Sunday, February 14, 2021

Unwinnability and Wishbringer

Someone asked recently about a version of Zork modified to avoid all the classical Infocom annoyances. Expiring light sources, random combat, crucial items stolen by the thief... Zork comes from an era where all these were simply part of the fun. You saved a lot, optimized your run, and tried again if you failed. But the gaming world -- I include myself -- has largely decided over time that this isn't fun.
As far as I know, there is no such modified Zork. But it's an intriguing idea! In fact someone is working on a modified Planetfall in just this vein. A new MODERN command switches to a mode with no hunger timer and, ideally, no unwinnable situations.
When we talk about those early games, it's easy to fall into the trap of calling those unwinnable situations mistakes or design flaws. Once you do that, you're stuck searching for explanations of why the designers made those mistakes. Answers come in two flavors. Either (a) the state of the art was too primitive to understand that unwinnable situations were bad; or (b) the designers were padding out their games to a given level of difficulty or play time.
Speaking as one who played the games back in the day: this is just wrong.
This Giant Bomb article puts it well:
The early generations of text adventure games tended to have a lot of chances that could lead players to unwinnable states, as a way to make a game deeper and more challenging; this kind of game design was not yet considered unfair to players. It was usually considered a product of the game’s difficulty rather than poor design and encouraged (or, as its critics would say, forced) replayability.
The need for many failed run-throughs before one figures out the perfect approach was certainly an element of the challenge. But it was never padding. It was as much a part of the design as the treasures and the magic words.
There's a deeper confusion, though. The same article goes on to say:
Mike Dornbrook, Infocom's Head of Marketing, conducted a customer survey in late 1984 which showed a clear correlation between the Infocom games players considered their favourites and the games they had actually finished. This piece of marketing intelligence led to the more foolproof design of Wishbringer and later games.
-- ibid
Wishbringer was Infocom's second junior-level game. It is friendlier than earlier games, more forgiving, and more generous in its design. Most of the major puzzles have at least two solutions. But it's a misconception to say that Wishbringer avoids unwinnable situations, or that Infocom changed its design stance after 1984.
It's true that Wishbringer was immediately followed by A Mind Forever Voyaging, which was focused on narrative exploration rather than puzzles. But this was just halfway through Infocom's high season. Their later slate (Spellbreaker, Leather Goddesses, Trinity, Lurking Horror, ...) embraced the model of failable game design and refined it into some of their best-regarded work. With Journey, Infocom redesigned IF from the ground up (discarding the parser!) -- but doubled down on failable puzzles, structuring the game around a limited resource system which took much trial and error to solve.
To untangle this, we're going to have to dig into the ways in which games can become unwinnable. How does it happen and why?

First, consider the difference between losing the game and the game becoming unwinnable. If you attack a troll and it hacks you to death, it's very clear that you've lost! RESTORE (or UNDO) and try again. Whereas if the troll breaks your sword and you run away, you live -- but if you need the sword elsewhere, you're hosed. You might make progress elsewhere in the game, but ultimately you'll have to throw it away and redo it.
Or maybe a broken sword is a clear sign of failure? Adventurers don't waste resources! There's an ambiguity here: if the game signals that you've made an irreparable mistake, you know to back up and try again. But signals are a matter of convention and culture.
(These are the issues I was trying to get at with my old "cruelty scale". When you've fallen into an unwinnable situation, do you know it? Reasonable players may disagree, particularly if they're from different decades. Some games clarify with pointed mutters: "You feel like that was a mistake," or perhaps by penalizing your score.)
Here's a subtler question: have you really lost progress just because you need to restart and try again? In an adventure game, progress is understanding the world and solving puzzles. Solving a puzzle again is just a bit of extra typing!
I'm pushing a point here, but the classical parser games really did have a different rhythm than we're used to nowadays. Even on the old eight-bit machines, a experienced player could replay a section of a game nearly as fast as they could type. It wasn't fun, no. But if you already expected to need to restart a few times, then exploring with unwinnable mistakes hanging over your head just wasn't a serious penalty.
This balance shifted with the later graphical adventures. In the third-person (Sierra/LucasArts) games, walking and other animations made replaying a much more tedious chore. In the Myst-likes, the loading times of those graphically rich scenes slowed you down nearly as much. It's no surprise that the "unstuckable" game model had its renaissance in the graphical era.

