Baseball can often be boiled down to a simple struggle between the wills and skills of one batter and one pitcher. At its base, reptilian core, this conflict is straightforward and Darwinian: pitchers want to record outs, while batters want to avoid making outs. But depending on the game situation, not all outs (or all non-outs) are the same. With the bases empty, any out will do-but with a runner on third and fewer than two outs, many fly balls and some ground balls can result in a run, while strikeouts and popups almost never do let anyone cross home. On the other hand, with a runner on first and fewer than two outs, a ground ball that results in a double play is usually better for the pitcher than a strikeout or popup.

These differences lead us to think that pitchers may use different approaches in different situations. We’ve heard it a thousand times: “That free pass to Aaron puts runners on first and second, so Haddix will be trying to get Adcock to bounce into a double play to at least preserve his no-hitter and send this one into the fourteenth.” Well, most of us haven’t heard exactly that (as much as we may have wanted to), but there seems to be a general belief that pitchers may (or should) try to alter their approach to induce the specific category of out that would be most beneficial in a given situation.

To get some sense of whether this actually occurs, let’s look at some aggregate pitching numbers for the 2005-2009 seasons (props to BP intern Dan Malkiel for the data assist). The chart below shows results for different categories of plate appearances, with intentional walks removed from all calculations. DP lists all plate appearances in which there is a runner on first base (and any other base) with fewer than two outs-essentially the situation when we might expect pitchers to try and induce a double-play grounder. R3 lists situations with a runner on third or runners on second and third, and fewer than two outs-times when a strikeout or popup might be most beneficial. Note that for these calculations I’ve put “first-and-third” and “bases loaded” situations into the DP bucket and not the R3 bucket, Lastly, I’ve listed the totals for all plate appearances during this time period (2009 numbers through last week):

Situational Pitching Splits, 2005-09
                        Change         Change           Change       Change
Split    PA-IBB  GB/PA   GB/PA  BB/PA   BB/PA   K/PA     K/PA  HR/PA  HR/PA
    DP  171,451  36.1%  +3.25    6.7%   -1.49   15.1%   -2.59   2.8%  +0.11
Non-DP  726,088  32.9%    --     8.2%     --    17.7%     --    2.7%    --

    R3   21,681  32.9%  -0.59    9.7%   +1.84   16.1%   -1.19   2.1%  -0.66
Non-R3  875,858  33.5%    --     7.9%     --    17.3%     --    2.7%    --

  All   897,539  33.3%    --     7.9%     --    17.2%     --    2.7%    --

Here we see the rates of ground balls, walks, strikeouts, and home runs per plate appearance, and the amount of change seen in those rates during a given situations compared to the aggregate of all other situations. So the +3.25 value for the change in GB/PA during the “DP” situations means the GB/PA rate (36.1 percent) is 3.25 higher in DP situations than in non-DP situations (32.9 percent).

And it’s that +3.25 that first stands out-there appears to be a measurable increase in GB/PA rate during the situation when ground balls may be most beneficial for the pitcher. We also see a small reduction in BB/PA, and a larger reduction in K/PA, during these double-play situations. This seems to lend credence to the idea that pitchers are more likely to be “pitching to contact” during these situations-avoiding walks and strikeouts by working more in the strike zone in the hopes that a routine two-hopper will get them out of a jam. There is also a minute increase (0.11) in the HR/PA percentage-but that increase is due entirely to the smaller percentage of plate appearances ending without contact, not due to batters marking harder contact, as the HR/Contact percentage (not shown) actually drops from 3.7 percent in non-DP situations to 3.6 percent in DP situations.

This article will focus on the DP situation, with a follow-up article on the R3 situation soon to follow. But just looking at this chart, the most interesting value in the R3 category is the -1.19 change in K/PA, just when strikeouts would seem to be most beneficial. Hitters seem to be doing a slightly better job of exerting their will in this situation-pitchers may want to get strikeouts, but hitters are more likely to be able to shorten up and figure out a way to put the ball in play.

