Better Casting Through Science (?)

Casting is a very important part of running one-shot pregenerated LARPs.  Making the wrong decision about which player should play which character can result in unhappy players and make the game worse.  On the other hand, if you can find that perfect player for a role, they can bring up the level of the entire LARP.

If you ask five LARP GMs for the best way to cast games, you're likely to get ten answers.  Some notable ones include:

• Just give brief descriptions of all the characters and let players pick their top five.
• Casting is an art, not a science.  Ask open-ended questions and go with your gut.
• Everything players write is significant, including margin doodles, handwriting, and when they got the questionnaire response back to you.
• Interview them on the phone or in person.
• We've developed a complex Excel spreadsheet that tells us how to cast people.

At Alleged Entertainment, we've developed a complex Excel spreadsheet that tells us how to cast people.  What follows is a trip into the dark recesses of that spreadsheet.

First, A Little Philosophy

In order to explain why we cast the way we do, it's helpful to understand the assumptions that led us here.  Each of these points could probably be an essay on its own.  For now, I will simply lay them out with a little bit of explanation.

Dislikes are more important than likes.  In other words, if you're stuck in a position where you're forced to choose between giving a player a character that they're 100 percent okay with but aren't excited to play, or giving them a character that has aspects they'll really like but other aspects they'll really hate, it's better to choose option A.

Impartiality matters.  I always hate it when I see GMs playing favorites and giving awesome roles to their friends just because they're their friends.  It's unfair, especially to newbies.

Players should control their casting destiny.  If you've played scheming villains in the last five games, and been an awesome scheming villain, that doesn't mean you don't want to try something new this time.  We give players a new casting questionnaire every game for a reason.

Good casting is more important than keeping secrets.  Yes, asking players whether they're okay with playing a murderer could reveal that the murderer is one of the characters.  If you mean it to be a big surprise, asking about it on the questionnaire could be a spoiler.  But casting someone as the murderer when they're not actually okay with that would be way worse.

Always give the nuclear option.  Maybe you love almost everything about this character.  They've got a tormented soul, a heart of gold, the chance for redemption, political intrigue, and powerful magic.  This is the character you've always wanted... except that they're a stranger in town, and that will ruin the game for you.  We're not here to judge.  If there's something that will absolutely kill your experience, we should allow it to trump everything else on your questionnaire.

The Infamous -3 to 3 Scale

If there is one feature for which our casting questionnaires have come to be known, it is surely the -3 to 3 scale.  We use this scale for almost every question.  It looks like this:

Every page of the questionnaire reiterates the meaning of each of the options, as follows:

• -3 means "I will not enjoy the game if I have this"
• -2 means "I strongly prefer not to have this"
• -1 means "I prefer not to have this"
• 0 means "I'm okay either way about this" (and will never exclude you from playing such a character)
• 1 means "I prefer to have this"
• 2 means "I strongly prefer to have this"
• 3 means "I will not enjoy the game unless I have this"

If one were to remove -3 and 3 from the scale, it would be roughly equivalent to the (very common) 1 to 5 scale in use on many other LARP casting questionnaires.  Thus, our scale differs from the usual ones in three ways:

We provide the nuclear options, namely, -3 and 3.  We will disregard everything else in your questionnaire rather than violate these constraints.

We use negative numbers to signify dislikes and positive numbers to signify likes.  Mathematically, this makes no difference whatsoever, but it does help clarify the scale for users.

We provide rigid definitions of what each number actually means, and reiterate them often.  This helps avoid the situation where some people answer near the middle for everything, while others are all over the place, and it isn't clear whether they really mean that or not.

Constraints

Let's take an entirely hypothetical character named Luke.  He's young and a bit naive, having grown up on a farm.  However, unbeknownst to him, Luke has a great destiny to fulfill.  He's also been lied to his entire life about the identity of his parents.

If we're trying to cast a player to play Luke in our LARP, we want to make sure that that player isn't averse to any of these traits:

• young characters
• naive characters
• characters who don't start the game with a lot of information
• characters who are highly involved in pushing forward the story
• characters who have surprises coming to them

In addition, the following traits about Luke may appeal to some players, but are not strictly necessary for playing him:

• he's good in a fight, if the player decides to get involved in those
• he can train to unlock supernatural powers

These could easily be translated into a grid of -3 to 3 scale questions.  For example, have players rate these potential character traits using the scale:

• 18-25 years old
• naive
• ignorant
• plot-central
• likely to be surprised during the game
• involved in combat
• in training
• with supernatural powers

(Some of these could probably be rephrased better by separating them out into semantic sections, such as "traits" vs. "activities.")