But we should take a closer look at the ways a game might become unwinnable.
Let us assume that you are hosed. Why? Because you need something to win the game and you don't have it. Either you've lost access to it and you can't get it back, or you've run out of it and you can't get more.
In practice, these situations usually fall into a few general types:
  • An area gets closed off partway through the game. You left some critical item there and you can't get back to it.
  • You run out of time. Or you run out of something that is consumed over time: lamp fuel, food, health.
  • You accidentally destroy or lose an item while experimenting.
  • You use up a resource trying to solve a problem, but it's the wrong problem.
  • Random events in the game ruin your path to victory.
What led authors to create games with these situations? What did the games, and the intended audiences, gain from them? Let's go point by point.

An area gets closed off partway through the game

What we gain: A sense of a living world (if the game geography changes), or a dynamic storyline (if the player is pushed into a new area).
The fun this adds: The challenges of planning and thoroughness. You have to sweep an area for every possible resource. If you can't snarf them all on your first run-through, you at least have to remember what was available. When facing later puzzles, you have to consider everything you've encountered in the game, not just what you've got right now.
(Yes, I am taking for granted that more challenges mean more fun. Obviously you don't have to be into this kind of fun! But when we analyze these old games, we have to think about what fans got out of them -- and the fans, by definition, were the people who enjoyed them.)
How do we get this in a modern game? Think about narrative games which aren't parser IF. They typically don't have a DROP action! If the player can't leave objects scattered around, the problem of closing off areas becomes much easier to manage. You just have to make sure the player picks up everything they need -- either as part of the narrative, or in order to solve the puzzles leading to the section break. (This is the old cliche of opening a chest and finding a key underneath a coil of rope. No getting out of that room without the rope!)

You run out of time (or light, or food...)

What we gain: A sense of tension.
The fun this adds: The challenge of optimization. First you have to figure out what to do; then you have to figure out how to do it in the limited time available.
(Captain Verdeterre's Plunder is an example of building an entire game around this challenge.)
How do we get this in a modern game? The familiar gimmick is to narrate a time limit, but really sync the clock to the player's progress. The mine timbers may shudder threateningly, but they won't actually fall until the exact moment the player dives clear. The idea isn't to fool the player -- it's usually a pretty transparent gimmick -- but to give them a mood to play along with.
(I recall Christminster as the first time I noticed this gimmick, but I'm sure Infocom used it first.)

You accidentally destroy or lose an item

What we gain: A sense of a systematic model world. You can eat food and then it's gone. You can throw items down a pit and then they're gone. You can set things on fire and then they're gone, or perhaps (if the game model is ambitious) scorched and ruined.
The fun this adds: These systematic interactions can be worked into puzzles; everyone loves a systematic puzzle mechanic. But the puzzles won't make sense unless the player has already experimented with the interactions. That means giving the player rope (no pun intended) to hang themselves.
How do we get this in a modern game? This is a tricky one. You have to come up with ways for every single resource to be replenishable, which can feel artificial.

You use up a resource solving the wrong problem

What we gain: A sense of deep exploration. There are many ways to approach a problem, and you may have to explore several of them before you're done.
The fun this adds: You've solved a bunch of puzzles, but the solutions are mutually exclusive -- you need one item in three different places. Now you have a metapuzzle! Go back and juggle the previous solutions until they fit together. You may have to take another stab at one of the puzzles and find a new solution to it; but first you have to figure out which one.
(Enchanter is my earliest glowing example of this idea. That dispel scroll solves all sorts of puzzles -- but you can only use it once!)
How do we get this in a modern game? This is the road that led me to Hadean Lands. My solution was (a) the RESET command, which effectively replenished every resource by restarting the story; and (b) the goal system, which removed the tedium of retrying a sequence of puzzles. Everything was geared towards letting you grapple with the metapuzzles.

