor BABIP on two-strike counts. Hamels allowed 16.0 percent of his contact to be on three-ball counts in 2008, and allowed 15.4 percent of his contact to be on three-ball counts in 2009. He allowed 51.6 percent of his contact on two-strike counts in 2008, and 51.4 percent of his contact on two-strike counts in 2009. He has not been particularly different in detail with the first pitch either, getting behind 1-0 a total of 38.3 percent of the time in 2008 against 39.2 percent in 2009, getting ahead 0-1 48.6 percent of the time in 2008 and 49.9 percent of the time in 2009, and allowing the first ball to be hit into play 13.0 percent of the time in 2008 and 10.9 percent in 2009. The drop in first pitches put into play is not statistically significant (t=1.34), but it is notable. However, his BABIP was .248 on those pitches in 2008, and .309 in 2009, so fewer first pitches put into play does not seem to be a culprit of extra BABIP.

Instead, Hamels BABIP has been different on each type of batted ball:


Year  0-0  1-0  2-0  3-0  0-1  1-1  2-1  3-1  0-2  1-2  2-2  3-2
2008 .248 .208 .267 .000 .325 .246 .326 .263 .264 .195 .355 .216
2009 .309 .341 .250 .000 .321 .325 .273 .286 .278 .284 .313 .524

Shockingly, Hamels’ BABIP on full counts has shot up. He has also allowed four home runs on full counts in 2009, compared to just two in 2008. However, he appears to be walking and striking out people at the same rate: 26 walks and 32 Ks in 2008, and 24 walks against 28 Ks in 2009, so it appears unlikely that he is grooving pitches on full counts or anything like that. Hamels also appears to be struggling on two-strike counts, allowing a BABIP of .338 on them in 2009, compared to .255 in 2008. However, it would be tough to argue that he is more hittable in these case as he struck 40.2 percent of hitters who he got two strikes on, similar to his 2008 rate of 41.5 percent.

It really does appear that Hamels may have simply seen a great deal of good fortune in 2008, and a great deal of bad fortune in 2009. I have argued before that as pitchers and hitters are constantly trying to out-think one another, they probably end up canceling each other out enough that hitters guess right a pretty stable amount of the time in the long run, and that this perhaps may be why BABIP is not persistent for pitchers. If a pitcher tried to repeatedly throw the same pitch on the same counts, he would eventually be caught up with, and would need to make adjustments accordingly. All thirty teams have pitching coaches who are responsible for making sure that their pitchers do not do that, so it is unlikely that any pitcher could be consistently predictable or consistently tricky in consecutive years. Thus, BABIP would appear to have little correlation in the data. Hamels may be more predictable in certain counts than in previous years, but it has not led to an increase in line drive rate or an increase in the rate of balls hit to the outfield. In fact, his home-run rate is lower on a per-batter basis: 3.1 percent in 2008, and 2.9 percent in 2009.

Taken all together, Hamels simply is the same pitcher he was last year. He throws the same pitches at the same speeds, and induces the same amount of contact and same magnitude of contact. He throws as many balls and as many strikes. He is neither the type of pitcher who will consistently put up ERAs as low as 3.09 or as high as 4.32; his skill level is more likely around 3.65. However, deviations like this in BABIP are far from normal. For a pitcher who allows about 600 balls in play in a season, the standard deviation of his BABIP will be about .019 points from his true value-meaning that one-sixth of the time he will be .019 points or more below his true BABIP in a season, and one-sixth of the time he will be about .019 points above his true BABIP in a season. When that happens, replacing about 35 hits with outs or vice versa, a pitcher’s ERA will typically swing by a full run in either direction. Seeing as Hamels has a high strikeout rate (which correlates with low BABIP), a slightly below-average ground-ball rate (which also correlates with low BABIP), and plays in front of a strong defense, his true BABIP this year should probably have been around .290. If he has normal luck on other rates next year, and exhibits the same skill level, he will likely end up with that ERA around 3.65, even allowing for variations as seemingly significant as those between 2008 and 2009.

