I would like to take a moment to introduce myself. My name is Peter Bendix, and I’m new around here. I’ll be contributing weekly articles (if not more often) about baseball. Many of my articles will be statistically/sabermetrically themed, but many also will not. Hopefully, you will be able to enjoy my articles whether you consider yourself a hardcore sabermetrician, or if you think a sabermetrician is some kind of ancient dentist in the African safari.
I welcome any and all feedback, suggests, and constructive criticism. Thanks for reading, and I hope you enjoy!
I should first tell you that I’m a Cleveland Indians fan. More accurately, I’m a *baseball* fan who’s originally still from
Fausto Carmona was a key ingredient in the Cleveland Indians’s success in 2007. Carmona seemingly came out of nowhere to post the second-best ERA in the American League (John Lackey passed Carmona in his last start of the season with seven shutout innings, leaving Lackey with a 3.01 ERA and Carmona 3.06). Carmona was atrocious for the Tribe in his first season in 2006, posting a 5.42 ERA and gaining unwanted notoriety for a week-long stretch at the end of July where he blew three saves and was charged with four losses.
Savvy observers might have realized after 2006 that Carmona was far better than his 5.42 ERA indicated – Carmona was very unlucky on balls in play, giving up a .336 batting average on balls in play (BABIP). Additionally, although Carmona was and remains a groundball pitcher, he gave up an inordinate amount of homers, as 15.1% of his fly balls became homers. These factors likely mostly had to do with luck. In fact, his “expected fielding independent” (xFIP) – which normalizes these lucky components of ERA – was 4.46. However, no one could have predicted the degree to which Carmona broke out in 2007.
Or could they?
In 2006, Carmona had severe splits depending on whether the first pitch of the at-bat was a ball or strike. ESPN keeps track of the result after the first pitch, and Carmona’s splits are amazing: if the first pitch of the at-bat was a strike, opposing batters hit .231/.293/.299; however, if the first pitch was a ball, opposing hitters crushed him to a tune of .383/.474/.625. Basically, Carmona got batters out 18% more often if he delivered a first-pitch strike – additionally, if a batter managed to get a hit after being down 0-1, it was far more likely to be a single. Furthermore, hitters’ BABIP was very different depending on the first pitch: if the first pitch was a strike, hitters’ BABIP was .320; but if the first pitch was a ball, hitters’ BABIP was an incredible .532.
In 2006, Carmona threw first-pitch strikes to only 54% of the hitters he faced. Nearly one in two hitters were hitting .383/.474/.625 against Carmona due to his inability to throw first-pitch strikes. However, he threw 60% of his pitches for strikes overall. This suggests that his low first-pitch strike number is somewhat of an aberration – there is no reason a pitcher’s ability to throw a strike should differ depending on the count (perhaps if the count is 3-0 or 0-2, but I’d posit that there should be no difference between overall strike % and first-pitch strike %).
Now let’s look at 2007. Carmona’s splits weren’t quite as drastic, but they were still quite large: if the first pitch was a strike in 2007, opposing batters hit .220/.248/.307 (remarkably similar to their line after a first pitch strike in 2006). Once again, however, if the first pitch was a ball, batters did well, hitting .285/.386/.420. That’s nearly a 14% different chance of making an out depending on whether the first pitch was a ball or a strike. Once again, we also notice a difference in BABIP depending on the first pitch: if the first pitch was a strike, hitters’ BABIP was .270, whereas if the first pitch was a ball, hitters’ BABIP was .314. While this is not as drastic a difference as in 2006, it is still a rather large difference for a stat that is supposed to be random.
In 2007, Carmona managed to throw first pitch strikes to 58% of the batters he faced. While a 4% increase may not seem like that much, over the course of a full season (of, say, 862 batters, the amount Carmona faced in 2007), that’s 34 more hitters who’s chances of getting on base are 14% less than they otherwise would be. In other words, in 2007, basically one batter during each one of Carmona’s 32 starts was replaced with a far worse hitter. That may not sound like a lot, but it certainly is.
From a statistical standpoint, we can attribute much of Carmona’s turnaround from 2006 to 2007 to regression to the mean. Carmona’s .336 BABIP in 2006 regressed to .280 in 2007. He gave up fewer homers per fly ball (15.1% in 2006, down to 11.8% in 2007). His splits between whether the first pitch was a ball or strike continued (although he did improve against batters to whom he delivered a first pitch ball), and his overall stats improved due to his ability to throw more first pitch strikes.
Carmona was lucky to have a 3.06 ERA in 2007 – his xFIP was 3.99. However, when we examine the stats more deeply, we can see that Carmona’s true ability only changed marginally from 2006 to 2007 – basically, he threw some more strikes, which is almost certainly an improvement in skill, but only a rather small one. However, his stats changed drastically. Thus, while we could not have predicted a 3.06 ERA from Carmona, had we done this exercise before the 2007 season we would have likely predicted an ERA around 4.50, and we would have realized that if this young pitcher had improved only slightly in his ability to throw first pitch strikes (something that it is not unlikely that a young pitcher can improve upon), Carmona would have been a candidate for an ERA around 4. We might not have been right, but our prediction would have been a lot closer than most peoples’, and Carmona’s explosion onto the scene wouldn’t have surprised us nearly as much.