# A Quick Note on Situational Pitching

Earlier this year, when I delved into Zack Greinke’s 2019 season, I was impressed by how militantly Greinke follows one common-sense pitching rule: he does absolutely everything he can to avoid walks when the bases are empty, then pitches to avoid contact as soon as a runner reaches base. It’s a classic piece of pitching strategy, but the lengths to which he’s willing to change his approach to match the situation were eye-opening. Greinke’s dogmatism got me thinking: are there areas of pitching where context is so strong that it dictates specific strategies?

Luckily, there’s been a wealth of research on the intricacies of pitching to the situation. Colin Wyers, the current head of R&D for the Astros, wrote an excellent investigation of the topic that holds up well today, but many more writers have tackled this problem. Mitchel Lichtman periodically addresses pitching strategy, Matt Swartz took a good look at the question, William Spaniel considered one narrow case in an interesting way — the list goes on and on. Rather than attempt to go toe-to-toe with these excellent analyses (preview: I’d lose), I’m going to take a slightly different tack. Instead of talking situational pitching broadly, let’s look at a couple situations where behavioral change makes sense and see if pitchers can actually exert any control over it.

**Bases-Empty Aggression**

This is the question that got me on the case in the first place. Greinke unquestionably pitches to avoid walks with the bases empty. All pitchers do, to some extent — “never walk the leadoff batter” is a pearl of wisdom nearly as old as baseball itself. Just how bad is that walk? “Never allow a home run to the leadoff hitter” is just as true, yet you never hear that. What makes walks stick in peoples’ minds so much?

In a word, context. Home runs are better than walks for a batter in every situation, of course, but how much better they are isn’t constant. The linear weights that drive wOBA and wRC+ look at how many runs the average walk or home run adds to a team’s expected production, averaging across all cases proportionally. For example, a home run is worth 1.67 runs on average, or 1.4 runs more than a random plate appearance. A walk is worth considerably less, albeit still valuable: .59 runs, or .32 above average.

Those values consider every situation where each event occurred. We’re not interested in that; we’re looking for how many runs each is worth with the bases empty. Let’s start with the bases empty and no one out. A home run here adds one run to a team’s scoring expectation, naturally enough. To work out the value of a walk, we can turn to a run expectancy table. At the start of an inning in 2019, a team projects to score .54 runs. With a man on first and no one out, that value jumps to .93 runs. Therefore, we can say that a walk is worth .39 runs to the offense in that situation.

The difference between these bases-empty walks and the average value of a walk is meaningful. A bases-empty walk costs the pitcher 39% of a home run (.39/1) relative to a random plate appearance. That’s significantly higher than the percent of a home run that a walk is worth overall — 23% of a home run, or .32/1.4. The effect is there for doubles, too — a random walk costs a pitcher 41% as much as a random double. With the bases empty and no one out, it costs 65% as much.

These differences might not sound like much, but they add up. It’s difficult to ascertain the exact value walking people and pitching to contact selectively is worth, because redistributing where walks occur changes the relative value of each hit. Still, the math is clear. Hits are, relatively speaking, at their least deleterious to a pitcher with the bases empty. If a pitcher can allow extra contact in exchange for a lower walk rate, bases empty and no outs is a great place to do it.

This logic flips with two outs and the bases empty. With two outs, a runner on first isn’t particularly valuable to the hitting team, so giving up a bases-empty walk isn’t very costly — it’s worth a mere .13 runs in expectation. A home run, obviously, is still worth a run. The value swing is extreme. Take a look at a walk’s cost (or value, if you’re a hitter) relative to each type of base hit in a few bases-empty situations:

Outs | % of Single | % of Double | % of Triple | % of Home Run |
---|---|---|---|---|

0 | 100.00% | 64.70% | 47.70% | 39.20% |

1 | 100.00% | 63.70% | 40.20% | 26.40% |

2 | 100.00% | 56.70% | 46.10% | 12.80% |

The rule of thumb should be something closer to: avoid walks strongly against the leadoff hitter, and a bit with the bases empty and one out. By two outs, however, avoiding powerful contact is at a premium, so change your strategy.

This begs a question: can pitchers actually control when they issue their walks, selectively changing their walk rate based on the situation? The answer, in a broad sense, is a resounding yes. Consider this: with the bases empty and no one out this year, major league pitchers have walked 7.7% of the batters they’ve faced. In all other situations, that rate jumps to 8.4%, even after stripping out intentional walks. This effect persists from year-to-year as well: pitchers truly do pitch more aggressively when it makes sense to.

Can we measure and predict individual pitcher skill in this? That’s considerably dicier. The ratio between a pitcher’s bases-empty unintentional walk rate and overall unintentional walk rate isn’t stable at all. There’s almost no year-to-year correlation (r-squared of .06), and what’s there is mostly explained by walk rate. The lower your walk rate, the lower your ratio of empty walk rate to overall walk rate tends to be. Perhaps a few pitchers are outliers in terms of this skill (looking at you, Greinke, Dallas Keuchel, and Bartolo Colon), but for the most part, good-control pitchers use their good control to walk fewer batters when walks are at their most painful.

**Hunting for Grounders**

Another classic rule of thumb when it comes to situational pitching is to get grounders with runners on base. The direction of this one is pretty clearly correct — you can get a double play with runners on base but not with the bases empty, and double plays are worth, well, double the outs. How much is it worth? Let’s look at an example. The value of a ground ball with no outs and the bases empty is easy to calculate — all we need are the frequencies of hits and outs. With that in hand, we can work out the value of the grounder.