You need to be logged in to comment. Login or Subscribe
buffum
10/21
Fantastic. Took as rather bland statement that sounds a little like hand-waving ("His peripherals are the same, he's just less lucky") and really looked at it from lots of angles and supported the claim in a meaningful and persuasive way. Thought question: if the Dodgers THINK Hamels is more hittable, is it possible that they will be more relaxed and confident at the plate, producing better RESULTS? I don't actually know how to investigate that, but it makes a fair amount of intuitive sense and would seem to agree with what small personal experience I've got. But then, I've felt "in the zone" before, too ...
Richie
10/21
Agreed. Fantastic look at the issue. Thank you.
SamHughes
10/21
Terrific piece. Not sure if this has been addressed elsewhere, but is he throwing roughly the same mix of fastballs, changeups & curves this year? More (mediocre) curves *might* explain some of that higher BABIP, though probably not. So would games when he's having trouble with his changeup. But I guess that would probably show up in the line-drive rate or other numbers....
j11forbes
10/21
If I'm remembering correctly, Hamels has actually thrown significantly less curveballs this season. I believe someone used this to argue that he has become more predictable this season. However, I can't remember the source for this so take this information with a massive grain of salt.
swartzm
10/21
Maybe a little, but I doubt enough to explain anything-- he went from 13.7% curveballs to 10.5% curveballs (and from 54.8 to 59.1% fastballs, and 31.5% to 30.3% changeups). I doubt that would be enough to help hitters predict him more. The curveball has always been pretty much a show-me pitch for Hamels.
DLaRoss
10/21
The problem Could it be that he was always predictable, but batters didn't have the chance to study and adapt until after that season?
SoxOsPhils
10/21
Awesome, this article is a great example of why we all subscribe to this site! Thank you for all the thought that went into this.
3n2sports
10/21
I can't wait for hit f/x to finally offer us the full story for these kinds of analyses. The only question that remains to me is: Are batters hitting him harder when they do hit him? All we need are some bat exit speeds on the different buckets of contact and we'd know for sure whether this is all a matter of luck or not.
swartzm
10/21
I wonder if we would really learn much from that information though-- we would need to look at a lot of hit f/x to confirm whether pitchers or just hitters are actually responsible for speed off the bat. My suspicion is that pitchers primarily control whether the ball hits or misses the bat, and what direction and trajectory it has if the batter does hit it. I would guess that if line drive rate lacks persistence, so does speed off the bat. Then again, the argument you're making would need to extend all the way to say that the contact he surrendered to the outfield was hit harder than the contact he surrendered to the outfield last year-- but that no more balls were hit hard enough to make it to the outfield. I kind of doubt that, but it would interesting to see.
tide182
10/21
Great stuff - really great stuff. I had been traveling and hadn't gotten the chance to read any of the stuff here at BP for a while, so to come back with a piece like this was just awesome. Loved that you continued to persist with questions beyond the first layer of data analysis in order to reach your conclusion - completely agree with SoxOsPhils. Keep it up!
agentsteel53
10/21
King Cole's '08 wasn't bad at all. He declined some in '09. hmm wait a minute!
louisma
10/21
I can see your reason for calling it a decline (increased H/9 and BB/9), but his K/9 ratio also increased, and he allowed the same number of runs in more IP, so I would say he was still just as good, on a true talent basis, in '09.
jkelley3
10/21
Nice analysis but that ERA differential is huge and still leaves me scratching my head. Could it be that he just does get out of trouble well? He hasn't looked comfortable pitching out of the stretch this year. Is there a "pitcher's RISP"? If so I wonder what this differential would be for Hamels from '08 to '09.
jkelley3
10/21
Correction: meant to write, "Could it be that he just doesn't get out of trouble well (this year)?"
swartzm
10/21
I checked that, but it actually went the opposite way. You can see his splits on baseball-reference.com, but his K/BB was actually better in 2009 with runners on, and his OPS w/ and w/o RISP difference was comparable.
aaronbailey52
10/21
2009 Contact % = 72.5% ? Or did hitters really swing at 24% fewer pitches?
swartzm
10/21
Oops! That is supposed to say 75.2%. Bad typo. Sorry.
ofMontreal
10/21
Wonderful article. Reminds me of the old saw about one extra hit a week is the difference between a great hitter and an average one.
patjmoore1
10/21
Echoing everyone else, great analysis and use of numerous approaches to the issue. I have been scratching my head about Hamels all year and am doing so a little less now.
brianjamesoak
10/21
Will wondered if anyone could explain Hamels' year outside of the Verducci Effect about two months ago in his column and I responded with a shorter response that was nevertheless right in line with what is here. http://baseballprospectus.com/article.php?articleid=9360 Yeah, the thing about Hamels was that he was never as good as he appeared so a universal-type karmic correction was in order for him, one might say.
bldxyz123
10/21
We just call that regression to the mean... In this case, Cole Hamel's true mean.
brianjamesoak
10/21
Nah, he had a worse year than he should have this year so it's the proverbial 'correction' rather than regression to the mean.
bldxyz123
10/22
You are right. Technically, since his 'true mean' is between 2008 and 2009, he more than regressed.
bldxyz123
10/21
Amazingly thorough and persuasive work, Matt. Nice job. I wish it were a simple matter to look at such things for every pitcher. For example, I know there was lots of analysis to suggest that last year's Brad Lidge was an aberration, but is his awful year this year just a dramatically different roll of the dice? Similarly, how differently did Chad Billingsley perform between the first and second half of the season beyond the conventional "overused" and "its in his head" explanations?
brianjamesoak
10/21
With Lidge it was more of a bad year all around. He was still striking out batters but not at the rate he did the year before, and striking out well over a guy an inning can make up for an ugly walk rate. This year the walk rate wasn't just ugly, it was horrendous, and without the gaudy strikeout rate to help make up for it. Add to that, a sky-high homer rate and his situation wasn't good. You walk a guy in most innings and allow homers afterward a quarter of those times and he really earned his earned run average. Whether he can bounce back is another question. All his problems stemmed from poor location, which was rumored to be due to imperfect health. The K rate shows that he still has talent so an off-season of rest might do him some good.
yadenr
10/21
Great piece. Nice writing, thorough analysis...more please. One thought- did you look into strand rate? Another statistic that can indicate good/bad luck (or sometimes skill) over a year.
NathanJM
10/22
How about the effects of the defensive changes? I can't remember off the top of my head, but if Hamels is pitching roughly the same, some could be luck and expected to eventual cancel out, but some could be reflecting changes in the defense's ability to get to balls, which might indicate a portion of "decline" (not his) that we can expect to continue in 2010...
DLaRoss
10/23
Here's another theory: other teams have found tells for Hamels' pitches. http://www.philly.com/dailynews/sports/20091022_Paul_Hagen__Tipping_or_not__Phillies_starter_Hamels_still_struggles.html
blanck28
10/24
One of the better articles in recent history, and one of the reasons I'm very happy BP kept Matt in the bullpen even after BP Idol. No offense, Ken, but I thought Matt had a higher VORW for the Idol season.
kimberg
11/02
I'm surprised at all the positive comments on this article. No attempt at statistical analysis whatsoever, just a long series of "seems to me" eyeball analyses of one descriptive stat after another. Probably there's a good point here, but this is no way to analyze numerical data.