Wouldn’t You Prefer a Nice Game of Chess?
By Anna Rettberg
Casey Muratori
Have you ever read a Choose Your Own Adventure book? You know the kind of book I mean  —  where you get to the bottom of a page and it says something like, “If you decide to bargain with the man for an extra pound of frozen peas, turn to page 47. If you decide to try using the trebuchet to get past the mutant fishpeople, turn to page 131.” You, the reader, carefully consider these options  —  clearly the best two options for the situation, and the only two worth considering  —  then you turn to the page indicated for your choice.
Or, if you’re one of those people who cheats at literally everything, including your entertainment, you read both pages and then back-decide which choice to make. Far be it from me to accuse you directly, but you know who you are. You probably have a bunch of GameFAQs tabs open right now.
Anyway, if you’ve ever read a book like this, then you already know how modern “interactive” story games work under the hood. The methods used in games today are only slightly more advanced than the methods authors were using in “interactive” books before personal computers even existed.
While small variations are easier, major variations are still built as manually-written branches.
There are some widespread improvements that have been made thanks to computers. Most games that feature “interactive” narrative now have easy ways for authors to insert or remove parts of their text based on the state of the game, making small variations easier. But on the whole, the major method of construction has not changed: authors still have to write out all the major plotlines of the game, in their entirety, by hand.
This is not a huge problem if the goal is to just make lightly interactive fiction, where the player gets a little bit of choice but basically plays the same story as everyone else who plays the game. But it’s a huge problem for those of us who want actual volitional fiction. And it’s not just a practical problem, it’s a theoretical problem.
How about Global Thermonuclear War?
As with most theoretical problems involving interactive narrative, the fundamentals were thoroughly explained in the exemplary 1980s documentary War Games. I would call your attention to the following scene, in which the computer is instructed to play itself at Tic Tac Toe by setting the number of players to zero:
Now, obviously, since it was filmed in the 80s, there are some differences in how this works on today’s more sophisticated computers. For example, on PCs you cannot always set the number of players to zero and have the computer play itself like you can on the WOPR. But everything else in the scene still applies.
What the computer was doing in this scene  —  if you didn’t see the movie and thus don’t know the context  —  was learning about how branching narratives are impossible to write by hand. One might also argue it was learning how good linear narratives are also impossible for certain screenwriters to write by hand, and that possibly the WOPR should fire its agent, but that was sort of a larger realization that it came to later, after the day’s shoot, when it was drinking away its troubles in its trailer.
But I digress.
The point is that even a game of Tic Tac Toe, as simple as it may seem, turns out to have over two-hundred thousand different possible playthroughs. Over two-hundred thousand. This is a game that has only 9 possible moves on the first turn, 8 on the second, 7 on the third, etc.  —  hardly anything anyone would consider a good game, let alone a great game. And it’s a game most people wouldn’t even bother playing as adults because the space of possible games is so small that even a completely untrained and unstudied human quickly gets good enough that they can never be beaten, only tied.
Things only get worse in terms of game complexity as you move on from Tic Tac Toe to games that people actually think are worth playing, like Chess. Chess has so many possible playthroughs that we don’t even know how many there are.
That’s not me being funny. We literally just don’t know. Type “Shannon Number” into Google if you don’t believe me.
So now project this back onto the authoring process for a narrative game where we want the player to feel like they have real control and are playing a game with real possibilities. In other words, something more engaging than Tic Tac Toe. Maybe it doesn’t have to be Chess, but it’s gotta be somewhere in between.
Hopefully you can see that solving this problem directly is intractable  —  writing substantially more than two-hundred thousand different stories isn’t something you can expect any game designer or author to be able to do. And if you wanted the interactivity level to be on par with chess, well then they’d have to write a minimum of ten duodecillion stories  —  that’s a one with 40 zeroes after it.
Thus we come to the other important thing War Games teaches us about manually-authored branching narratives:
The only winning move is not to play.
Writing volitional fiction by hand isn’t possible. It’s just that simple. There’s literally no point in trying.
That’s why last week I said that the reason nobody has been able to ship a volitional fiction title yet is because we simply don’t have the technology. Whatever game you’re playing right now that’s got some narrative elements to it already is  —  to a first approximation  —  as volitional as game fiction is going to get until there are some technological breakthroughs.
Yes, there are plenty of small things games have done to help reduce the writing workload, and yes, there are probably some more of those to be discovered. But small improvements like these do not even come close to providing the kind of amplification necessary to cross from the sub-Tic-Tac-Toe scale to the Chess scale of interactivity.
And at the risk of starting yet another major grumpy grumblefest thread on Hacker News or Reddit or wherever else, I should also mention that academia has produced no practical solutions to these problems, so it’s not like games have been ignoring valuable existing research. No, that’s not me being unaware of your favorite story generation paper  —  that’s me being aware of it, and saying it fails to solve the problem. You’re welcome to disagree  —  but of course, if you want to, you’re going to have to go make a great volitional fiction game based on said research and prove me wrong.
So what do we do now?
Well, unsurprisingly, a problem of this magnitude that has resisted easy solutions for so long is not going to be solved overnight. If we want volitional fiction to become a reality, somebody needs to establish a long-term initiative to improve the state of the technology.
A little over a year ago at Molly Rocket, we decided to be just such a somebody, and we’ve been working hard since then to set up a long-term initiative that we believe can make real progress. Just what that will look like over the next few years, if we’re lucky, will be the subject for next week’s post.
Until then, thanks for reading, and we’ll see you on this great series of tubes we call the internet!