Random events in the game ruin your path to victory

What we gain: Okay, this one is the joker. Even in 1978, nobody was happy that the thief was running around Zork stealing your treasures. Yes, it provided the challenge of persistence -- you had to be bloody-minded enough to retry the game when the thief boned you. (Or, in Zork 2, the wizard.) But it wasn't a surprise that Infocom retired this theme from their repertoire after about 1981.
(It turned up again in Beyond Zork with its randomized mazes and monster combat. And, yes, everyone called it a design flaw. I don't know what Moriarty was thinking.)
How do we get this in a modern game? Do the math -- or run the test suite -- and verify that no combination of random events can hose the player.

All right, we've walked the bounds. Where does Wishbringer lie in this territory?
(Spoilers for Wishbringer in this section.)
(I will confess that I did not replay the whole game for this article. I played through the opening chapter, familiarized myself with Festeron, and pulled down the source code for reference purposes.) (Oh, this is the 1985 version of the game -- release 69, serial 850920. Note that the 1988 "solid gold" release has bugs in its time-of-day code!)
Wishbringer's opening salvo is a strict time limit. The game starts at 3:00 pm; you must reach the Magick Shoppe by 5:00. That's 120 turns. If you optimize your route -- which is to say, on your second or third try -- you can make it under 40 turns. But the typical player will examine the scenery, go back and forth a bit solving the initial puzzle, look around town -- and run out the clock.
Looks like the story's over. But don't despair! Interactive fiction lets you learn from your mistakes.
The "mistakes" here include mapping the town like a diligent adventurer. That definitely takes more than 120 turns.
The town holds a few other ways to wreck yourself during that initial exploration:
A shower of sparks erupts from the back of the game machine, and the video screen goes black.
That's unfixable. There's also a seahorse who dies three turns after you encounter it, unless you offer aid. And so on.
But let's say you skip the tour, head straight towards the Magick Shoppe (the game's feelie map is helpful), and reach your goal by, say, 4:30 pm. The shopkeeper introduces your true quest, hands you a joke snake-in-a-can, and shoos you out to a changed landscape.
You now face your first serious puzzles: getting down from the hill and back across the bridge into town. The hill path is a fog-shrouded maze, but you were explicitly advised to map it on your way up; if you didn't, you can bumble your way through by trial and error.
As for the bridge, it follows the game's master schema by offering two solutions. You can hand the troll the snake can -- now containing a live snake in this altered world -- which scares him off. Or, if you've brought the horseshoe and the gold coin from town, you can WISH FOR LUCK and give him the coin. (It's not legal tender in Witchville, but "luckily" the troll doesn't notice.)
But note that both these solutions are failable. If you haven't searched the town for both the coin and horseshoe, and you release the snake before meeting the troll, then the puzzle becomes impossible. In fact, the "easy" path of the wish is rather more likely to be foreclosed -- a beginning player probably won't have scoured the town for items.
I didn't play much past that point, but I ran into another amusing failure back in town. (1) Get thrown in jail before extracting the stone from the can; (2) crawl down into the tunnels; (3) get lost in the dark; (4) squeeze the can. The stone falls into the dark and is lost forever. Whoops.
So Wishbringer is already three for five: time limits, closed-off areas, and accidentally destroyed resources.
Then there are the wishes themselves. Each can be used at most once. Some have more than one puzzle application; all can be wasted through experimentation. That checks off both "accidentally destroyed resources" and "solving the wrong problem". No, you're not strictly making the game unwinnable, since every puzzle can be solved without recourse to wishes. But if the intended experience of the game is "win using wishes, then replay and try winning without any", then wasting wishes pushes pretty hard into breaking-the-game territory.

Let's be clear: Wishbringer has many affordances for beginning players. It begins in a tightly bounded area with a handful of rooms, two takeable objects, and one puzzle. Solve that and you gain access to the town; but the game firmly (and repeatedly) reminds you that the Magick Shoppe is your first goal. NPCs interact with you in brief scenes with clear direction. You can (and, in some sense, should) go off-track; but the track is clearly laid out.
There are also, despite my litany above, several guard-rails against unwinnable states. If you open the snake-can inside the Shoppe, the proprietor resets it for you. If you drop an item in the fog outside, it will be found waiting for you at the bottom of the hill. And I think that the game makes an effort to always show your failures as clear results of your actions, or (at worst) of your missed actions.
What the game does not have is a systematic design of avoiding unwinnable states entirely. It just avoids a few of the worst ones! Nor does it make any attempt to lead the player through to victory on the first run. Wishbringer is, in this sense, exactly like Hitchhiker or Spellbreaker: a game which wants to be approached iteratively. You will try some stuff, fail, back up to an earlier save, retry, save again, realize what you really needed the first time, do a definitive run of the first chapter, save, explore the town for a while, fail in a new way...
The saga of the unlosable adventure game begins a few years later. ...I should close with a rhetorical flourish by naming it, right? Only I don't have any idea when the idea took hold. I didn't play enough of those early Lucas/Sierra games. (Certainly by Loom (1990) we all knew what the score was, no pun intended. But that feels late.)
I don't even remember if Infocom got there. Did Nord and Bert have unwinnable states? Sheesh, it's been way too long.
Well, there's your blog post hook. Enjoy!


