# What a Baseball Jam Is and Is Not

Much of the time, a jam isn’t confusing. What counts as a jam, and what doesn’t, tends to be obvious. The bases are loaded with nobody out in the ninth inning of a tie game? That’s a jam. There are two outs and nobody on in the fourth inning of a blowout? Not very jammy. A jam is a gut thing, and gut things don’t come with explicit rules, but you often know a jam when you see one.

Last week, I asked you, the FanGraphs community, to define what a jam is. Not exactly that, I guess — more like, I asked you to help come up with a jam definition. I presented you with a dozen different situations, and then thousands of you voted on whether the situation counted as a jam, in your own book. I didn’t know what the results might yield, but I figured it would help us in the in-between. Between the obvious jams and non-jams, there are iffy jams. I wanted to try to identify a cutoff.

Let me acknowledge, again, that jams are gut feelings. They’re situation-dependent in more ways than I could include in a poll, and there are presumably elements of momentum and opposition quality that matter to some extent. This is all basically for fun, and for exploration, and nothing is conclusive. We haven’t arrived at a set definition. But we can at least see where the crowd stands. What’s a jam? What isn’t a jam? I have a better idea now than I used to.

Let’s get right into it. Each of the dozen polls came down to being a simple yes-or-no question, so based on that, I just took Yes% to be Jam%. I compared Jam% against three different but somewhat related measures: the probability of at least one run being scored, the average run expectancy, and the leverage index. To begin with, here’s Jam% against run probability:

There is, of course, a relationship, but I don’t think it’s very strong. For example, the fourth-highest Jam% comes with the fourth-lowest run probability. And then there’s the point over there in the lower right, with the highest run probability but the fourth-lowest Jam%. There’s more to it than this, so let’s move on to run expectancy, or the average number of runs expected to be scored in the inning given the situation presented:

That’s better. That’s pretty good! I guess, even though “good” in this case is arbitrary. I just realized I haven’t told you yet that run probability and expected runs are based on the 2016 regular season. That shouldn’t change very much. Anyway, this looks better than the first plot, in that it looks less random, but there’s still a good amount of scatter. There’s one point at 78% Jam% and a run expectancy of 0.70. Left to right, the first nine points form a horn. If you wanted to set a jam cutoff using just this data, you’d put it around a run expectancy of 1.2 or 1.3, assuming you’re satisfied with a community Jam% north of 50%. It’s still clearly not perfect.

The plot using leverage index is just silly. I still need to include it:

Technically, if anything, the voting yielded a negative relationship. That doesn’t make a whole lot of sense, and the better interpretation is just that jams are not really leverage-dependent. That still doesn’t feel exactly right, since we know jams occur in tight situations, but leverage also captures other elements, elements that have little to do with determining jam situations.

What I have for you now is a big table of information. This shows Jam%, and all the different measures, and this also provides reminders of what the situations presented actually were. I know that this is probably a lot to take in, but it’s necessary that the numbers all be included. This will also help the community in further analysis, since I know that I’m not getting to everything. I’ve given this all a first glance, but some of you might decide to go deeper.

The Big Table of Jams and Non-Jams
Situation Bases Outs Inning Margin Jam% Run% Exp. Runs LI
1 123 0 1, top 0 91% 83% 2.27 2.0
2 –3 0 4, top 0 30% 83% 1.32 1.2
3 -23 0 4, top 0 95% 82% 1.90 1.6
4 123 2 6, top 0 78% 30% 0.70 4.1
5 123 2 6, top +3 60% 30% 0.70 3.2
6 12- 0 6, bottom -1 51% 61% 1.46 3.0
7 1-3 1 6, bottom -1 65% 64% 1.18 3.0
8 -2- 1 7, bottom -2 3% 39% 0.67 2.1
9 1-3 2 7, bottom -2 33% 26% 0.48 3.0
10 123 0 8, top +4 87% 83% 2.27 3.0
11 0 9, bottom 0 3% 28% 0.51 2.3
12 1– 0 9, bottom -1 5% 41% 0.87 5.4

Here, you can see the problem with just trying to lean on leverage index. By far the highest leverage in the table occurs in situation 12, with a runner on first and nobody out in the bottom of the ninth, with the home team down one. Yet only 5% of voters felt like that was a jam, which is only two points higher than situation 11, showing the start of the bottom of the ninth of a tie game. That, also, is a high-leverage situation, in that anything late and close is a high-leverage situation, but jams seem to require baserunners. Especially baserunners in scoring position. It seems that all jams are scary, but not all scary situations are jams. One needn’t pass through a jam to lose a game in the ninth.

It’s funny — the strongest linear relationship is between Jam% and the raw number of runners in scoring position. It’s basically the same as the relationship between Jam% and the raw number of runners, period. In that sense, identifying a jam is pretty easy. Every jam here involves at least two baserunners. There is one situation with two baserunners that isn’t a jam, in the table — situation 9, with runners on the corners and two outs in the bottom of the seventh, with the home team down two. That still generated 33% pro-jam support. Compare it to situation 8, which got just 3% support. That one had the higher run probability, and also the higher run expectancy. But the number of runners seemed to sway the voters.

What I like about situation 1 is that it shows a jam can occur extremely early in a game. Not all that much needs to hang in the balance. Yet situation 3 does show that timing matters a little bit. Compared to situation 1, it’s missing the runner on first. But it takes place three innings later, which moved a small handful of voters.

Also, you can compare situation 3 and situation 2. One of those doesn’t look like a jam, while the other appears to be an obvious jam. The difference: a runner on second, to go along with the runner on third. The odds of a run scoring are basically the same, but the runner on second boosts the chance of a crooked number. That was enough to move two-thirds of everyone.

Sticking with the comparisons, the only difference between situations 4 and 5 is that 4 is a tie game, while 5 comes with a three-run margin. The larger margin turned the situation into 18% less of a jam, indicating that the score does play a role. Jams are a function of several variables, and although margin isn’t a primary determinant, it doesn’t look like you can have a jam when the game starts to feel more in hand.

I don’t need to keep going. I’m not going to arrive at a perfect, all-inclusive definition. Nevertheless, this exercise has been illuminating, and I appreciate all your participation. One new thought that has occurred to me is that it could be possible, based on this, for a pitcher to work into a jam, and then work out of it, even without getting out of the inning. A jam situation with one down might no longer look like a jam situation with two down, and yet still there are not three down, which muddies things. The pitcher hasn’t technically escaped the jam until the inning is over, meaning the jam has kind of a lingering effect.

I’ll leave you with a point right by the cutoff. Where’s the line between a jam and a non-jam? According to 51% of you, situation 6 qualified as jam-worthy. That has the home team trailing by one in the bottom of the sixth, with nobody out and runners on first and second. Situation 7 got 65% jam support, even though it’s just situation 6 with an out that moves the lead runner to third. These things are complicated.

Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.

Member
Member

Should have consulted Councilman Jamm.