RSS Twitter Facebook YouTube
Expand Menu

Carrasco's Conundrum: Summing the parts

Carrasco's Conundrum: Summing the parts
Carlos Carrasco is 11-19 with a 5.29 ERA in 48 career big league appearances. (Photo: AP)
March 27, 2014
Share via: Share: Facebook Share: Twitter Share: Google Share: Pinterest Share: Print Share: Email

In 2012, Max Scherzer, though generally regarded as a good pitcher, was no one’s idea of a world-class run preventer. His 3.74 ERA was certainly above-average, but based on this result statistic, ERA, alone, no one could have possibly predicted he was poised for a Cy Young-caliber year in 2013.

Yet if one were to look beyond these result statistics to component statistics – Scherzer’s incredible strikeout rate and better-than-average walk rate – one might well have drawn the conclusion that Scherzer was poised for a breakout. When Scherzer did win the 2013 Cy Young award, it was these component statistics that provided a predictive basis that he would be something more than a merely good pitcher.

It’s an accepted tenet of saber orthodoxy (and, incidentally, completely true) that certain component statistics – strikeout rate, to the greatest extent – are effective predictors of result statistics. Component statistics like strikeout rate succeed as predictors for two reasons – there is a causative relationship between component and result, and they stabilize much more quickly than result statistics.

The relationship between strikeouts/walks and ERA, in this case, is obvious and intuitive: the more strikeouts, the fewer balls in play and possible hits; moreover, it correlates very strongly to ERA – correlation exists. Predictive components must also stabilize much more quickly: ERA fluctuates inconsistently, based on random factors like BABIP that have more to do with the defense than the pitcher himself. For these reasons, Strikeout Rate, a component statistic, is regarded as an effective predictive component stat, predicting the next year’s ERA moreeffectively than same-year ERA itself.

The case against Carlos Carrasco from statistics isn’t particularly difficult to make; Carlos Carrasco’s FanGraphs page is much less a repository of stats than a neon sign blaring ‘Carlos Carrasco is bad at throwing baseballs.’ Every statistic – from ERA to Strikeout rate to Walk rate – suggest that Carlos Carrasco is an extraordinarily mediocre pitcher. Yet second-degree component statistics (namely, swinging strike rate) suggest that Carrasco is a pitcher whose first-degree component statistics (strikeouts) should improve.

Second-Degree Statistics: Stats behind stats behind stats

Strikeout rates have hitherto been treated as cause statistics; however, it’s not true that strikeout rates are themselves root causes – strikeouts, obviously, are any combination of three or more foul balls, called strikes, or swinging strikes, the last of which must not be a foul. In this sense, if one considers runs allowed a zero-degree statistic, a true result, and if one considers strikeout rate a first-degree statistic, a cause of zero-degree statistics, then one might in turn consider strikeout rate's components second-degree components.

Each of these three are component statistics of strikeouts – each of them have a clear causative impact – but not all of them are predictive components, since not all of them, per a FanGraphs article on expexcted strikeout rates, correlate strongly with strikeout rate.

Called Strikes Swinging Strikes Foul Strikes
0.01 0.81 0.20

Source: FanGraphs, Mike Podhorzer

Among the three strike types (excluding balls put in play, which - shockingly - correlate negatively with strikeout rate), only Swinging Strike rate has anything approaching a strong correlation with strikeout rate; foul balls rates, while they are entirely interesting, are not good predictors of strikeout rate, and counter-intuitively, called strikes as a percentage of total strikes correlates not at all with strikeout rate. In this case, the only second-degree component statistic with useful predictive ability is swinging strike rate.

Carrasco - The Second-Degree Disconnect

And for most pitchers, these result statistics bear out the component statistics; a pitcher like Jeremy Guthrie, for instance, has extremely low strikeout rates, but one finds that this abysmal strikeout rate is a direct result of an equally abysmal swinging strike rate. For the Jeremy Guthries of the world, those pitchers who have compiled long MLB resumés, the results align with the components – the poor components predict poor results, and he ends up carving a perfectly lucrative career out of bunches upon myriad bunches of below-average innings.