  1. Jimmy Maher in the Digital Antiquarian credits Ballyhoo for inventing the narrated time gimmick.

    I don't remember whether Ballyhoo has unwinnable states, though.

    1. Ballyhoo indeed does have at least one unwinnable state, and this guy found it.

    2. The Black Sanctum (1981) includes at least two narrated time gimmick spots (the snow comes in only after you're safely inside and further in, and there's a kidnapping that only happens after you're up to a certain point).

      The second time is just a plot point mostly, the first one is somewhat plot-affecting since it means you can't die early on just by wandering outside for too long. There's an item outside that you can accidentally leave outside and make the game unwinnable though once the snow comes in; it's an item "in the open" while the player doesn't have much inventory so it is unlikely they'd get stuck there, but it means it's not entirely airtight -- the intent is more having tension without a set timer, though.

  2. I once talked to Al Lowe about the ridiculous number of "an area gets closed off" unwinnable states in Leisure Suit Larry 2. Some are funny or somewhat intuitive, but several are, "There is no logical reason I should have done that random thing five and a half hours ago."

    He apologized, said he would never make a game like that now.

    1. I didn't really address the question of whether an action is *logical*, since that's even more subjective than whether an outcome is obvious.

      (Is "pick up the gold coin" something you had a logical reason to do an hour ago? Or "pick up the cheese sandwich" for that matter? The traditional klepto adventurer would say "heck yes", but these issues still arise.)

    2. Not to mention, "pick up the gold pixel that doesn't look like a coin nor mentioned in the room description"

  3. From LucasArts, "Indiana Jones and the Last Crusade" had no unwinnable states, but only because the game has an alternate solution for most puzzles where you can use an arcade sequence to fight your way through. "Loom" really was the first truly Merciful LucasArts adventure game, followed later in 1990 with "The Secret of Monkey Island."

    Sierra had earlier games for kids with no unwinnable states, including "Mickey's Space Adventure" (1984) and "Mixed-Up Mother Goose" (1987).

  4. Thanks for the replies! I didn't realize Loom was a trailblazer in this sense.

  5. This brings to mind the new double genre that build failure and retry into the base game play. Rogue-like: fail and retain nothing but knowledge. Rogue-lite: fail but gain persistent bonuses (better tools, weapons, etc). The early unwinnable state adventure games were similar to rogue-like in that one aspect, but it seems much easier to create interesting random variation for replay in an action game than in a puzzle games.

    1. (There's less than 10 comments here, and I managed to not read yours that told my point 12h earlier)(I'm blaming the Internet)

  6. I haven't tried to replicate it recently, but I'm fairly certain that Nord & Bert -- and in particular the Manor of Speaking section -- included some fail states where the game would, for lexical reasons related to that section's puzzle, refuse to allow any of the inputs you needed to give to solve it. My memory is that I managed to only lose but "soft-lock" the game, putting it into a state where it had not crashed but where I could not even operate the metaverbs.

    1. Besides Manor of Speaking (where you can run out of words) I think you can make the game unwinnable in Shaking a Tower by going up the beanstalk without proper preparation--the beanstalk isn't critical path AFAICT but you can't get back down. Though this might just mean that I couldn't figure out how to get back down.

