Antony & Cleopatra: The Murder of Marlon Brando Post-Mortem

Right, so, bread’s proofing and I’ve got some time before the Thanksgiving feast kicks off, so here’s a post-mortem.

Antony & Cleopatra: The Murder of Marlon Brando wasn’t the game I wanted to enter. I realized a few weeks before the comp that the game I wanted to enter was never going to get finished, at least not in the time that I had, so I threw this together instead. I think it was, uh, extremely mid, verging on bad, but fortunately for me it seems the judges disagreed, putting it at the two-thirds mark (24th out of 74). It also has the lowest standard deviation of the comp (.92), so I’m forced to bow to the wisdom of the crowds and admit that, well, maybe it’s fine, I guess.

Why might I think it’s terrible? Well, I wrote it in a little under three weeks, starting on September 12, 2023 and ending on September 30, 2023. Twine tells me that it has 44,346 words, about ~2,300 words a day, a writing pace that for me can be described only as frenetic. It’s actually somewhat worse than that, because while I had a prototype to go off of (more on that later) I had to hack together all of the menus and UI and some (not all) of the multiplayer networking code. As a result vast amounts of the game are literally first drafts, thrown together in a mad rush, with not a single revision. My first two playtesters (four, really, since it’s intended to be played in pairs) came back with a gigantic list of typos, and I’m sure that many more remain.

Still, I don’t think it’s all bad. The prose might be, you know, mediocre, but I think some stuff worked well, too! But before we talk about the final product, I’d like to talk about the prototypes.

Oh, and I’m going to assume you’ve played the game, or don’t care about spoilers. So, you know, be aware.

The prototypes, and some multiplayer IF theory

Way back in November of 2021, I’d just played Alexisgrad, and I was sitting around thinking “Huh. You can do that?” I figured, hey, why not try? I wrote 4k words (about 3 minutes’ worth of gameplay) of a cyberpunk thriller in ChoiceScript, with explicit instructions to the players to tell the other person, out loud, what the options they picked were, and after playtesting it once very quickly rediscovered that it’s actually really weird to be on a call with somebody and be playing through a scenario, but not talking to them. This was an issue with Alexisgrad, too, where the roles were opposed and if you got on a call there’d be long stretches of silence where you just waited for the other person to read their text and get to the option. Awkward.

Right, I thought, so! I want my players to talk to each other. How do you get the players to talk to each other? Well, the mind obviously jumps to the traditional TTRPG style, but the TTRPG decision space is so much less bounded than you’d have in a traditional IF-style game, and you’re not doing fixed narratives, and often you have a DM or other social mediation mechanisms to smooth everything out and frankly, it’s incomparable with the tight narrative structure I enjoy in IF. The next option is split-information puzzles, where you force the players to talk to each other by giving them a puzzle split in two, like the We Were Here series or Keep Talking and Nobody Explodes or The Past Within, but those are based specifically around puzzles, which, like, isn’t the narrative IF experience I enjoy. There’s, like, Arkham Horror, but that’s full-visibility, fully-cooperative, and while it might be fun to read out the cards in spooky voices it’s actually kind of awful to play, because it shares the Pandemic problem where it often devolves into one player and n bystanders, and it’s, you know, awful. The worst. Pandemic needs to be shot into the sun, but that’s a whole other thing, you can ask me about it if you want to hear me rant. Anyways, what other examples can I think of?

Hey, remember that Sherlock Holmes game?


Long story short, I borrowed my dad’s copy of Sherlock Holmes: Consulting Detective, read through the first few cases to refresh my memory, and figured hey! you know, if I took this and split the locations up between the users, that’d work.

What’s Sherlock Holmes: Consulting Detective?

Sherlock Holmes: Consulting Detective is a book-based detective game. Aside from the book itself, the game comes with a variety of important feelies, such as a big map on London, a bunch of single-day newspaper articles, and a directory of various people and businesses. You read the introduction, which introduces the main conflict, and then, based on the clues in the introduction, you visit various locations. When you visit a location, you read the corresponding passage in the book, which may give you further clues to investigate, or might be a red herring. Eventually, once you think you have a good idea of what the solution to the mystery is, you go to the last page where it gives you a quiz, which you’re probably going to fail, because these are tuned to be pretty difficult, and then in the conclusion Sherlock Holmes comes in and explains to you who did it and why and how you were expected to come to that conclusion. Also it grades you on how many locations you visited, which seems really harsh, since you’re not likely to get the answers even if you visit all the locations, but I guess Sherlock Holmes fans enjoy that sort of thing.

