The Stack

Yesterday, Black Mirror: Bandersnatch, an interactive movie, was released on Netflix, and my entire Twitter feed immediately became very annoyed. Apparently it was accompanied by a suite of breathless articles about how daringly innovative it is, a claim that ignored decades of prior experimentation, including four earlier interactive releases by Netflix itself. Oh, but those were aimed at children! Black Mirror is serious grown-up drama, and high-profile at that.

But also, it’s Black Mirror, which means the whole thing was constructed around the constraint of making CYOA dismal. It does this by going meta. Most of the meta is is reasonably restrained, almost even subtle, but a couple of branches take a dive into the crassly meta. The story is of a game developer named Stefan Butler in 1980s Britain — a distinct branch and era of gaming history, presented here condescendingly but, based on what I’ve read about that scene over at The Digital Antiquarian, fairly accurately — descending into madness as he works on a game with CYOA-style branching narrative, based on a CYOA-style book whose author also went mad in a similar way. Part of his madness is a sense that he has no free will, that someone else is controlling his actions. There’s a sense that he’s coming to be aware of the metafictional truth because he has some memory of the failed branches you’ve put him through; in some cases, he “wakes up” from a branch as if it’s a dream, and one early choice seems to change how multiple characters behave in its replay (something that I find myself thinking of as an “Undertale choice”).

It’s all very thematically tight on paper, but it all hinges on the idea that Stefan lacks agency because he’s under the viewer’s control, and it undercuts this idea by not giving the player a whole lot of agency either. It feels like most of the off-trunk choices just result in immediate failure and rewind, or maybe one other choice before failure and rewind. Some of the choices even deny player agency by using a choice to assert authorial control. At one point, you’re given the choice of “shout at dad” or “pour tea over computer”. The story needed Stefan to behave irrationally as a result of his lack of control, so it put the irrational behavior into viewer choice. But neither of the choices reflected my desires, so I was just as powerless as Stefan. The writer either expected an audience of sadists, who would relish such a choice, or wasn’t thinking about the the experience of the interactivity at all there. At another point, the player is given a meaningless choice between two ways for Stefan to fidget just so he can be shown successfully resisting your command. Well, good for you, kid. You sure showed me. How about we stop fighting and team up against the writer?

A lot of this is the result of treating the format as a gimmick rather than a medium, but some of it, particularly the shallow structure and inconsequential choices, can be blamed on technical limitations. The fact is, streaming video makes it hard to do the sort of narrative interactivity we’re used to seeing in games, as I learned while working on the Netflix adaptation of Minecraft: Story Mode. ¹My involvement was slight, but it mostly involved removing things: pruning branches, getting rid of conditional logic, simplifying it into a form that Netflix could handle. And the result was still far and away the biggest, most complicated interactive video they’ve got. People who say that Telltale games are just barely-interactive movies have no idea what they’re talking about. Throwaway callbacks are suddenly expensive, because they require an entirely separate video stream. Choices have to be spaced out — you have to give about two minutes between choices because it has to buffer both branches in advance to keep playing smoothly. This also means that the video clip that plays in the background of a choice has to complete playing in full, which I found particularly irksome. You could make your choice in the first second, but Stefan would just sit there indecisively while his dad repeats “Well? Which do you want?” and similar filler. Streaming video just isn’t the ideal medium for this sort of thing.

But it may be the most accessible. If this is what it takes to get interactive narrative deeper into the mainstream than it already is, should I really complain? And, as gimmicky as it seems to those of us steeped in the stuff, it probably at the very least serves as a good showcase of the platform’s capabilities. One of the first choices you get, of which of two music tapes to listen to, has a very obvious callback after the story has trunked, as if just to tell us that it’s capable of keeping state. (This isn’t the only piece of state-tracking, but it’s the only really obvious one.) At another point, there’s something that’s almost a puzzle: you use a special UI to enter a telephone number that was clued in a subtle and cryptic way earlier. The solution is thrown in your face while the UI is up, so it isn’t actually relying on the viewer to solve anything. Maybe it did in an earlier draft. Regardless, what it’s communicating is “We could have made this a puzzle if we wanted to. That’s something we can do.”