  7. A lot of these "unwinnable triggers" that people think as annoying in an adventure game are exactly what is the salt of the fun in many roguelite (FTL or Out There come immediately to mind) : you learn that you should avoid some things behind, prepare for random encounters, and the graphical interface affords you to go faster with a "speed time 10x" and you don't need to stop anywhere if you already know the story of that place. More often than none you end up is a shitty situation, that is not defined as "lost" but, after a few runs, you "know". The difference being that lately, it seems that adventure games are advertised as "story first, game second" (to the point that "adventure" has been replaced by "narrative") that it feels like a betrayal to not be able to know the end of the story because of the game part (and Hades, which is a good melage des genres kind of nails it with its god mode)

  8. The thing is... with recent advances in game design, it has no sense to has an unwinable state. For "recent adveces" I am thinking about the Jon Ingold deisgn ideas for Heaven's Vault, in what you have various points of going out a scene without the traditional "block because you didn't solve all the puzzles".

    In that philosophy you can either exit a scene with all puzzles completed, or with only some of them, even with bad solvings and information that is wrong, or even you can just... go out of the scene to return later. Even I think you can fail completely a scene and progress further in the game.

    Of course, Heaven's Vault riggers this. It is procedurally rigged so you can always progress, and the whole narrative is designed than even lazy players that don't care about "getting it right" can read the ending (that remembers me about the endings of Rite of the Shrouded Moon, the great and sadly underrated game by Tiger Style).

    Anyway... nowadays we have a lot of resources and design lore to avoid unwinable states: You can provide alternative resources, you can provide alternative puzzles, you can recognice that the game is in a "not optimal state" and provide significant narrative to support that.

    But, to rework those old games to REMOVE completely those aspects... is just to not to understand what makes the genre funny. If I want to throw my magical sword down the well just because of "what happens", please, just let me ruin my game that way, because that the game allow that kind of dumb behaviour is part of the fun of it.

  9. Anyway, definitively there are old games that have "bad design" attached. I'm enjoying a lot the Renga in Blue series, and preciselly I was thinking the other day that reworking and updating some of those old but interesting games could be funny: some of the are pretty short, Scott Adams like. Maybe those remakes could be made pretty quickly.

    Who has the time?!!! Not myself.

    But, someone who really is in position to accomplish this are the guys bringing back the Magnetic Scrolls cathalog. But I ignore how much Magnetic's games had or had not that kind of dead-man walking scenarios.

  10. I like how Lemmings deals with this by limiting unwinnable situations to a level, but within a level all of these things apply - limited dwindling skill supplies, sometimes very tight timer, dwindling sacrificed lemming supply, multiple potential paths which creates complex resource usage calculations, diagonal thinking, sometimes you click in a Lemming crowd and you get a guy going the wrong way and become doomed... as long as it only takes a few minutes to undo your unwinnable state, you're okay.. and I guess that's the thing that changed which caused the whole shift to the Lucasarts style design.

  11. Huh, I never thought of unwinnable states so charitably before. I think the tedium of re-doing what I've already done is worse than being 'stuck.' Good point in that re-typing is faster than watching an animation again etc., definitely. I think this works when you do look at these experiences in a more meta way, because approaching these stories / games iteratively, if you're not viewing it that way, these re-tries hurt some of the magic and immersion (at least personally). Not sure if that makes sense. I think iterative approach works when you're viewing the game as something to solve, but I'm not sure if it works well when it's a 'story.' If I get booted out and have to re-try it just reminds me that I'm not living in that world, I'm approaching it from outside.

  12. I've had a problem with things like your cruelty scale for a long time. The implication always seemed to be something like, "Those uncultured players of the '80s had no idea they shouldn't be enjoying those flawed games." This post is the most nuanced take I've seen.
    I guess I never picked up on the "modern" sensibilities because from 1985 to 1995 our PC only had CGA. So my last Sierra game was King's Quest III, and I never played a LucasArts game until ScummVM was a thing. Also, I never even heard of A Mind Forever Voyaging until I found Classic Text Adventure Masterpieces of Infocom on the store shelf. So for me, an adventure game was always a game to be won, a series of puzzles to be solved. Death or getting stuck are just temporary consequences of doing it wrong, just like in Doom or Super Mario Bros.
    I appreciate your acknowledgement that "even on the old eight-bit machines, a experienced player could replay a section of a game nearly as fast as they could type." Not strictly true for the C64 with its slow 1541 disc drive, but it's an important point. To this day, I can get anywhere in a game like Zork II or Planetfall within a few minutes. Dying isn't really the same thing as starting the game from the beginning, because you know so much more than you did at the start.
    So not only do I think unwinnable states are at most a small cost to the player, but eliminating them can have a large cost. Playing a game that "makes an attempt to lead the player through to victory on the first run" can feel like it's not a game at all, but rather a demo or a movie. I play IF for an interactive experience, not an "immersive" one. I don't merely want to watch a story play out; I want to solve it.
    Also, there's another way to get past the bridge troll in Wishbringer, but it's clearly a bug: PUSH TROLL UNDER WATER.