You would then, if possible, only cast people as Luke who answered 0 or more to all of the first five questions.  In addition, you would take the last three into account when choosing from amongst that pool of people.

This distinction between "qualifying" questions (i.e., ones that decide which of the players is eligible for a particular casting) and "scoring" questions (i.e., ones that come into play when deciding between qualified players) is central to our casting method.

Writing a Useful Questionnaire

Knowing that we want to use the answers to the casting questionnaire this way informs the way we create the set of questions.  One naive, but effective, method for doing this is to simply go through the entire set of characters by hand, writing down every distinguishing character trait they have and everything they're likely to do in the game, and turn each of those into a question.

Of course, you'll probably end up with around 150 questions.  Better hope your players are really dedicated, because if not, nobody is going to answer the whole thing.

The first line of defense against this problem is to think about combining similar questions.  For example, you could easily combine "playing an asshole" with "playing a jerk."  Usually it's not so clear, though.  One way to figure out whether to combine two questions is to ask whether someone might conceivably answer positively to one and negatively to the other.  If you can't think of a situation in which they wouldn't, you can probably combine the questions.

It may also be possible to simply avoid asking some of them, particularly if they're not as important as others.  Cutting questions from the questionnaire is a difficult and often contentious process, and usually just comes down to a judgment call.

In addition to the numeric questions, it's important to give space for freeform answers as well.  Sometimes this simply takes the form of a "is there anything else you want us to know?" essay question at the end.  We've also tried asking more game-specific ones, and also leaving blank spaces at the end of each page of numeric questions for specific thoughts on the questions on that page.

Finally, it's usually not a good idea to add questions to the questionnaire just because most LARPs ask them.  If you can't think of how you would use it in casting, why ask it?

Crunching the Numbers

We now have a scale with which to ask questions and interpret their answers with a good deal of precision.  We also have a set of questionnaire responses that we know how to apply to our set of characters.  Now it's time to cast.

The method is fairly straightforward.  For each character in the game:

1. Start with the entire set of players.
2. Remove any who answered "3" to a question that this character doesn't match.
3. Remove any who answered "-3" to a question that this character does match.
4. From the remaining players, figure out who is qualified for this character.  Start by removing any who answered a negative number on any qualifying question.  
1. If this leaves you with an empty set of players, try again, only removing players who answer -2 to any qualifying question.  
2. If you still can't find a qualified player, consider all the remaining players your qualified set. And consider changing the character in the next run.
5. Score each of the qualified players using both the qualifying and the scoring questions for the character.  Just sum up their answers for these questions to determine the player's score for the character.

Once you've done this, you will be left with separate scored lists of players for each character in the game.  At this point, casting becomes a constraint-solving puzzle.  There are many ways to solve these, but here's how we often do it:

1. Start by looking for players who only match a single character, or characters who only have a single qualified player.  Cast those.
2. Every time you cast a player, remove them from the scored lists for all the other characters.
3. If you see a player who's significantly higher-scoring for a particular character than any other player, this might be a good indication to cast them there.
4. Oftentimes, we wind up with a "middle group" of player/character matches at the end that are roughly equivalent.  Using freeform answers can sometimes give you inspiration on how to cast this group.  Also, if you know some of the players, this may be a good time to bring in that knowledge to inform how you think they would do with the characters.

Before making any casting decision of any kind, make sure to re-read their entire questionnaire, including any essay questions they wrote.  I really can't stress this point enough.  Do not blindly rely on the casting algorithm.  It's there to help, but it is not omniscient.  If a casting choice goes against something the player wrote under "anything else you want us to know," or it just feels wrong, don't do it!

After you've got a cast, do one more sanity check.  Read through the questionnaires for each player again, out loud, and think about them in light of the character you gave them.  Seriously.

Casting Matters

On that point, at least, I think most LARP writers can agree.  And if something worth doing, it's worth thinking about.

This numbers-driven way of casting is not for everybody, or every LARP.  I personally don't use it for every game I write.  And like every system, it has its strengths and weaknesses, and being aware of these is key to successfully applying it.

Regardless of how you choose to cast LARPs, I hope that reading this has inspired you to consider how best to cast.  Thanks for reading!