If we accept that pitchers seem to be tailoring their approach to get more ground balls in double-play situations, the next question is doubtless whether this is universal, or if some pitchers are better at this brand of situational pitching than others. To start with, here’s a chart showing the top ten pitchers since 2005 in GB/PA during the DP situation:

Highest Ground-ball Rates in DP Situations, 2005-09 (390 PA min.*)
                 DP Split  Other  DP Split  Other   GB/PA    K/PA    GB+K
Pitcher           GB/PA    GB/PA    K/PA    K/PA    Change  Change  Change
Derek Lowe        53.2%    49.5%   12.2%    16.1%   +3.74   -3.85   -0.11
Fausto Carmona    52.9%    45.0%    9.5%    15.2%   +7.91   -5.77   +2.14
Aaron Cook        52.3%    48.4%    7.1%    10.7%   +3.97   -3.65   +0.31
Brandon Webb      50.9%    48.4%   16.9%    19.6%   +2.52   -2.58   -0.06
Jake Westbrook    50.7%    47.5%    8.8%    14.3%   +3.18   -5.49   -2.31
Tim Hudson        50.2%    46.3%   11.8%    15.2%   +3.84   -3.43   +0.42
Chien-Ming Wang   49.9%    49.5%    7.9%    11.7%   +0.38   -3.85   -3.47
Greg Maddux       46.9%    43.8%    9.3%    13.9%   +3.16   -4.68   -1.52
Carlos Silva      46.7%    39.7%    6.4%    10.3%   +6.97   -3.88   +3.09
Zach Duke         46.6%    41.2%    9.6%    12.2%   +5.33   -2.52   +2.81

*Why 390 PAs? Because it's a sizable sample, and it ensures that
 Tim Lincecum is in the sample; a dash of Tiny Tim always adds zest.

Not surprisingly, this list is chock full of pitchers who support the Infielder Full Employment Act during any situation, not just when a double play is in order. Eight of the top ten pitchers in overall GB/PA are also in the top ten for GB/PA during the DP situation; the other two (Roy Halladay and Jamey Wright) make the top twenty, while their replacements here (Silva and Duke) can be found in the top twenty in overall GB/PA rate. Each of them gets even more ground balls when it counts, lead by Carmona’s increase of almost eight batters per hundred.

But that’s not unique to the pitchers on this leaderboard-of the 113 pitchers in the sample, fully 101 saw their GB/PA rate go up in a DP situation. As we saw in the first chart, some of this increase is due to a decrease in walk and strikeout rates, and thus an increase in contact; even if a pitcher’s GB% (per ball in play) stays the same, his GB/PA would increase. However, when I looked at GB% instead of GB/PA, the overall rate increased by 1.81, and 80 of the 113 pitchers saw their GB% go up as well. So it appears as if both overall, and in a large majority of individual cases, pitchers are getting more groundballs by both allowing more balls to be put in play, and getting a larger percentage of those to be hit on the ground.

Let’s take a look at who’s best at improving their GB/PA rate in DP situations relative to their work in other situations:

Biggest Increase In Ground-ball Rates in DP Situations, 2005-09
                 DP Split  Other   DP Split  Other    GB/PA    K/PA    GB+K
Pitcher           GB/PA    GB/PA     K/PA     K/PA    Change  Change  Change
Jeff Suppan       45.2%    35.9%    10.3%     12.8%   +9.26   -2.52   +6.74
Chris Young       28.1%    19.6%    17.1%     21.8%   +8.55   -4.69   +3.86
Jason Jennings    40.5%    32.5%    13.2%     15.5%   +8.04   -2.28   +5.76
Fausto Carmona    52.9%    45.0%     9.5%     15.2%   +7.91   -5.77   +2.14
Carlos Zambrano   40.0%    32.1%    13.7%     21.6%   +7.90   -7.90   -0.00
Chad Billingsley  38.7%    30.9%    16.9%     22.7%   +7.76   -5.90   +1.86
Oliver Perez      27.9%    20.5%    15.5%     22.5%   +7.45   -7.00   +0.45
Bronson Arroyo    37.9%    30.6%    13.9%     16.4%   +7.42   -2.63   +4.79
Livan Hernandez   39.8%    32.5%     9.2%     12.6%   +7.30   -3.40   +3.91
Ian Snell         36.4%    29.2%    16.8%     19.0%   +7.18   -2.26   +4.93

