Are Relievers Wilder Upon Entry?
On Wednesday afternoon, Liam Hendriks entered a tough situation. There were two outs in the bottom of the ninth, but his margin for error was nonexistent. The bases were loaded, and the White Sox were locked in a tie game. One hiccup in command, four slightly misplaced pitches, and the game would be over.
Do pitchers have less command when they enter? Is it worth worrying about whether a pitcher might not have it that day? I have no earthly idea, so I decided to investigate. First things first, though: I wasn’t actually sure what I was investigating. Time for some experimental design.
What about the walk rate, but only on the first batter faced by a new reliever? That’s certainly a number I could look up. That checks in at 8.1% from 2015 to present (I used the Statcast era even though there’s no Statcast data involved in this query, just for consistency’s sake). Over the same time frame, the overall reliever walk rate is 9.3%. Case closed, let’s go get brunch.
Only, that’s a bad comparison. We’re not comparing apples to apples. If we’re actually going to look into whether pitchers are particularly likely to come in and not have it, we need to compare like to like. Take the immortal Sugar Ray Marimon, who made 16 appearances for the Braves in 2015. He was a one-hit wonder, though “wonder” might be strong: he compiled a 7.36 ERA in 25.2 innings before decamping to Korea.
One thing he was particularly inept at: walking the first batter he faced. Six of those 16 batters walked down to first, a gruesome 37.5% walk rate. After that, he settled down: his walk rate against all other batters was 7.9%. Clearly, Marimon performed far worse against the first batter he faced. I certainly wouldn’t bring him in with the bases loaded — though, yeah, I probably just wouldn’t bring him in if I could avoid it.
On the other hand, Austin Brice was locked in when he entered the game in 2020. In 20 appearances for the Red Sox, he didn’t walk the first batter he faced a single time. Spectacular! Of course, he walked all other batters at a 19.4% rate, which is not great. His 5.95 ERA was in spite of his performance against the first hitter he faced, not because of it.
Clearly, I could cherry pick examples all day — I have a whole spreadsheet full of them. Jace Fry was downright Briceian last year; he didn’t walk any of the 18 first batters he faced, then issued free passes at an 18.5% clip the rest of the way. Trevor Rosenthal went full Sugar Ray in 2018; he walked a whopping 40.9% of the 22 initial batters, then 27% of the remaining batters — yeah, okay, he was pretty wild that year.
The main point of this name-dropping is to drop names. The secondary point, though, is to introduce the next way I analyzed the data. I took every reliever-season from 2015-21 and split them up into two parts: the first batter of an appearance on one hand, every other batter faced on the other. This gave me 4,098 observations, from Alex Claudio’s 83-appearance 2109 season all the way to a bevy of one-appearance journeymen.
From there, I calculated each pitcher’s “first batter walk adjustment.” That’s a mouthful, but it’s simple in practice; it’s their walk rate against first batters minus their walk rate against all other batters. Walk 20% of first batters and 10% the rest of the way? You have a walk adjustment of 10%. Go from 10% first-batter walk rate to 20% the rest of the way, and you’d check in at -10%.
Next, I weighted each observation based on the number of first batters faced. I’m honestly not 100% sure that this is the right weighting; my thinking is that this will sufficiently smooth out randomness over the sample by de-emphasizing extreme values accrued in small samples, but I’m not a statistician, and I don’t even play one on TV, so if you have a better suggestion for how to weight things, I’m all ears.
What are the results? The average reliever actually runs a lower walk rate on their first batter of the night. It’s not a huge margin, but a 1.5 percentage point walk rate decrease is nothing to scoff at. Limit it to pitchers who made 20 or more appearances in a given year, and nothing much changes; that number clocks in at 1.8 percentage points.
I’m not sure if I’ll call this result counterintuitive, but I expected pitchers to run slightly higher walk rates while they adjust to the game mound and the umpire’s zone. On the other hand, they’re at their freshest, and relievers aren’t noted for holding their form as they tire.
Further research is merited here, because plate appearances aren’t created equal. The base/out state, as well as the score of the game, affects how damaging a free pass is. The identity of the batter matters as well, though that’s likely offset somewhat by the size of the dataset I used. I’d rather look at a different permutation today, though, so I’ll table those for my future self or an enterprising reader.
Here’s what I wanted to know: does this first batter adjustment affect all pitchers equally? Are a few control artists who fill the zone on the first appearance skewing the sample, while the Rosenthals of the world take a few batters to locate the zone? I split the dataset up into five quintiles based on overall walk rates to see:
| Quintile | Overall BB% | Entry BB% | Rest BB% | Difference |
|---|---|---|---|---|
| Lowest | 4.7% | 3.8% | 5.0% | -1.2% |
| Low | 7.3% | 6.1% | 7.6% | -1.6% |
| Medium | 9.2% | 7.9% | 9.6% | -1.7% |
| High | 11.6% | 10.4% | 12.0% | -1.6% |
| Highest | 16.2% | 15.4% | 16.5% | -1.1% |
Oh. Huh. So much for that theory. Every group is sharper, when it comes to walk rate at least, on the first batter they face. There’s not much difference in how much they improve by, though. It seems to be one of two things: an inherent trait, or something unique about the first batter that I haven’t considered.
You might think that this is a handedness issue — and indeed, I thought that. Relievers with the platoon advantage have walked 8.2% of their opponents since 2015, while those without it have walked 10.5%. Managers usually bring in relievers to get the platoon advantage. There’s your gap!