Ultimately, it’s a first-released work of IF by a new writer — not new to writing, but new to IF specifically. It may have a larger budget than your typical Comp entry, but it’s about the same length. It should be welcomed as such, but also criticized as such.

I did one more stream, this time playing Desktop Dungeons from the beginning. Why the switch? Partly because I figured the few people who watched were probably getting bored with Gearheads, but mainly because I wanted to see if the same problems would manifest. They did. My software reported that in excess of 98% of frames were being dropped.

So I’m thinking I’m giving up streaming, at least for a while. It’s not that the problems are unsolvable. It’s that they’re unsolvable without buying additional hardware, and that’s more than I’m willing to do on a lark. All the streams I’ve done so far have been from a 2015 Macbook Air, over wifi. “You fool,” shouts every streaming advice article on the ‘net, “you should never stream over wifi!” And I was all ready to hook the machine up to my router via ethernet, and was even pleased to discover that I had an ethernet cable long enough to reach from the router to my good streaming spot, but then I realized that a 2015 Macbook Air doesn’t have an ethernet port. Ethernet-to-USB adapters exist for this purpose, but that falls under “buying additional hardware”.

OK, so maybe I could stream from my desktop PC instead? I could stream a game, certainly, but this machine doesn’t have a camera, and it seems like watching the player’s face as they gurn and overemote is an essential part of the experience. I could buy a webcam, but again, that would be buying additional hardware. At this point I start getting clever and thinking: What if I hook up the laptop’s video output to the desktop machine? I’m pretty sure I have a video out adapter for the Macbook somewhere. That way I could use the laptop’s camera. I could even play the game entirely on the laptop but use the desktop machine for streaming — which sounds more efficient anyway, because that way the game and the streaming software aren’t competing for cycles. But you’ve probably already spotted the flaw in this plan: my PC has no video input.

Even if I fixed my hardware problems, my chief limitation is probably just the upload speed from my ISP. So I’m giving up, but I imagine I’ll have the means to do proper streaming at some point in the future. I mean, I didn’t get as close as I did on purpose. Laptops just happen to come with cameras now. Open-source streaming software exists. It’s kind of amazing how easy it’s gotten to do video broadcasts over the internet that fall just short of adequate.

As for Desktop Dungeons, I may get back into playing it. I still haven’t beaten the final dungeon.

I did another stream. I’m still having the same framerate problems, which I mysteriously didn’t have in my first session. But instead of describing that, I want to go into a little detail about some mechanics in Gearheads that I haven’t described previously.

In addition to varying the toy types available to both sides, most levels distinguish themselves by varying the terrain. Outside of the special levels, which all have unique graphical themes, there are four main variations, apparently called “kitchen”, “garden”, “frozen pond”, and “factory”. The kitchen levels are just plain tile floors without any special features. Garden levels sometimes have strips of mud, or heavy rocks, or insects crawling around and getting in your way — all things that impede your progress and force you to wind your toys more.

Rocks are also sometimes found on the Frozen Pond levels, but Frozen Pond’s more distinctive feature is the cracked spots that turn into holes when trodden on repeatedly. One the game’s merrier points is watching the computer AI obliviously send toys into these holes. Diagonal-moving toys (the kangaroo and the zap-bot) seem particularly prone to this. I have to admit that I myself can’t always predict the trajectory of a diagonal mover with any accuracy, but at least I notice when it leads into a hole and stop launching more along the same path. Even without these features, though, ice levels are slippery. This greatly affects how toys interact. Less friction makes it harder to immobilize a heavy pusher like Big Al by throwing mass in front of it, but also means that collisions tend to send the lighter toys careening backward. And that means that they tend to escape collision effects that require proximity, like the Deadhead’s scream or Disasteroid’s blaster. So you have to get used to things operating by different rules, although the result does kind of come down to “Try to win quickly with fast things”.

