January 17, 2011
Ahead in the Count
One of the pitchers I enjoyed watching the most while I was growing up was Tom Glavine. Even though I was a Phillies fan and frequently saw him victimize my favorite team, I was impressed by the expertise he demonstrated on the mound, and how he perfected his craft. Glavine remains the premier example of a pitcher who out-pitched his peripheral statistics; he was greater than the sum of his parts. For the amount of strikeouts, walks, and ground balls that Glavine got in his career, he should never have been able to keep runs off the scoreboard as well as he did.
It was not that Glavine was particularly talented at preventing hits on balls in play—his .285 career BABIP is similar to his teams’ .289, albeit lower than league-average. Instead, I would argue that he knew when and how to challenge hitters, and as a result he was able to prevent home runs when they would cost him the most, and was willing to walk those hitters that were likely to do the most damage.
Glavine allowed 2.3 percent of hitters to homer when the bases were empty, but only 1.5 percent when men were on base; to some extent, he minimized the damage of the long ball by allowing solo shots. He unintentionally walked 8.9 percent of hitters with runners on, but issued only 6.2 percent unintentional free passes with none on. With runners in scoring position, his strike-zone evasion became more extreme, as he unintentionally walked 11.3 percent of hitters in those situations. The seeming result was that Glavine was able to find his way out of jams, selectively pitching around the hitters that were likely to knock runners in, and attacking the weaker ones behind them. During the 2003-08 years when SIERA can be computed, Glavine allowed a 4.84 SIERA despite only a 4.11 ERA.
Do many other pitchers exhibit slightly less extreme tendencies to pitch to the situation (with Glavine being the most extreme example along a continuum)? Or, are most pitchers very similar to each other in this regard, with Glavine unaccompanied in his strategic prowess? I set out to answer this question by looking at situational pitching, identifying situations in which pitchers would find walks, strikeouts, and ground balls most useful, and seeing if some pitchers outdid their usual tendencies in these situations more than others. If so, I was curious whether SIERA had been picking up some of this effect.
Last month, I looked at ground-ball wizards to understand SIERA better. Since there is a negative coefficient on the squared term of ground balls in SIERA, I knew that the more ground balls a pitcher allows, the increased effectiveness each grounder has. In other words, the difference between otherwise similar pitchers with 60 and 50 percent ground-ball rates is greater than the difference between otherwise similar pitchers with 50 and 40 percent grounder rates. Could it be that ground-ballers are also able to induce worm killers when they need them most?
While ground balls are usually good for a pitcher, they are most useful with a man on first base and less than two outs because of a potential double play. The first thing I checked is whether pitchers who were able to get more ground balls in double-play situations in one year were able to repeat the feat the following year.
Among the 461 pitchers with 500 batters faced in consecutive seasons between 2003 and 2010, the correlation between getting extra ground balls in double-play situations one year and the next was only .096.
Given the small sample size, I thought it would also be wise to check the ground-ball rates of pitchers in double-play situations against situations which did not especially call for a ground ball much at all—when the leadoff batter came up. The year-to-year correlation between the differences in grounder rates in double-play situations versus against leadoff hitters was just slightly higher, .105.
To see if SIERA might be picking up this effect, I checked whether this was a particularly common tendency for ground-ball pitchers. Do these hurlers get ground balls when they need them more than other pitchers? I checked the correlation between ground-ball rate and the following year’s ground-ball rate in double-play situations net of the following year’s overall ground-ball rate, and found that the correlation was basically non-existent: -.018. The next year’s difference in ground-ball rates in double-play situations versus worm-killing rates against leadoff hitters was just -.048.
Even though there were a few pitchers who were able to get ground balls when they needed them, these pitchers were not especially likely to be ground-ball pitchers. Essentially, ground-ball wizards got worm-beaters equally in all situations.
However, strikeout pitchers seemed to do better at inducing grounders when they needed them. The correlation between strikeout rate and the added ground-ball rate in double-play situations the following seasons was .066, and the correlation between strikeout rate and the difference between ground-ball rates in double-play situations versus the ground-ball rate against the leadoff hitter was .144. In fact, a regression on this data shows that the previous year’s strikeout rate is actually a better predictor of strategically inducing grounders than the previous year’s measure.
Pitchers who walked more batters had a slight tendency to get double plays when they needed them, too. The correlation between walk rate and the following season’s added grounder rate in double-play situations was .058, and the correlation between walk rate and the following season’s added worm-beating rate in DP situations relative to ground-ball rate against the leadoff hitter was .066. Neither number is staggeringly large, but it appears that pitchers who walk many hitters, and especially pitchers who strike out a lot of hitters, are both more likely to get grounders when they need them.