I figured, ok, great. As, like, a game design technique, this is actually perfect for getting people to talk to each other, right? You have a “puzzle” (the mystery), and you have “keys” (the clues from the various passages) and so splitting them up between the players would necessitate discussion. Unlike with We Were Here or The Past Within, the gameplay would be text-centered and not centered around clicking on things, which leads us another step closer to the narrative side. I was eager to see how it went, so I wholesale copied the first scenario in the book, about a murdered munitions magnate, into ChoiceScript. I made three changes, however.

  1. I divided the game into turns, and put a turn limit on it.
  2. I restricted the game so that each player had to visit different locations every turn.
  3. I split the amorphous blob that was the player(s) in SH:CD into two distinct players, Antony & Cleopatra (because I guess I had Shakespeare on the mind or something), and renamed every other character in the scenario.

This renaming wasn’t to hide the origin; after all, this (being a ChoiceScript game) was intended only to test out the design, and was never intended to be released. It was done because I found it hard to remember any of the original names, and I found it way easier to remember the names of people I already knew. The French ambassador became Napoleon and the Russian ambassador became Rasputin. The lady who’s having an affair became Marie de Champagne, because she was the patron to the author of The Knight of the Cart and had opinions on love and marriage (yes I know that’s a silly free association chain). There’s not really any intentional joke to the three company executives being Brando, Dean, and Hepburn, it’s just that I had them on the mind and threw them in. There’s an engineer in the SH:CD scenario, and I was like, who was a famous engineer that I’ll remember? That’s right, Vitruvius Pollio.

Sorry. I don’t know many famous engineers. I’m sure there are many famous engineers out there who aren’t, like, two thousand years dead, it’s just I haven’t been keeping up, you know?

Anyways I tested it and it went hilariously. It got people talking to each other, but it also revealed some structural issues with that specific approach. Here’s the problem with breaking up the locations in SH:CD.

  1. The game requires extremely close reading to pick up on the subtle hints and clues.
  2. It’s really, really hard to pick up on subtle hints and clues when you only have half the text and the person telling you about the other half may or may not have picked up on the appropriate bits of information on their side.

One pair of playtesters, a father and daughter pair, didn’t selectively summarize their passages to each other so much as give their opinions on them, meaning that any of the cues which might lead them to spot contradictions were lost, leaving them totally confused when it came time to take the quiz. Another pair of playtesters, fully aware of their own human fallibility, resorted to simply copy/pasting the entire text of the passages to each other for every passage. I was very impressed with the ingenuity on display, but also, lol, lmao. Most people failed pretty hard, which wasn’t unexpected, since I was adding difficulty onto an already difficult game, but watching them interact was really helpful.

At this point, I had a different multiplayer experiment I wanted to make, but I realized that in order to get there, I’d need a working Twine multiplayer template, so I briefly ported the ChoiceScript proof-of-concept into Twine. This was my second prototype, which was built expressly to get the multiplayer architecture down, which I then used in Ma Tiger’s Terrible Trip, another experimental multiplayer game.

Ma Tiger’s Terrible Trip and the AI Problem

No, not that AI problem. This AI problem is: if you’re playing a choice-based game, how is having another person piloting another character different from having an AI do it, and by extension, from just writing the other character as a NPC? If you’re playing a game of chess, and there’s no chat or anything, how do you know if the other player’s a bot or a person? Well, you don’t, for the most part! So…if you’re playing a game, what does having a player on the other side add?

In some contexts the answer is obvious; if you’re playing, like, League of Legends, the players are there to curse at you in chat when you mess up. (This is a joke; League of Legends is a team-based game, has an extremely large decision space, and as far as I know there just aren’t any bots that are equivalent to players…yet. I don’t actually play League of Legends, maybe I’m wrong on that, but I remember reading that to be true.) More to the point, in many competitive games it’s true that bots just aren’t as good as people yet or that it’s more interesting to beat a person than a computer. So, sure, competitive games with large decision spaces, great.