Factory levels tend to be the most complicated. They have three distinctive elements: impassible obstacles that raise and lower in a set rhythm, pusher tiles that act like conveyor belts but don’t look like them, and portals. The pushers can be oriented in any cardinal direction. If they’re oriented horizontally, they’ll make things easier for one player and harder for the other, and presumably for that reason, they always seem to be placed in symmetrical pairs. Portals always come in pairs, one that’s an entrance for left-going toys and an exit for right-going, and another that’s the reverse. I tend to think of portals in games as shortcuts, but in this game, they tend to function as the opposite: two rows will have paired portals poised to catch toys just before they reach the finish line and send them back to traverse the same distance again. Not good for scoring points, but great for keeping defensive units like Krush Kringle on the screen longer.

I’ve done a second streaming session. I intend to keep doing these until I get it right, which I clearly haven’t yet. I managed to keep up the patter better than in my first session, and I managed to get all the way from level 1 of Gearheads level 24 in one go, but my changes to the OBS streaming settings seem to have made the lag/framerate problems worse, not better. I’ve looked at the resulting video, and it’s at the point where it’s best described using seconds per frame instead of frames per second.

Well, I have some more ideas to try, starting with reducing the resolution (default resolution is a waste for this game) and shutting down all other devices that use the Internet. One obvious thing that I actually mentioned on the stream is to switch from the Mac laptop I’ve been using to my more-powerful Windows desktop, but for whatever reason, DOSBox runs a lot better on the laptop. If I play from the desktop, improving the streaming won’t matter, because a stream that perfectly replicates what I’m seeing on my screen would still contain noticeable hitches.

In short, via streaming, I have managed to turn Gearheads into the sort of technical challenge that I’ve always found easier to blog about than game content. If you want to watch tonight, the URL is https://www.twitch.tv/muckenhoupt.

So, I spent more or less a full day of tinkering with streaming software, gradually finding answers to questions like “How do I make the game fill the screen?” and “Why isn’t there any sound?” The software I used, OBS, is capable of recording video instead of (or in addition to) streaming it, so you can fiddle with settings until it’s right. Nonetheless, there seem to be some streaming-specific problems that you don’t get from a local recording. I set twitch to keep a recording of the stream, and the recording has problems with laggy pauses. It seems to me that they correspond to points where there’s a lot of sound. I’m trying fix that by following advice online, but the only way to know if it works is to do another stream. So there’s my motivation to do more streams right there.

I didn’t make any progress in the game. I got as far as that Disasteroid vs Cockroach level a few times while trying to set up the streaming software, but no further. During the stream itself, I decided to start Tournament Mode from level 1, so I didn’t get anywhere near there. Doing this reminded me of something I had forgotten: Every level you win gets you an extra life. So there’s a potential strategy of starting from level 1, where it’s easier, just to accumulate enough lives to get through the later levels. But you’d have to be a lot better at the game than me to apply this strategy. I’ve only made it as far as level 25 once, so any strategy that starts with “play all the way through level 25” is a no-go for the time being.

I didn’t really give advance notice of the first stream, because I didn’t want much of an audience for my first try, but I’ll probably be doing them daily at 6:00 PM Pacific time (9:00 Eastern, 2:00 AM UTC) for a little while, whether anyone watches or not.

Let’s rewind a bit. Apart from Galaga: Destination Earth, which is shelved due to technical problems, the last game I started but didn’t complete on this blog was Gearheads. Can I polish it off before year’s end? Maybe. I was pretty far along, having gotten through the first 24 of its presumed 36 levels.

Trying it again just now, it took me a full game to get used to the controls again, but in my second go, I started from level 25 and and reached level 30, a special level where you have Disasteroids (the unstoppable killer robots) and the computer opponent has cockroaches. This doesn’t mean as much as it sounds like, though. As I’ve noted before, this is a game that, despite never having been a coin-op arcade game, is built on the coin-op arcade game model. It doesn’t save your progress, and it gives you a limited number of lives before you have to start over. You can start over from level 13 or 25 if you want, but if you can only get through six levels before game over, you’re not going to reach the end.

I recently noted how steady incremental progress kept me coming back to Creeper World 3: Arc Eternal. Lack of progress has, I think, had the opposite effect with Gearheads, making me slightly dread the prospect of devoting effort to it, even though it’s a perfectly good little game, and not even all that hard, due to the luck factor. But my impatience to win hurts the experience. I’m considering spicing up the experience by trying my hand at streaming, thereby transforming my motivation from trying to get a game off the Stack to trying out a new way of playing games in general. It would, at least, add some new challenge to the experience: the challenge of figuring out how to stream a Windows 3.1 game running under DOSBox on a Macbook.