The “GB/PA Change” column shows the amount by which each pitcher was able to increase their GB/PA rate relative to other situations. Only Carmona appears on both lists, while Jeff Suppan appears to be the most astute pitcher around as far as inducing more ground balls when they’re needed, going from a solid 36 percent to a spectacular 45 percent in DP situations. You’ll notice a few other pitchers on this list who wouldn’t take umbrage if you called them “crafty veterans”. But take a look at the “K/PA Change” column-every pitcher here realizes a reduction in K-rate, but there’s a wide variation, sometimes to the pitcher’s detriment. For example, Carlos Zambrano’s modern-day alchemy seems to directly convert strikeouts into ground balls. This is likely not helpful for Big Z’s run prevention, since it’s the baseball equivalent of splitting a pair of tens at blackjack-the strikeout is almost always a win, while the ground ball might be a double play, but could also result in a base hit, an error, a double over the bag, or any number of other bad outcomes.

To better understand which pitchers are best in DP situations overall we’ll need to take more into account than just ground-ball rate. The chart below ranks pitchers by a quickie measure I’m going to call PRIDE (Positive Results In Double-play Environments) just to keep from typing “GB+K-BBHBP” repeatedly. Unsurprisingly, PRIDE is calculated by summing the rates for strikeouts (always helpful) and ground balls (a desired result), then subtracting the rates for walks and HBP (usually harmful).

Highest PRIDE Scores in DP Situations, 2005-09
                                             PRIDE   PRIDE
Pitcher          GB/PA     K/PA    BB+HBP    Score   Change
Brandon Webb      50.9%    16.9%    6.9%     60.96   -0.37
Derek Lowe        53.2%    12.2%    5.4%     59.95   +0.80
Felix Hernandez   41.2%    21.4%    6.1%     56.44   +2.86
Chris Carpenter   43.1%    16.5%    3.3%     56.24   +0.01
Roy Halladay      44.8%    14.6%    3.5%     55.90   -1.43
Tim Hudson        50.2%    11.8%    6.9%     55.04   +0.94
Greg Maddux       46.9%     9.3%    2.1%     54.12   +0.38
Andy Pettitte     40.7%    19.3%    5.9%     54.05   +6.25
Aaron Cook        52.3%     7.1%    6.3%     53.13   -0.06
Jake Westbrook    50.8%     8.8%    6.5%     53.03   -1.69

One might expect that Brandon Webb would top this list, with Derek Lowe lurking in the corner. But now pitchers with other great peripherals like Carpenter and King Felix show up as well. These are the starters you may most want on the mound in a double play situation-they’ll get you groundballs and/or strikeouts, without too many free passes. The PRIDE Change column shows the change in the PRIDE score in DP situations compared to others, and again, with one notable exception, these pitchers are here mostly due to their innate abilities, not their savvy situational pitching skills. So who improves their PRIDE score the most during double-play situations?

Most Improved PRIDE Scores in DP Situations, 2005-09
                                       PRIDE    GB    K    HBP    PRIDE
Pitcher        GB/PA     K/PA  BB+HBP  Score   Chg / Chg / Chg    Change 
Jeff Suppan     45.2%   10.3%   5.6%   60.96  +9.26/-2.52/-3.68  +10.42
Jason Jennings  40.5%   13.2%   7.1%   59.95  +8.04/-2.28/-3.21   +8.97
Adam Eaton      36.0%   14.3%   7.9%   56.44  +6.29/-0.23/-2.02   +8.08
Chris Young     28.1%   17.1%   7.3%   56.24  +8.55/-4.69/-2.98   +6.85
Noah Lowry      36.4%   13.2%   7.5%   55.90  +5.89/-2.60/-3.51   +6.80
Ian Snell       36.4%   16.8%   8.9%   55.04  +7.18/-2.26/-1.61   +6.54
Zach Duke       46.6%    9.6%   3.6%   54.12  +5.33/-2.52/-3.50   +6.31
Andy Pettitte   40.7%   19.3%   5.9%   54.05  +3.69/+1.39/-1.17   +6.25
Jon Garland     41.1%   11.2%   4.1%   53.13  +4.04/-0.58/-2.71   +6.18
Jeff Francis    37.9%   13.9%   5.4%   53.03  +5.11/-2.26/-3.10   +5.95