Only, it doesn’t really work that way. Relievers had the platoon advantage in 60% of their first batters faced in this sample. That’s a big number, but it’s not particularly elevated relative to their overall mark. Since 2015, relievers have had the platoon advantage on 53.1% of all plate appearances. Getting the advantage that extra handful of times “should” be worth roughly 0.2 percentage points. We’ll have to look elsewhere for any variables getting in the way of our answer. It seems, for the most part, that relievers simply have control, on a relative basis at least, when they enter the game.
Hendriks, by the way, wriggled out of his bases-loaded jam. He didn’t even reach a three-ball count. The risk of a reliever coming in wild is never zero, but just like it does in aggregate, it failed to rear its ugly head on Wednesday. Hendriks lost the game the honest way in the 10th, with a pair of singles driving in the runner that automatically started the inning on second — okay, the semi-honest way. But his leadoff wildness didn’t do him in — and yeah, it doesn’t appear to do many pitchers in, as it turns out.
Ben is a writer at FanGraphs. He can be found on Bluesky @benclemens.
Ben, I love your articles, especially the ones that end up with “Man, IDK.” I mean this sincerely.
I did not expect this result.
But then I remembered that, anecdotally, when it’s my team bringing in the reliever, he always walks the first guy…while when it’s the OTHER team bringing in the reliever, they always strike my guy out.
Ben, would eliminating intentional walks affect the result? The reliever (almost) never intentionally walks the first batter faced. If they are entering with runners on base, they may be more likely than average to intentionally walk subsequent batters.
Good call — looks like that eliminates about a third of the discrepancy.
Might relievers walk more when entering in the middle of an inning? Relievers entering at the beginning of an inning presumably had at least 1/2 inning to warm up. Some of those relievers entering mid-inning may not be loose yet if the manager wasn’t expecting to need him yet. I could see this possibly leading to more first batter walks.
This is extra cool because it contradicts my mental anecdata and I’m happy to be corrected.
I would compare walk rates by base-out situation, first batter faced for a reliever is most likely to take place with no one out, none one. Possible walk rates are lower in that situation vs. say 2nd and 3rd two outs.
Other potential tests of interest:
* Zone rate in lieu of walk rate. Not really a great proxy for command, but fits the “control” or “wildness” concept a bit better.
* K%-BB% instead of BB%. Or, maybe K%+BB%. Basically, wildness doesn’t necessarily mean walks. Wildness could mean grooving pitches. So I guess maybe you could do “heart %” or “power cross %” or even “down the middle %.”
With the second bullet, my “concern” is that if we ask whether relievers are wilder, and see a lower walk rate upon entry, that could be either because A) relievers are in fact not as wild when entering or B) relievers are wild but miss in the zone and the pitches get put into play at an elevated rate, meaning fewer opportunities for walks.
I don’t think relievers are really prone to “laying one in” anymore the way you think of a starter throwing a down the pipe first-pitch fastball to the game’s first batter in days of yore, but if they were to any extent, that’d torque the walk rate downward.
I’ve always wondered about this one, thanks for scratching that itch! Now I am wondering about starting pitchers. Do they walk the first batter less than average? It’s a bad way to start the game if you’re a pitcher.
Let me add a potential wrinkle. Compare the walk rate of the second faced batter to the remaining batters, then the walk rate of the third faced batter. To the remaining batters,… Ceteris paribus, I imagine a reliever gets less and less effective over time. I would also imagine the walk rate to increase as the reliever tires. How do you test this? I’m sure some statistical tests exists that provides a factor that estimates the rate at which any variable decreases over time. One way I envision it is a simple x and y graph. X measures walk percentage and Y = the number of batters faced. Connecting the dots produces a curve. The slope of that curve at any point would provide the rate of change. Dang baseball stats are fun! Thanks again, Ben for another breath of fresh analysis…
This is basically what I expected. There is usual a very exact plan for the first batter – that’s the matchup being leveraged more often than not. After that, sometimes there’s a plan, sometimes the guy just has to face two more batters.
That said, I kinda figured you’d find for the null hypothesis here – no effect.
There could be very mental/adrenaline effect-type explanations for certain individuals, but I prefer to stick with more tangible explanations. Think of it another way, if relievers were wildest upon entry, we never would have developed the TLR special of using three pitchers to retire three batters. *Somebody* would have noticed it’s a negative EV strategy. These meltdowns stand out so much specifically because they’re rare.
great article
Perhaps a small matter, but if you’re talking wildness, what about WP and HBP on the first batter faced?
I would venture that the reason is; relievers are often brought in for situations where they have to get the guy currently up to bat. As mentioned earlier, eliminating intentional walks from your sample helps with this somewhat. But most importantly, they have to attack the first batter more often (see Hendriks example) where they may have more breathing room and be more willing to pitch around other batters.
So essentially, that means the pitchers are coming in ready-to-pitch and not extra wild, and the data is showing the different game state/situations in which they are presented. Great stuff!
Ben, additional question for you. Do batters swing at a higher percent of first pitches from the first reliever they see in a game (or any reliever they see in a game, I suppose)? I wonder if batters, going from a starter with a more varied arsenal to a pitcher with a typically more limited (fastball/whatever) arsenal, expect first-pitch fastballs in reliever situations, so they swing at them at a higher rate than normal. If they whiff or get in the hole, they’re already at a disadvantage and less likely to walk. The IBB seemed to account for the a third of the differential, so I wonder if this doesn’t account for some small additional percentage.