Now that I’ve broken silence, I should probably say something about what I’ve been playing for the past few months. I pretty much skipped the IF Comp this year — I tried, but it was an especially big year, and I just wasn’t in the mood for it, and wound up playing less then 10 games total. Instead, I have to confess that I spent an enormous amount of time on Creeper World 3: Arc Eternal. Probably more than it deserved, but I found it a tremendously easy pastime to default to.

I’ve posted briefly about Creeper World and its first sequel before. All three games are basically novel real-time strategy games, in which you expand a network of nodes that carries the “packets” you need to build and power weapons to fight an enemy called “creeper”, which is a fluid. Creeper emerges from “emitters” and just kind of pools and spreads out until it starts damaging your structures. The third game is back to the top-down view of the first, but brings along some of the mechanical improvements of the second game: that you can harvest ore (if it’s available) to produce your own “anti-creeper” that physically acts like creeper but is on your side, and that you can actually destroy the emitters instead of just parking cannons around them to destroy any creeper the moment it gets emitted. Destroying emitters makes for a much more satisfactory and conclusive-feeling victory.

It then adds some new mechanisms of its own, steadily increasing the complexity by introducing new things you can build and the conditions that make them necessary, as is customary in RTS campaign modes. There’s a “forge” that lets you mine “aether” to research upgrades, terraforming machines that slowly reshape the land per your instructions, and so forth. I particularly like the “guppy”, a flying non-combat unit that carries a cargo of packets to a designated landing spot. This lets you leapfrog past creeper-infested areas and build in areas disconnected from your base, enabling tactics otherwise impossible. I suppose it’s the game design pattern of “Impose arbitrary restrictions, then grant the player special powers to overcome them”, but it’s a well-done example of it.

Still, even as the mechanics get more complicated, the winning strategy remains more or less the same. Rush to grab as much land as you can defend. Defend it. If you have enough power to keep a stable border with the creeper, you can spend any excess on building what you need to break the stalemate and grab more land. Levels vary what challenges they present, and what resources they provide to meet them, but the rhythm of the game remains constant.

Until Farbor.

Farbor is the second-to-last level in the campaign, and it makes you hurry. In it, a number of enemy drones are collecting ore — the same ore you use to create anti-creeper — to build a monstrous invincible spacecraft, Sinistar-style. You only have so much time to stop them. You can build weapons to shoot down drones and slow down progress, but some of them are are out of reach, and you still have to fend off creeper while you do it. It seems utterly impossible to do everything fast enough. In fact, it’s not as difficult as it seems, because it moves the goalposts a couple times. If you fail to keep the ship from being built, you’re told that it’s going to go over to another building that will power it up and make it unstoppable. And if it reaches that building intact, you’re then told that it will take a full twenty minutes to power up. But the first several times I tried the level, I quit and restarted well before that point, when all seemed lost. And I remained in that state for a couple of months.

During those months, I tried the bonus levels. And that’s where it turned from a game to a habit. By the time you reach Farbor, you’ve unlocked two distinct sets of bonus levels: Tormented Space, which consists of ultra-hard levels, and Prospector Zone, which has fairly gentle levels full of collectible nubbins. And in both cases, there are many, many levels to try. I’ve barely made a dent in even the Prospector Zone, let alone Tormented Space. Probably because of the quantity, there’s a certain sameness to the maps after a while. They vary on a limited set of axes: map size, whether or not there’s ore and aether, whether the map is fully-connected or broken into islands, what enemies you’re facing in addition to mere creeper. The standard strategy still applies, without a lot of the frills they added to the main campaign. But the fact is, that seemed to be what I wanted at that point. I could always make progress. I could always sit down at night with a level I had never seen before, and finish it. And that apparently appealed to me while I couldn’t do the same with Farbor.