Unfortunately, most choice-based narrative games - and even parser-based puzzles games - don’t have large decision spaces. Part of the ability to “feel” the other player can be expressed through the decision space - if you’re playing WoW, Leeroy Jenkins might charge in where other players wouldn’t. If you’re playing an FPS, maybe this player’s tic is switching weapons while spinning in a circle; maybe it’s jumping at the start of every round. Even absent any integrated voice chat, you can feel the unique presence of the other player. If there are six options you can pick and six options the other player can pick, and both of them are “in-character” for your characters, well, what do you want here? Do you want the other player to role-play? If you want them to role-play, the desire becomes equivalent to just having an NPC. Do you want the other player to react to what your choices are? Well…they can’t; they’re too restricted by the author’s ability to write out scenarios.

Anyways, so, that’s why I think that, in order to “feel” the player, you need to provide either an expressive decision space or you need to push the players into interacting in channels outside of the game - that is, voice chat, text chat, whatever. But, given that I don’t know of many examples of authored narrative multiplayer games, I wanted to pick at this some more.

The Game I Wanted To Make

So I wanted to test this out. I was going to make it an investigation game where you and your partner poked around in an abandoned settlement, requiring the players to cooperate out-of-band, but I didn’t finish it. A few weeks before IFComp I gave up. RIP that game, perhaps it will return one day.

The Game I Did Make

See, IFComp’s kinda unique in that, even if your game isn’t very good, people will play it and write things about it and, sometimes, give you really good feedback. So I figured, hey, I wanted to test out ways of making multiplayer narrative games, that didn’t get done, but I also have this old prototype over here, what if I just threw that in? Then I looked at the prototype and went absolutely not, because being a transcription of a SH:CD scenario ported to Twine with a bad UI and a deeply terrible, unfun design, like, yeah, I’d get feedback, and the feedback would just be “this is bad tho.” So I rewrote everything in three weeks, keeping the bones of the scenario with minor tweaks, but completely redoing the prose, the gameplay, the quiz at the end…

Here’s a summary of what changed:

  1. I rewrote the entire text of the game, resulting in a massively expanded experience; the original is 1457 words (as per the wc command), and the expanded version is 44346 words (as per Twine’s summary), a 30x increase.
  2. I added dialogue options for the two players, as the original was static text, as per the gamebook.
  3. I changed the nature of the daily visits, as a result of prototype feedback. In the prototype, players visited different locations. In the new version, both players visited the same location, every time.
  4. I added synchronous elements, and a shared calendar for planning. The idea here is that by having a linked calendar which changed when the other player scheduled visits, that would increase the “feel” that you were directly cooperating.

I guess that’s not really interesting, from a post-mortem perspective, being essentially a changelog, but I’m honestly surprised at how many words ended up in the final product. I intended it to be a quick, mostly throwaway game to rustle up some commentary; I ended up with an extremely ramshackle but respectably-sized game which was, at its core, still just a straight port of the first scenario from Sherlock Holmes: Consulting Detective.

Most of it was, as I said before, literally first draft work, and so I threw it out into the wild, hoping to get feedback on the design and the experience of playing it as a multiplayer game, but not expecting it to be taken seriously. The best-case scenario, for me, was to scare up a bunch of reviews with new ideas on how best to leverage multiplayer in narrative games.

The Reception

Well, I did get some reviews of that nature, so I’m happy! People also seemed to like it in general, some even on the basis of its writing, which…surprised me, given my opinion of its quality, but hey, I’m not the boss of anybody. A lot of people commented on the setting, and how it was underutilized, which is funny because from the author’s side there is no setting, only what seemed vaguely funny to me at 1am when I needed to bang out another clue point with [insert name here].

Uh, I gotta go prep for Thanksgiving, which is unusually good timing, so let’s wrap this up.

Wrapping This Up

Here’s my takeaways:

  1. I need to schedule my time better.
  2. Help, text-based narrative multiplayer games are hard.
  3. Uh, sorry, I really gotta go, I might come back to this later.
  4. Whoa this is 5+ pages in Google Docs, I sure can spit out words when I choose not to revise at all.
  5. Sorry for the extremely meandering nature of this post-mortem.
  6. Thanks for reading, please ask questions below if you’ve got’m!

Thanks for sharing! Wow, this is a fascinating look at the whole process, so far the post-mortem season has been mostly me being impressed at how long some of these ideas were percolating and how different the initial conception was from the end result.

(Re: your beta testers who copy-pasted the text to each other . . . that might have been more efficient than what my partner and I did on the first screen, which was literally reading aloud good chunks of the description to try to determine if we were getting something different . . .)