Second Thoughts Game #96-97: Cleveland Takes Two in Detroit
Kluber Provides An Introduction to Run Differential vs. W-L
Corey Kluber is not merely an excellent pitcher, he has taken it upon himself to educate about run-differential at the expense of his ERA.
At this point, Corey Kluber’s dominating performances – on Saturday, 8.2 IP, 2 R/ER, 7 H, 1 BB, 10 K, 114 Pitches, 79 Strikes, 16 Swinging Strikes – come less as a shocking blindside than subsequent episodes in the ascent of a remarkable pitcher. While Felix Hernandez’s career year has shut down the AL Cy Young discussion for now, Kluber – along with 9 other pitchers – has an argument for being the 2nd-best pitcher in the AL.
It’s worth recalling that Kluber’s 2 ER-allowed afternoon may not have represented how well he truly pitched on Saturday. At the end of the eighth inning with 101 pitches, Cleveland was winning 6-1 after the offense scored two insurance runs in the top of the ninth. Given that Kluber’s dominance was most pronounced late in the game – including a seventh inning in which he struck out the side and a 1-2-3 eighth – it was unlikely that Franconawouldn’t have let Kluber try to earn the complete game; moreover, perhaps more importantly, Cleveland was in the first game of a doubleheader. If Cleveland could get through the first game without using even one reliever, that would have been an unqualified, unmitigated success. It was a possibility that could not have escaped Francona’s notice, and while it may not have been the primary driver, it was undoubtedly a factor in sending out Kluber for the ninth. It was a ninth inning that saw two doubles and a run that would not have been charged to Kluber in a closer game.
On one hand, this sort of situation would at face value suggest the validity of ‘pitching to the score,’ but it might more appropriately be called managing to the score. Those hits did not occur because Kluber was taking score-related liberties to get three outs, they occurred because Miguel Cabrera and Nick Castellanos had seen Kluber three times earlier in the game. Give a hitter like Cabrera four times to see a pitcher’s complete arsenal, and the odds distinctly favor the hitter – one needs consult Danny Salazar to confirm that.
On the other hand, ‘managing to the score,’ or pursuing strategies that one knows to be sub-optimal because the game outcome is unlikely to change, might prima facie cast some doubt on the validity of run differential as a predictive mechanism. If a team loses by 15 rather than 10 in a blowout because a manager used a the last man in the pen rather than the second-to-last man in the pen, or if a manager uses a fringe reliever solely for mop-up duty and he costs 5 runs over the course of a season, it isn’t apparent that those five runs are reflective of a team’s ‘true’ talent level.
Nevertheless, while run differential – or its second- or third-order derivative stats that use root batting stats in lieu of runs scored, such as team wOBA-projected runs in place of actual runs – has its flaws, the root stats are nevertheless more reflective of the overall performance of a team to that point than is Win-Loss record. The reasons for this are many – Team W-L in one-run games having virtually no predictive power, in addition to the fact that gradient component statistics (runs) are, under most circumstances, far more predictive than their corresponding binary outcome stats (win-loss). The foremost reason for this is because of the imperfect alignment of input to output – a one-run win does not indicate the same mastery as a ten-run win, but the standings do not reflect this dissonance in mastery of the game; the same applies to losses. Moreover, W-L is a binary team outcome in the sense that there are only two possible outcomes, win or loss, whereas its component statistics, be it run differential or wOBA for/against, have a large spectrum that, over an entire season, can provide an extremely detailed synopsis of a team’s season and can situate a team’s performance with precise nuance. Run Differential does not account for injured players or underperforming players, so it is limited – but on the other hand, neither does Win-Loss.
The Indians started the Detroit series with a Run Differential of -8 on the year. At the end of the day, Cleveland stands with a +5 run differential. Run differential, wOBA-based or otherwise, isn’t perfect, with flaws that its proponents will readily grant, yet these flaws do not outweigh their predictive ability. It’s hard to imagine, after all, that the 13-run swing in run differential over three games in Detroit, or anything about any series in Detroit - be it Kluber’s two-run, eight-inning dominance, McAllister's successful 7-inning outing, or the renaissance of struggling offensive figures – could possibly be described as ‘cheap.’
John can be reached on Twitter at @JHGrimm. He can also be reached by e-mail at firstname.lastname@example.org.