Also, the predictability makes it oddly satisfying for a game about warfare. Especially in a large level, you tend to set up your weapons and just leave them alone for a while, operating as a big machine, packets zipping along their lines, guppies sailing back and forth on their errands. Sometimes you place a bunch of weapons in one part of the map and then turn your attention somewhere else, and then when you look at the weapons again, they’ve done their job and cleared all the creeper out of an area, which is highly gratifying.

Eventually I decided that I had gained enough expertise to tackle Farbor again, and learned what I’ve already said, and finished the game. And I might have stopped there, except that completing the campaign unlocks the Alpha Sector. This is another set of bonus levels, also large but not as large as the others. But these levels were made by testers during the game’s development, apparently before they had established standards of style or balance. So they’re not as polished as the other levels, but by the same token, not as uniform. One person will have a bunch of small levels clearly made by just scribbling around in the level editor. Another will have a meticulously-planned puzzle, where only one approach works, or even just use the scripting system to make a puzzle that has nothing to do with the usual mechanics of the game. Others take things to extremes, giving you a map that’s vaster than what you’ve seen before, or one that starts both you and the enemy off with far more power than you’re used to, or that bombs you with waves of creeper spores every second instead of every couple of minutes. Some levels are jokes. All of them feel personal. And that makes them fascinating to me. I can’t always sit down and win a level, because some of them are just unduly hard — the game specifically warns you that there’s no guarantee that they’re all even possible. But I can always sit down and see something new.

Come to think of it, the Alpha Sector is a bit like an RTS version of Cragne Manor. It’s even more like the fan-made levels that lots of games collect, I suppose, but enshrined as somehow special. Context makes it feel less like a user-made DROD hold and more like the rejected puzzles in an official hold’s Mastery area.

Back in June, noted interactive fiction authors Ryan Veeder and Jenni Polodna sent out a call for contributions. For the 20th anniversary of Michael Gentry’s classic Lovecraft-inspired cosmic horror game Anchorhead, they wanted to make a collaborative tribute game, where each participant writes one room. They expected about a dozen people to express interest. Instead, they got more than eighty, including me, but also IF luminaries Emily Short and Andrew Plotkin, Kingdom/West of Loathing authors Zack Johnson and Riff Conner, and even Michael Gentry himself. It’s one of the largest collaborative IF projects ever. Not the very largest, though; apparently there’s a furry porn game that has it beat handily, furry porn inconspicuously leading the way as always. But it had more authors than the annual IF Comp has ever had. The resulting game, Cragne Manor, was released to the public just a few days ago, after a lengthy testing period where the authors shook out the problems created by putting all the pieces together.

Again, each participant was responsible for a single room, although some bent this rule by creating sub-rooms or just plain additional rooms only accessible from their main one. Part of the organizers’ core concept was that they wanted the game to be a mishmash of authorial styles and intentions, like a patchwork quilt. And so they insisted that each author work basically alone, with no knowledge of what other people were writing, apart from how it directly touched their own work, exquisite-corpse-style. The organizers provided the bones of a plot and setting (one Naomi Cragne searching for her lost husband Peter in the fictional town of Backwater, Vermont), and negotiated with each writer how their room fit into the map and the game’s puzzle structure. Some, for example, were told “Your room contains a book which is one of many that needs to be returned to the public library for a puzzle. Here’s the specifics of how to implement a library book for this game.” Some others were told “Your room should have a puzzle that uses an object from another room to obtain an object used in a different other room, and we need to coordinate on what those objects are.”

The result is, as expected, incoherent. It reminds me a little of Deadly Premonition. Near the beginning of Deadly Premonition, before you even get to the town where the murder you’re supposed to be investigating took place, you fight your way through a zombie outbreak. The moment you reach town, the existence of zombies is forgotten about. That’s what Cragne Manor is like. Individual rooms confront you with horrors beyond imagining, scientific marvels, and dire revelations about the Cragne family that are only acknowledged in that room. One author, tasked with making a bridge, decided to make it a rope bridge in a cavern, even though both ends of the bridge are ordinary streets in the town of Backwater. And yet, it’s somehow surprisingly coherent for such an incoherent work. Each room is basically its own independent reality, but they sometimes sync up in fortuitous ways. Multiple rooms contain mirrors that act as portals to the past, something that their authors thought up independently, creating a sense of a general mechanism. The aforementioned bridge room features the colossal skeleton of some extinct monster; shortly after crossing it, you come across a paleontological dig. Seeing the strange bones uncovered there, your mind automatically draws a connection to the ones under the bridge, even though they seemed to be in a completely different game.

Also, a few of the more ambitious writers created things to give a sense of cross-room connection beyond the organizers’ plans. Lucian Smith made a puzzle that follows you around and interacts with those library books I mentioned. Emily Short’s room, otherwise one of the simpler ones, contains a creepy pull-string doll that comments on random objects in your current room by scanning their descriptions for words that she guessed other people would be using. (This is useful in some places for identifying objects you failed to notice.) Nonetheless, most rooms are self-contained or almost self-contained. One of the game’s big challenges is getting used to the degree to which you should ignore stuff from other rooms. One of its big design problems is that several authors decided to make “obtain a cutting implement” puzzles, whose cutting implements can’t be used on each others’ cuttable items.

Mainly, though, the style and mood is wildly variable in a very fun way. Not every contributor was familiar with Anchorhead; not everyone who was familiar with it chose to imitate it. Some rooms are brimming with Lovecraft mythos references (something that Anchorhead itself notably did without, despite clearly bearing Lovecraft’s influence), and one or two even imitate his prose style. Others are ghost stories, or observations of small-town life, or surrealist, or comic, or gross. Adjacent rooms are often jarring juxtapositions. (Chris Jones’ meat packing plant bathroom — just the name of the room is full of promise! — is especially notable for pulling off a number of these weird juxtapositions within itself, as if reflecting the game as a whole.) There are crypts and tentacles and dark rituals and monstrous fungal blooms. And there’s lots and lots of books. Everyone knew that there was a puzzle track involving library books, and many people seemed to take this as permission to throw in journals and histories of their own. It’s been merrily pointed out that Backwater has more libraries than bathrooms.

The game is large. Just having more than eighty rooms makes it a large game in that sense, and some of the rooms are large individually, containing enough prose or puzzle content that they could have been released separately. Hanon Ondricek’s church scene, for example, is essentially a novella, and Andrew Plotkin’s workroom is a miniature Hadean Lands/Myst mashup, teaching the player a remixable system of magic words that can transport you to other worlds. (As with nearly everything in the game, those magic words only work in the room they were designed for.) On playing the full game, it was easy to feel like my own contribution was unusually slight, but I think that’s an illusion created by the fact that the larger rooms dominate the play experience.

Largely as a result of those large rooms, the last few rooms feel anticlimactic, as you use your hard-won inventory to perform a relatively simple ritual and wind up in a relatively simple and utterly disconnected endgame that doesn’t address anything that happened before. This is perhaps inevitable. A work in this genre should end in the protagonist coming to a realization that ties all their bizarre experiences together, and how could you possibly do that exquisite-corpse-style? For my money, the real climax of the game comes slightly before the ending, in a room that directly confronts Naomi with the fractured and mutable nature of her reality and identity, which she’s been oblivious to and which the player has been struggling to ignore through the entire game.

I highly recommend playing the game, although it’s probably best done with a group. Not necessarily as a group play session, but as a bunch of people who are discovering the game independently but in tandem, who can help each other through the more obtuse puzzles (some of which are pretty obtuse), laugh together at the more ridiculous things, congratulate each other on beating the larger rooms.

Kudos to Jenni and Ryan for tackling the unexpectedly mammoth task of integrating everyone’s disparate contributions into something playable. Communication is always the most difficult part of any large project, and actually making it against the rules didn’t help matters. One notable innovation they added is a divination device, discoverable within the first few rooms, in the form of a coffee cup — a subtle Anchorhead reference; some Anchorhead players carried a discarded coffee cup from the first few rooms with them for the entire game for no reason, so this time there’s a reason. Once you learn how to read it, the cup tells you whether you’ve solved all a room’s puzzles or not, and, if not, whether you have everything you need. During testing, I played the game for a while before this device was added, and found that it drastically improved the experience of the game. I wouldn’t necessarily want such a thing in a game produced under a single unified vision, but in Cragne Manor, it was immensely useful in clarifying the ever-shifting authorial intent.