To get a large enough sample, I looked at the outcome of every ground ball hit with the bases empty over the past two years, and I folded errors in with singles. That gives us something like so:

Outcome | Out | Single (+Error) | Double | Triple |
---|---|---|---|---|

Percentage | 73.89% | 24.03% | 1.97% | 0.11% |

We can plug that into a run expectation table and say that on average, a team will score .466 runs when it starts the inning with a ground ball. Since the run expectancy before an inning starts is .541 runs, a grounder to lead off an inning is worth -.075 runs. Now let’s take the same logic and apply it to the situation with a runner on first and no out:

Outcome | Lead Runner Out | Trail Runner Out | Double Play | Single | Double | Triple |
---|---|---|---|---|---|---|

Percentage | 22.10% | 13.63% | 33.68% | 28.20% | 2.22% | 0.18% |

With a runner on first and no one out, teams have scored .933 runs per inning this year. After a ground ball though, that number drops to .766 runs per inning. The ground ball costs them .167 runs, more than twice as much as it decreases run expectancy with the bases empty.

Now that we have the value of getting a ground ball with runners on base, the question becomes whether pitchers can induce grounders. Luckily, the answer to this one is easy: yes! All they need to do is throw a two-seam fastball, a pitch that consistently gets a higher ground-ball rate than every other pitch. Case closed? Sadly, it’s not quite as simple as that.

As I investigated last month, two-seam fastballs get more grounders, but they also get less whiffs, and contact can be dangerous with a runner on base. We can use the formula from that article to figure out the value per swing of two-seam and four-seam fastballs with a man on first and no one out by plugging in run values for grounders, line drives, and fly balls. The outcomes per swing look like this:

Pitch Type | GB | LD | FB+PU | Whiff | Foul |
---|---|---|---|---|---|

Two-Seam | 24.10% | 10.70% | 9.90% | 15.00% | 40.30% |

Four-Seam | 11.80% | 9.20% | 12.90% | 21.80% | 44.30% |

After summing the run weights of each event, we finally get our answer. For every 100 swings at a sinker with a man on first and no outs, the offense produces 2.4 runs below average. Four-seamers are a little worse for the pitcher — 1.6 runs below average production for the offense. As a side note, both pitches create below-average production because walks don’t involve a swing, and average production involves a decent amount of walks. Even though four-seamers are a more valuable pitch than two-seamers overall, the logic from above holds: it’s better to throw two-seam fastballs with runners on base.

But that’s not the end of the story. Breaking balls are even better when the batter swings (somewhere around 3 runs below average per 100 swings), because avoiding contact (and the ever-increasing chance of loud contact) is of utmost importance when a home run scores multiple runners. Maybe the rule should be fastballs with the bases empty and whiffs with men on base, not sinkers with men on base.

There’s one last question in this vein: we know pitchers can throw two-seamers to get more grounders. Can pitchers get more grounders *with their sinker* with runners on base? In other words, is there something they can do aside from pitch selection to influence ground balls? Swartz addressed this in his study, but I took another tack. I took pitchers who allowed 100 balls in play off of two-seam fastballs with the bases empty in 2017 and 2018, and I compared it to those pitchers’ ground-ball rates on two-seamers in double-play situations. That looks like this:

There was, essentially, no bias towards a higher ground-ball rate in double-play situations. Pitchers whose two-seam fastballs generate a lot of grounders do so in all situations, but pitchers who get more grounders overall don’t have any special skill to increase that rate with a runner on. In fact, the overall ground-ball rate was basically identical in both situations. Want a pitcher who can get a grounder with runners on? Look for a pitcher who can get a grounder, period.

Don’t walk the leadoff batter. Get a grounder to turn two. These are classic rules of thumb, and the math totally supports them. Where it gets confusing, however, is whether pitchers can exert a lot of control over them. From my vantage point, it looks like most pitchers are already doing what they can. Pitchers already avoid leadoff walks more than all other walks. They throw more two-seam fastballs with runners on base. The only problem is, batters also know these things. They’re trying to walk to lead off an inning and to hit the ball in the air with runners on base. Baseball is hard!

Unsurprisingly, it seems like pitchers understand situational pitching well. When someone walks a leadoff runner or doesn’t get a grounder with runners on, it’s probably not because they just didn’t realize what they should be doing. Baseball is unpredictable, after all — round ball and round bat, etcetera. If your team’s pitcher is out there nibbling at the edges of the strike zone to lead off an inning, you’re right to be frustrated. Know, though, that they know it too. Most pitchers decrease their walk rates by a little bit when game situation dictates they should. They’re trying their hardest to pitch to the situation. They just can’t all be Zack Greinke.

Ben is a contributor to Fangraphs. A lifelong Cardinals fan, he got his start writing for Viva El Birdos. He can be found on Twitter @_Ben_Clemens.

The effect for the DP might be slight. Interested in the slope of the regression line. is it 48%, or is it 54%?

I’m traveling right now, so I can’t run all that many calculations on it, but the slope of the regression line is 41%. That said, the deviations are really random: if you try to predict gb% increase with runners on using bases-empty gb%, the r-squared is 2%.