Unlike the current National League, here we can witness Jeff Suppan blowing away the competition-canny situational pitching and a willingness to take the ball may be what’s keeping him employed at this point. Most of these pitchers see relatively small drops in their strikeout rates (though in some cases they couldn’t get much lower) combined with larger drops in their walk rates in double-play situations. And then there’s Andy Pettitte, one of only seven players in the sample to actually see his strikeout rate increase. Add in his otherworldly ability to hold runners, and you have to wonder whether an opposing team’s run expectancy goes up less when getting a runner to first against Pettitte than any other pitcher in baseball.

So, I expect you’re curious about the trailerboards? Here they are:

Lowest PRIDE Scores in DP Situations, 2005-09
                                           PRIDE   PRIDE
Pitcher          GB/PA     K/PA   BB+HBP   Score   Change
Kyle Davies       32.7%     9.5%   11.5%   30.66   -3.94
Oliver Perez      27.9%    15.5%   10.8%   32.58   +4.33
Brian Bannister   28.9%    11.2%    4.9%   35.19   -5.27
Barry Zito        32.2%    14.0%   10.7%   35.44   +0.39
Jarrod Washburn   31.7%    11.6%    7.0%   36.24   -0.49
Casey Fossum      28.9%    17.1%    9.6%   36.39   -0.72
Chad Gaudin       32.3%    16.1%   11.9%   36.47   -3.38
Scott Baker       29.0%    13.1%    4.8%   37.33   -2.68
Matt Garza        31.7%    14.1%    8.4%   37.47   -2.52
Jered Weaver      26.0%    17.1%    5.6%   37.47   +0.64

Least Improved PRIDE Scores in DP Situations, 2005-09
                                        PRIDE    GB     K    HBP   Total
Pitcher          GB/PA   K/PA   BB+HBP  Score    CHG / CHG / CHG   Change 
Brian Bannister  28.9%   11.2%   4.9%   35.2%  -6.56/-2.48/-3.77   -5.27
Joel Pineiro     41.6%    8.9%   6.6%   43.9%  -0.27/-4.53/+0.37   -5.17
Matt Morris      40.7%    7.9%   6.8%   41.9%  +1.81/-5.85/-0.02   -4.02
Kyle Davies      32.7%    9.5%  11.5%   30.7%  +4.76/-8.02/+0.67   -3.94
Chad Gaudin      32.3%   16.1%  11.9%   36.5%  +0.10/-2.79/+0.69   -3.38
Wandy Rodriguez  33.6%   16.5%  10.0%   40.0%  +1.60/-3.59/+1.29   -3.28
Joe Blanton      34.5%   13.5%   7.1%   40.9%  -1.38/-1.36/+0.42   -3.16
Sidney Ponson    44.7%    7.9%  10.2%   42.4%  +2.82/-4.31/+1.38   -2.88
Justin Verlander 26.2%   19.7%   7.7%   38.2%  -3.26/-1.22/-1.61   -2.87
Scott Baker      29.0%   13.1%   4.8%   37.3%  +2.91/-6.28/-0.69   -2.68

Surprisingly, Brian Bannister seems to suffer the most from his approach during DP situations-his GB/PA rate drops by far more than any pitcher in the sample, partly due to his reduced strikeout and walk rates, but mostly due to an incredible drop of 11.2 points in his ground-ball percentage. Known both near and far as a self-aware, statistically-informed player and a stand-up guy, Banny might want to think a few more deep thoughts about how he should work with a runner on first.