In the case of Carrasco, he's thrown 238.3 major league innings. Most of those innings, however, were before he had Tommy John Surgery; given that the effectiveness of pitchers changes in the aftermath of this procedure, this inquiry shall focus solely on his 2013 effectiveness.

In 2013, Carrasco's 9.0% swinging strike rate was marginally below the league average of 9.3%; however, given that relievers typically have a higher-than-average swinging strike rate and starting pitchers typically a lower-than-average swinging strike rate, Carrasco's predictive component was very much in line with league average. Yet this component did not translate to results, as he struck out 13.8% of batters he faced, relative to the league average strikeout rate of 19.9%.

Overthinking It: The Criticism

Given this dissonance between predictive component statistic and result statistic, one can draw one of two conclusions: that Swinging Strike rate is not a universally reliable statistic, or Carrasco should expect his strikeout rate to increase in line with his strikeout rate.

Given the aforementioned strong correlation between swinging strike rate and strikeout rate, and given the extremely obvious causative relationship between the two, it's extremely difficult to argue that SwStr% is not an effective predictive component. The argument against Carrasco, as noted, is extremely persuasive at first glance - he's thrown 238 major league innings, and has seen his strikeout rate dramatically under-perform his swinging strike rate for the entirety of his career. Given this gap, and given the memetic insistence that Carrasco is a madman on the mound, the near-consensus has arisen that Carrasco has mental problems, and that these mental problems are responsible for this gap.

Yet it's not clear what bearing this would have on his pitching, if true. If it meant Carrasco threw fewer strikes, then there'd be reason to believe it; however, Carrasco's 66.5% first-pitch strike rate was well above the major league average. Combine this with his entirely solid swinging strike rate, and there's nothing that indicates that Carrasco should have trouble; if he did have issues and these issues did affect his pitching, one would think that it would show itself somewhere in Carrasco's stat line. Yet in nearly any pitching metric one would like to use, Carrasco comes out no substantially worse than average.

It seems likely, then, that this particular narrative is an invention of those who would wish to explain the dissonance between Carrasco's second-degree components and his first-degree results. It's not to say that this is false, inherently, but it is much more likely that this is akin to attempting to decode a signal from the white noise of one's television - it's possible that there's a channel trying to come through, certainly, but it's equally likely that the human brain, fantastic as it is at perceiving patterns, convinces itself to hear a pattern in the noise. Sometimes, it's just noise.

The argument for Carrasco is hardly a popular one, but the assertion that Carrasco will go into 2014 as an approximately average, or slightly below-average, pitcher is one that is reasonably well-supported by second-order statistics. His 2013 walk rate was higher than league average, but there's reason to believe that his second-degree components, swinging strike chief among them, should buoy him to a league-average strikeout rate - perhaps slightly above, perhaps slightly below.

The logic that suggested Scherzer would break out into the dichromatic monster of a pitcher that was to Cleveland the direful spring of woes unnumber'd - that is, the logic that suggested certain components predict results - is the same logic that portends a solid 2014 from Carrasco. With elite framing catcher Yan Gomes behind the dish (a substantial defensive step-up from Carlos Santana), as well as a secure rotation spot to begin the season, the ideal conditions are established for Carrasco to develop into an average major league starting pitcher - which, while not the most enthralling of endorsements, is still an extremely useful player in the #5 position.

As with everything Carlos Carrasco does, this requires him to do with his components what he's never done before: put the pieces together.

Stats courtesy of FanGraphs, Baseball-Reference

John can be reached on Twitter at @JHGrimmHe can also be reached by e-mail at

User Comments

March 28, 2014 - 2:17 AM EDT
Best article on this site ever.
March 27, 2014 - 5:35 PM EDT
"With elite framing catcher Yan Gomes behind the dish (a substantial defensive step-up from Carlos Santana),"

This is something that really excites me with all the starters, not just Carrasco.
March 27, 2014 - 5:23 PM EDT
"Don't think Meat ,you can only hurt the team."

Your Name:
Leave a Comment:
Security Code: