Funnyball: Can Improvisors Benefit From Using Data and Analytics?

asssscat-ucb-improv
First of all, I just want to say right up top that I know this is all very uncool. Improv is supposed to be fun, like all entertainment is, and it is super lame to take it too seriously like I am doing here. I admit it and I’m very sorry. Thank you for accepting my apology and agreeing that this is a good article now.

Okay, now that that’s out of the way, let’s talk about the potential to use analytics to make improv better. If you’re reading an article about improv online I trust that you’re enough of a nerd that you have some idea of what words like analytics, sabermetrics, or “data-driven” mean. In fact, in a weird coincidence, the first time I ever heard of Moneyball was when my friend and former teammate Brandon Scott Jones was reading it at a Harold team practice in like 2010.

Since then, the general idea of that book has soaked into a number of different industries. Nate Silver is the poster boy for analyzing politics from a data-driven perspective. Michael Lewis followed up 2003’s Moneyball with The Big Short, which applies roughly the same concepts to the housing market. The St. Louis Cardinals have enjoyed a significant amount of success using their own brand of Moneyball, which warranted it’s own book, 2016’s The Cardinals Way. And of course the Golden State Warriors are not only the most dominant team in basketball history, they were also named “Best Analytics Organization” at a recent conference at MIT. Which is all to say that there must be something to all this analytics stuff.

So then why not try to use it with improv? Is improv already perfect? Is it too good already? No. In fact, it’s often bad. And it’s for that reason that I realized I must plow ahead with this idea. Because if it can help even one bad improv show become good, then it will all have been worth it.

Frankly, I’m optimistic. After all, I’ve had to explain what improv is to tons of people over the years, and what’s already the most accurate frame of reference for improv? Basketball. An improv team practices between shows? Why? Isn’t it improvised? Well yes, but improvisers still need to get familiar with each other’s playing styles, they need to practice certain “plays” or formats together, they need to develop cohesion and a group mind. It’s all very similar. Even ASSSCAT 3000 is often described as “like if a bunch of NBA players got together and played a pick-up game.”

The big hump to get over seems to be that it’s just lame to apply these same principles to art. With sports, there’s a pretty clear rubric for success: points. Billy Beane could have probably recommended that all the As players wear Richard Nixon masks and if they started winning they’d be like, “Hm good thinking with the Nixon masks. Here’s a raise.” Art? Not so much.

Except…kind of. I would suggest that in the past few years, even if art (and comedy especially) remain firmly and inarguably subjective, a little bit of analytics has been sneaking in. The Bechdel Test, the litmus test that looks for whether a movie includes a scene of two female characters talking about something other than a male character, is an analytic. I’d put it on the level of a statistic like “defensive touchdown” or something. Like, can the defense get points on its own? That doesn’t have any necessary relationship to the outcome of the game, but surely it measures…something.

Further, criticizing or lauding (or even, in the case of the Hayes Code, directly influencing) the “moral” of a movie has to be considered an analytic. Or even putting more focus on the optical value of, say, increasing the number of female showrunners or avoiding an all-white Oscars, is evidence of ways analytics can help even in fields where a scoreboard isn’t the last word.

So how would this work? I emailed Los Angeles-based improv director and UCB institution Will Hines to find out.

First I had to clarify, for my own edification as much as this article, to what extent data is kept on individual improvisers as they go through the UCB system. “There are no statistics kept,” says Will. “Teachers give grades like ‘repeat level,’ a ‘pass,’ or a ‘superior,’ that’s it.”

Pretty much what I expected. But it turns out Will has thought about this as well, and listed some of the “statistics” that could be kept:

initiations
Tag-outs
Walk-ons
Offers made (New piece of information)
Offers accepted
Unusual things introduced or framed
Patterns hit (Doing something that has already been done at least once)
Emotional reactions
Specificity
Justifications/explanation
Heightening moves

Those are a pretty good baseline for gathering data on a given improv team or improviser, right?

To use the basketball parallel, imagine if there were no stats kept in the NBA. Now you start keeping track of shooting percentage — in my mind, that’s analogous to keeping track of, say, percentage of offers made to offers accepted. Analogous in the sense that it has a bearing on the overall quality of the show (without that bearing being directly tied to the objectivity of points). All else being equal, it’s preferable for a player’s field goal percentage to be higher, but it’s conceivable that a player shoots terribly from the field but adds value in other ways. And ultimately it’s always better to know if you’re terrible at shooting so then you can rebound more, a la Dennis Rodman. Or I think we all know an improviser who can sit on the back wall for an entire Harold and then add something in the second group game that takes the show to a whole new level.

But even with this basic hypothetical statistic there’s already some weirdness. As Will pointed out in a follow-up email, when you start down this road you quickly see that the basketball template is way better than the baseball one for the simple reason that these “offers made” and “offers accepted” numbers already involve multiple team members. For example, Improviser A’s “offer percentage” (“agreement ratio”??), as opposed to any stat in basketball, already depends on Improviser A’s teammates. So it seems that you could go a step further and say that improv stats are an even more team-oriented than even basketball. Or not. I don’t think it’s a dealbreaker either way, but worth mentioning.

No, when you get right down to how analytics could be helpful in improv, I think you have to get at something a little more general. When I used to coach improv more, I found myself going back to one sentiment over and over: do this or that because it just makes it easier on yourself. You want to develop full premise ideas in an opening because then you can hit the ground running in scenes without having to find something funny from scratch. You want to start scenes with clear, premise-based initiations so your scene partner isn’t sitting there saying “what the hell is s/he going for?” And just in general, you’re told to “play to the top of your intelligence” and “don’t be coy” because as a rule it’s more entertaining to watch people make choices and know what they’re doing. Now to be sure, there’s nothing stopping a basketball team from winning the NBA Finals if they’re all shooting 15% from the field. But it’s way easier not to.

So to that end, Will and I agree that our best bet is some abstract statistic that measures “agreement” or “on the same page-ness.” In his words, “things that tend to make shows better, rather than trying to measure how good a show is.”

He went on:

I like the idea of measuring “offers accepted.” It feels akin to on-base percentage in baseball to me. It doesn’t give you runs in and of itself, but it tilts so many factors in favor of the offense that it’s a good barometer of how successful your team is going to be. If you have an improv team that can reflexively confirm facts and things said; I think that’s a good barometer of success. Another one is framing unusual things. 

Repeating, calling attention to or demonstrating unusual things. If your team is sharp enough to notice what is unusual very reliably then that feels like a good barometer. 

Things like who initiates, who walks on, who edits: I don’t think that has any real bearing on how good the show is.

I like the concept of “offers accepted” too. But I think it could be a little more general. A big sabermetric stat in baseball (and theoretically any sport) is “WAR” or Wins Above Replacement, which purports to measure “a player’s total contributions to his team” by comparing his (unfortunately women’s sports are behind in analytics, but they’re catching up) contributions to the team against those of a hypothetical “replacement-level player.” So how much value an individual player adds to a team that the team couldn’t have gotten from a hypothetical market-average player.

Could you do that with improv? As a baseline, keep track of “offers accepted” over “offers made” for each individual improviser. Then factor in “initiations made” and “initiations accepted” (perhaps weighted because they’re initiations?), “clarifying support moves,” “audibles” where an improviser clearly had an idea but dropped it to support a scene partner’s idea, “rescue moves” where an improviser made a big choice to shift focus from a struggling teammate. I would also want to add some measurement like “Time Playing Real,” which I would define as “total time on stage acting like a normal person without doing anything ‘funny.’” And then it could all come together as like “Agreement Quotient” or something.

This of course brings up one concern I always have when people talk analytics in sports. Doesn’t it seem incredibly subjective? Even basketball stats like “shots off a drive,” when you get right down to it, are largely judgment calls. How refined of an improv fan would you have to be to keep track of something like “Improviser B understood that Improviser A’s initiation was from the pattern game”? But again, the sports crowd seems to have figured out a way to make it work.

Frankly, my biggest worry is that there’s something inherently contradictory about keeping individual stats in improv, which is all about teamwork and group mind. It could be absolutely terrible for team chemistry (in addition to ushering in a nightmare scenario where people sit at McManus after Harold Night discussing improv stats). Would the potential for small improvements be worth it, culturally speaking?

I don’t know. But at the end of the day, I think we’re safe. Because as Will pointed out, there’s really no way to keep stats while you’re teaching a class. You would have to get someone to do it, and then keep track over time, possibly during shows, and then set aside time to implement that data into your practice schedule. That’s a lot of work for something most people take up because it requires no preparation.

In any case, you may remember last winter there was a bit of a dust-up where a few NBA coaches and former players claimed that all this math stuff was taking away from the real game, which is best played according to gut and instinct. Charles Barkley even went on TV and argued that analytics were just an easy way for boring nerds to get in on this game that cool jocks invented.

With improv, of course, it would be the exact opposite.

Photo by Mindy Tucker.

From Our Partners