(*) MASTER NOTES: Kelvin Degrees—The Save-Situation Effect

I’ve said it before, and I’ll say it again: I love the BaseballHQ subscriber forums. In addition to providing plenty of smart, informed discussion about everything from the Commissioner’s concern about misuse of the 10-day DL to fantasy trade ideas (lots of them) to alerts about new episodes of BaseballHQ Radio to players for the rest of the season to ... well, you get the idea. The Forums are a fantastic resource. (And there’s no flaming or hostility, just a ton of mutual respect, making our Forums quite a few notches above the usual forums out there in the interwebs.)

The other day I happened across a thread about Kelvin Herrera’s odds of being traded. This interests me because I hope he is. I have Joakim Soria, and some other KC proto-closers are still free agents in my league.

As often happens, though, the thread veered out of its lane into a discussion about Herrera’s perceived poor performance in non-save situations.

One poster had noted that the Royals “need to be careful pitching Herrera in non-save situations.” Apparently he had given up eight runs in his previous six non-save situations (six of those runs in just two of the games, as we shall see). He also said, “I've heard more than once about closers that seem to excel when in a save situation but for whatever reason don't perform as well in non save situations.... Maybe this is just urban myth and the facts don't back it up.”

Another poster then suggested that “Herrera is what the old timers would call a ‘gamer.’ He's at his best when the pressure is on and the game is on the line. Maybe it's psychological.”

And yet another poster opined: “What (first poster) suggests seems to happen enough that folks think there's something to it.”

Of course, what people (and old-timers specifically) believe doesn’t prove anything. A surprising number of people think the earth is flat.

Fortunately, in many instances, especially in baseball, we can use stats to find get a clearer picture.

The availability heuristic

But before we get to the numbers, let me quickly discuss why anecdotal evidence has such appeal. It has to do with the “availability heuristic.” (A “heuristic” is a mental shortcut or model humans use to make sense of a situation that might otherwise take a lot of time and effort to actually solve.)

Regular Master Notes readers might recall that I’ve discussed work by the renowned team of Amos Tversky and Daniel Kahneman, who were pioneers in the area of behavioral economics. In their research, among many other things, they learned that ordinary people often assess the probability of an event by how easily they can think of examples of that event—even when basic arithmetic or plain common sense says otherwise.

As cavepeople, this was a useful model. Nowadays, though, we have many mechanisms to deliver examples. And certain kinds of events stay with us, and are therefore “available” to us, for all kinds of reasons. Most of those reasons don’t reflect actual probability, and in fact the events are often wildly unlikely. For example, heavy media coverage of crime and plane crashes (“if it bleeds, it leads”) causes such events to be more “available” to people, who therefore greatly overestimate how frequently crime and plane crashes occur.

Paradoxically, it’s the infrequency of these events—it’s the relative rarity that cause them to be memorable—that causes us to believe it to be frequent.

And “recency bias” is a form of the availability heuristic—a recent event is more “available” than events that happened earlier.

Now think how availability might affect perception of how often closers pitch poorly in non-save situations. A lousy outing by a top closer is rarer and therefore more extensively mentioned and more memorable than his many clean ninths.

We have an example in the thread with Herrera’s string of allowing runs in non-save situations. In fact, the issue was more focused: Herrera had a 2-ER flop on June 5 and a 4-ER debacle on June 8, both in a series at home to the powerful Astros.

The guy, quite possibly a Herrera owner, recalls these two blowups in non-save situations. They are “available” to him, so he assumes the examples support (or, worse, prove) the assertion.

The Facts

Fortunately, in this case, it’s easy to check the facts. And for Kelvin Herrera, at least, the numbers suggest he does “turn it up” when saves are on the line.

Because Herrera took over as closer after last year’s break, I went to his pages at baseball-reference.com and grabbed the game-log data for 2016 post-break through the 2017 season up to the break. Sixty-seven appearances, if you’re keeping score at home.

From there, it was pretty simple to calculate his stats for performances in save-situations and not. I ignored outcomes like ERA and WHIP because of the variance caused by sample size, especially once the already small game log is further subdivided into various contexts. I also chose not to look at Leverage outcomes for the same reason.

Instead, I focused on per-batter results, which start with the overall 273 batters faced (BF) and subdivide into cohorts that are still small, but at least mostly have 100 BF or more.

To set Herrera’s overall level, during the period in question, he had a 25% K%, 4% walks, 31% GB, 19% LD, 21% FB, with a 4% HR (all of BF, remember) and a 5.6 Cmd (K/bb).

The first indication that he amps it up with a Save on the line is in Cmd: 11.3 K/bb in save situations, 3.7 K/bb in non-save. He had a six-point edge in K rate (28%-22% ) in save situations, and a three-point edge in walks (3%-6%). He also gave up two more points of HR in non-save situations (3%-5%).

But we also know that not all saves are equally difficult. The whole movement towards intelligent bullpen management depends on ignoring saves to put the best relievers in for the most important game situations.

So I also parsed Herrera’s performance by whether he was protecting leads of one run leads (“critical”) or more (“non-critical”), and it turned out he was more effective when the heat was on.

Interestingly, someone in the thread also said he suspected that reliever performance “varies between high pressure and low pressure situations, which isn't the same as save versus nonsave situations.” He went on to say it’s hard to believe that Herrera “can be expected to perform better ... in save situations(,) but worse ... in a high pressure situation that falls outside of the save opportunity definition.”

Hard to believe, but true. At least for Herrera over this relatively short run.

Herrera’s K and walk rates in critical save situations (one-run lead) was markedly better than in critical non-save situations (one-run deficits or ties). With a save on the line, Herrera's K rate was ten points higher (34% to 24%). His walk rate was four points lower (3%-7%). And his Cmd rate was more than nine points better—12.0 K/bb in critical save games, 3.6 K/bb in critical non-save games. In fact, criticality in general played a distant second fiddle to save opportunity, as he also pitched far better in non-critical save situations than in non-critical non-save situations.

Keep in mind that the samples are still small. I plan to expand this study to all the top closers in baseball over multiple years. I’ll put it on my docket for this off-season for an HQ Research article. In the meantime, remember the availability heuristic and try not to depend too much on it.

And what about those anecdotal examples earlier? Well, for me, they’re just that—anecdotal. But here’s something curious: including those two blowups, Herrera has allowed 2+ ER in nine different appearances since becoming the KC closer. Eight of those nine—and all four of the games with 3+ runs allowed—were in non-save situations. I’m just sayin’.

More From Archives