This has taught us the following about a pitcher’s ability to strategically induce more of their ground balls in double-play situations:
Among other reasons, strikeouts are also great for pitchers because they prevent baserunners from advancing. This is most important when there is a man on third and less than two outs, when almost any contact could advance and plate the runner. This is why, to define situational strikeout rate differences, I used strikeout rate with a man on third and less than two out relative to overall strikeout rate, and also relative to strikeout rate in twin-killing situations. I was curious whether pitchers were particularly good at getting strikeouts in necessary situations on any consistent basis.
The answer appears to be a decisive 'no.' The correlation when using overall strikeout rate as the baseline was just .002, and using double-play situations as a baseline was just .023. In other words, and perhaps obviously, managers should turn to their best strikeout artist when the situation calls for a K.
Think about that for a second, because of what the results mean. Strikeout pitchers did not reach back and strike more hitters out in these situations as compared with other pitchers. The correlation between strikeout rate and the next year’s situational strikeout rate jump was -.089 using overall K rate as a baseline, and -.026 using double-play situation strikeout rate as a baseline. If anything, pitchers who got fewer strikeouts were apparently more likely to reach back for extra cheddar when they needed it most, but even that effect was very weak.
The correlation between situational strikeout rate and walk rate was -.04 regardless of the baseline. Ground-ball rate correlated weakly with situational strikeout rate too, but again only slightly. Using overall strikeout rate as the baseline led to only a .042 correlation, and using double-play situational strikeout rate as the baseline, grounder rate only had a .088 correlation with situational strikeouts.
In general, it appears that there are not strategic strikeout pitchers. Specifically, we have learned that:
Even Tom Glavine had a similar strikeout rate with men on and with bases empty.
Unlike ground balls and strikeouts, strategic walks are obviously a part of the game’s strategy, because a pitcher can walk anyone they want. Of course, intentional walks are frequently issued with first base open and a good hitter at the plate, and some pitchers obviously get ordered to do this more often than others. Interestingly, though perhaps unsurprisingly, some pitchers are also more prone (or more willing) to pitch around hitters too.
The correlation between walk rate with men on versus with bases empty is .265, and the correlation between walk rate against the leadoff hitter relative to with men on was .250. Limiting this only to unintentional walk rate, we still get correlations of .219 and .209 using the same two baselines. This makes it clear that pitchers are able to walk batters strategically, and that not all unintentional walks are really that unintentional.
The correlation between walk rate and strategic walk rate is unsurprising as well. Pitchers who walk more hitters in general are less strategic with their command, while pitchers who walk fewer hitters are often doing so wisely.
The correlation between walk rate and the following season’s added walk rate with runners on (versus with bases empty) is -.130. Using unintentional walk rate instead, this number jumps to -.168.
For completeness, I used a different baseline—walk rate to the leadoff hitters, relative to walk rate with men on base. The correlation between that and the previous year’s walk rate is .096; using unintentional walks and the same baseline has a correlation of .129.
These tests highlight that pitchers who issue fewer walks unsurprisingly enough have relatively lower walk rates with the bases empty, and relatively lower walk rates to the leadoff hitter—just like Tom Glavine.
This is not especially correlated with strikeout or worm-killing skills, with all such correlations checking in below .10. The lesson is that the pitchers with better control are the ones to look for when trying to determine who will walk batters strategically.
In summary, as far as a pitcher’s ability to strategically walk batters, we've learned:
Unsurprisingly, Tom Glavine’s walks vary the most by situation.
Situational Pitching and SIERA
SIERA will actually pick up on some of these skills, because it is a regression-based estimator. For instance, pitchers who walk fewer hitters issue less damaging walks on average, so SIERA will note the correlation between walk rate and ERA (which includes more walks being bad in general and less control leading to more damaging walks in general), and it will attribute higher SIERAs to match the higher ERAs. Similarly, since pitchers with high strikeout rates are better able to get ground balls in double-play situations, SIERA will note the added correlation between strikeout rate and run prevention and credit pitchers extra for this skill.
While situational pitching is a rare skill, it certainly exists... in small doses. Undoubtedly talented in-game strategists like Glavine have existed, and there probably have been a number of less extreme Glavinesque arms over the years. However, very few pitchers have much of this overpowering ability to sequence strategically. Perhaps the best advice any pitcher can follow is that consistently pitching well is the best remedy for preventing runs.