      Heh. I knew you could get rid of your hands by putting them under water, but I never thought to try it with the troll.

      Maybe it's nostalgia talking, but I honestly can't remember more than a couple of times where I got an Infocom game into an unwinnable state in a way I found annoying. There's the murder mysteries, of course, but there you could argue that it's part of the formula.

      Even there, The Witness is - I think - extremely forgiving. You can spend your time talking to suspects, looking for clues, making plaster casts of footprints, etc... or you could win the game without ever leaving the scene of the crime. The story will make less sense, but it can be done.

      So, unwinnable states... I did fall into the trap in Cutthroats where you die on what would otherwise be the final move. I may have missed that all-important object at the start of Sorcerer, but I'm not sure about that. There's the infamous one with the microscopic fleet in Hitchhiker, but that game actually gives you a second chance to solve that problem.

      So the worst offender I can think of is probably Journey. It's very easy to miss the all-important clues scattered throughout the game (I did), or run out of magical essences right before the end (oh, yes). I think the problem is that in addition to being quite unforgiving, there's often very little to tell a correct solution from an incorrect one.

      I love Journey a lot, but I don't think it loves me back quite as much.

      But on the whole, I think annoying game mechanics are a bigger problem for me. Planetfall has one puzzle where you have to repeat the same thing a couple of times. In Zork II, some of the Wizard's spells mean instant death if they happen at the wrong time.

      And in Zork III, there is a small chance on each move you spend in the Royal Museum's past - and you have to spend several moves there - that the guards will find and kill you. I think there are some other such random instant-death situations in the game, but in this case you won't get resurrected. That's just bad design, if you ask me.

      But Infocom were hardly the worst offenders. I can't think of anything in any Infocom game that made me anywhere near as frustrated as the slot machine in the original version of Space Quest I. (The SCI remake made it a lot more tolerable.)

  13. "I've had a problem with things like your cruelty scale for a long time."

    That scale was an attempt to be purely descriptive! Attaching labels like "cruel" and "polite" was a bit of humor -- and remember that was half my lifetime ago. I might have been being a bit of a jackass about it.

    My current take on that scale is that it is *very* much of its environment: playing parser IF on mid-90s interpreters. For an updated retrospective on the whole idea, see this post:

    As for the unwinnable states themselves, the fact is that we've had 25 more years of game design since Wishbringer. We have "unlosable" games like _Outer Wilds_ and _Heaven's Vault_ and _Telling Lies_ which are in completely new territory of interactivity. Trying to pin everything down to the options that Infocom discovered when I was a teenager just isn't enough any more.

  14. Speaking of hunger meters, it's interesting that they still survive in the "difficult" modes of CRPGs. For example, the Bethesda Fallout series. It's expected that you will play through the game in normal mode, and then, for an added challenge, play on difficult mode where things like eating and sleeping aren't just optional things you do to heal yourself, but are required at intervals.

  15. So, only now do I realize that I only ever played the Solid Gold version of Wishbringer, so I never encountered the actual time limit. And that it's a bug. I assumed the time limit, while stated to give a sense of urgency, had been deliberately removed to be more generous to the player! (Consider Zork Zero telling you that the day seems to be getting a bit long.)

  16. I wonder about this. Turning dramatic narrative into boring puzzles?
    I wonder if losing the game is actually a bad thing. Considering that it's interactive, I think a dramatic telling of the impending doom is a viable solution in literary genre. Just like a CYOA stories with multiple endings where not all stories end gloriously, but all are interesting. So, perhaps the solution doesn't lie in eliminating unwinnable situations, but to actually set a timer to the effect of "if you meander too much, then I will end the game for you whether you like it or not!"