Next week I’ll take a deeper dive into the “Runner on Third” situation, and see whether pitchers or hitters tend to have the upper hand in their ongoing battle for situational dominance.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
given that so called groundball pitchers would probably want to get the groundball in most normal situations, their approach probably wouldn't change when it is a dp situation. perhaps comparing their gb ratios in dp situations with those in runner on third situations would better showcase the results of situational pitching.
One step further...I'm going to suggest that we don't know that the data shows that ground ball pitchers change their approach in double play situations. I don't see where Ken controlled for situations where the pitchers might be trying to pitch situationally in the data he is comparing to the double play data.

So instead it's possible that the numbers he is seeing in double play situations is indicative of the 'normal' ground ball inducing talent, and the larger sample differs because it includes situations where the pitcher is trying to keep the ball out of play, via strikeout or walk.
Good question. I split the numbers three ways: DP (any time there's a runner on first and fewer than two out); else R3 (any time there's a runner third, not a runner on first, and less than two outs); and OTHER (any time there's 2 outs, or bases or empty, or there's only a man on second). In R3, which only happens in 2.6% of plate appearances, you would definitely expect situational pitching; in OTHER (78.4% of plate appearances), any kind of out will do (with the tiny caveat that batters may be trying to hit to the right side to move a man on second over).

The charts I ran compare DP to the sum of R3 and OTHER, and shows GB/PA +3.25. Comparing DP to OTHER (which I think controls for the effects you're describing) also shows GB/PA +3.25. The number hardly changes, primarily because R3 doesn't happen that often.
Your "Most Improved PRIDE Scores in DP Situations, 2005-09" list includes nine National Leaguers and Andy Pettitte. Might I suggest removing all situations in which the pitcher bats? It seems likely that a sizable number of times in which the pitcher comes to bat with less than two outs involve a running on first base. Because the batter before the pitcher is often walked.
Very true -- I didn't control for that, and I should have (or at least run separate lists for AL and NL). I'll fix that for the next article. The other (primarily) AL pitchers who score +5 or more are Jon Garland, Kevin Millwood, Tim Wakefield, Carlos Silva and Gavin Floyd.
Thanks Ken. It's an insightful read, nonetheless.
I'll second Brian Kopec's comment on finding some way to control for confounding factors, and also ask if you've looked at whether the delta rankings are stable from season to season.

GB% is essentially a binomial random variable, and in 200 PA (best case split for someone who just qualifies for the study), +/- 2 standard deviations is +/- 6.5% of ground ball rate. That's a large swing relative to the PRIDE deltas you're drawing conclusions from; you might be measuring pitchers' serendipity more than their situational ability.
In aggregate, the deltas are pretty stable from season to season -- the first chart, if I broke it down by season, would show increases in GB/PA in each year of the sample between +3.1 and +3.5. I didn't look at variation by pitcher between seasons of the sample because each season by itself is a pretty small sample -- but I would suspect it would vary more, of course.

I agree that the individual rankings, based on much smaller sample sizes, aren't as compelling -- they're more for fun (though the numbers for Suppan and Bannister really interested me). But the aggregates (GB/PA and GB% going up in total, and going up at least some for most pitchers) seemed compelling.

ken, is it possible to show how many times, or what the rate was, of actual double plays in this situation? It might not make sense for individual pitchers, as they lose control over the outcome after they throw their pitch, but maybe it shows something on a team level about the quality of the defense behind them?
I can do that -- I have the data. I'll touch on that next week.
It makes sense to consider both K and GB as "good" things but giving them equal weight gets away from the situational idea - Ks and GBs are always good (that's why they both lower QERA), it's just that Gbs are particularly good in DP situations and you need a means yto pick that up. I suggest looking at how much a GB reduces run expectancy compared to a K in a DP situation and weighing accordingly.
Interesting idea. Although GBs are better in DP situations than non-DP situations, they may or may not be better than Ks in DP situations -- a K is always an out, while a GB could wind up with one out, two outs, a single, a double, a triple, etc. The data I was working with isn't currently sliced that thin (I don't have the pre-PA and post-PA base/out states, f'rinstance) -- but I'll take a look at this. Good suggestion.