§50   The design of puzzles

The odyssey of ‘Zork: Mimesis’ begins in a field behind a white house. You climb in through an open window, take the water and sack lunch from the table, go in the living room and move the rug aside to reveal – a blank floor! … Soon the owner of the house – an underemployed, alcoholic bricklayer – is covering you with a shotgun as his unfaithful, neurotic wife dials 911. The puzzle-free, super-literary action continues as you are funneled through the criminal justice system…
— Roger Giner-Sorolla, conceding that mimesis is not everything

Without puzzles, or problems, or mechanisms to allow the player to receive the text a little at a time – whichever phrase you prefer – there is no interaction. Inevitably, puzzles are obstacles. Here, Arthur Dent will not be able to meet aliens and have a generally wild time until he has got hold of a babel fish:

>examine machine
The dispenser is tall, has a button at around eye-level, and says “Babel Fish” in large letters. Anything dispensed would probably come out the slot at around knee-level. It bears a small label which reads “Another fine product of the Sirius Cybernetics Corporation.”
>press dispenser button
A single babel fish shoots out of the slot. It sails across the room and through a small hole in the wall, just under a metal hook.
>remove dressing gown
Okay, you're no longer wearing your gown.
>hang gown on hook
The gown is now hanging from the hook, covering a tiny hole.
>push dispenser button
A single babel fish shoots out of the slot. It sails across the room and hits the dressing gown. The fish slides down the sleeve of the gown and falls to the floor, vanishing through the grating of a hitherto unnoticed drain.
>put towel over drain
The towel completely covers the drain.

(For the final solution, still some way off, see ‘The Hitchhiker's Guide To The Galaxy’.) A good game mixes easy puzzles with hard, but no puzzle should be so simple that one obvious command solves it. On the other hand nor should its solution, once guessed, take ridiculously long to achieve, or require endless repetition: such as to fetch something pointlessly distant, or to solve an eight-discs Tower of Hanoi, or to keep juggling objects so that only three are carried at any one time. Here are two basic pitfalls:

The “Get-X-Use-X” syndrome.   By convention, every word or phrase in a cryptic crossword clue is used exactly once to account for some part of the answer. The equivalent in adventure games is the equation “one object = one puzzle solved”, where the player picks up a bicycle pump and looks for a bicycle, picks up a pin and looks for a balloon, and so on. Once used, an object can be dropped, for it surely will not be needed again. But this convention rapidly drains away realism, and most designers try to break the equation in every way possible: with red herrings (one object = no solutions), collection puzzles (many objects = one solution), multiple solutions (any of several objects = one solution) and multiple usages (one object = many solutions).

The “What's-the-Verb” syndrome.   In ‘Ballyhoo’, “whip lion” and “hit lion with whip” are inequivalent and only one of them does any taming. The following, from ‘Sorcerer’, can only be called a bug:

>unlock journal
(with the small key)
No spell would help with that!
>open journal
(with the small key)
The journal springs open.

(For a third example, the wording needed to use the brick in ‘Zork II’ is most unfair.) In many games the “examine”, “search” and “look inside” verbs all perform different actions, and it is easy to accidentally design a hidden treasure in such a way that only one of these will find it. (Though at the other extreme, excessive tolerance for verbs leads to everything being “moved”, not “pushed”, “pulled” or “rotated”.) Similarly, in the “What's-the-Noun” syndrome, an object stubbornly fails to respond to reasonable synonyms, such as “sword” for “gladius” or “football” for a soccer ball. But perhaps a remark on a sad subject might be intruded here. The Japanese woman near the start of ‘Trinity’ can be called “yellow” and “Jap”, terms with a grisly resonance. The game shows nothing but respect for her: should it allow the player to do otherwise?

Variety is valuable, but logic and fairness are paramount. “Is the writer pulling a rabbit out of a hat or do you see the fuzzy ears first?” (Dave Lebling). Gareth Rees suggests that one way to ensure that puzzles are consistent with the game containing them is to write a sample transcript of play first:

It stops me coding anything until I have a puzzle fairly well fleshed out in my mind. Too often it's tempting to start coding something one way and then discover that later developments need a different approach.
It makes me think like a player (I try to … include a selection of the silly things that I would be liable to type if I were playing the game …). Often when coding it becomes habit just to fail to deal with situations and responses that are tricky to write. Having them appear on the script forces me to say to myself, ‘It may be tough to code but it'll appear natural in the game and that's worth it’. I also find it hard to get into the habit of providing interesting responses to failed actions, and the script helps with this.

Another approach is to chain backwards from a goal, repeatedly asking “how can I obstruct this further?”, so that the plot line becomes, like a computer drawing of a fractal curve, more crinkly with each iteration. Peter Killworth, in his book How to Write Adventure Games for the BBC Microcomputer Model B and Acorn Electron (whose opening words are “Adventure games are like avocado pears”) describes an entire game this way (‘Roman’, 1984). Thus you need to pay a debt to a Senator, so you need to steal a bust from a temple, but that means impersonating a priest, by sacrificing a chicken with a gladius, which means catching a chicken (you scare it with a cat, but the cat must be attracted by a mouse, which you need to catch with a mousetrap): and the gladius isn't just lying around, either. You also need a torch, which

… needs soaking in oil first, just as candles need wax to burn. So we'd better organise a pool of oil through which the player can walk … When the player gets to a source of flame – how about a brazier of coals, which will have to be untakeable? – he can attempt to light his torch. It isn't oily, it burns to a stump well-nigh instantly … If it is oily, it'll catch fire … No, that's too simple. If a player is soaked in oil too, he'll probably catch fire too! … We'll create a damp, misty area, where the player is assassinated by a runaway slave, unless he enters while on fire. Then he will be safe, because the mist will condense on his body … So the poor player, staggering around and on fire, will try the mist, but to his disappointment the torch will go out permanently too! The solution is trivial – he must drop the torch before entering the mist.

(Killworth's games are not known for their qualms and the player, it will be noted, forfeits the game without any warning by lighting the unoily torch or exploring the misty area.) This kind of plotting, with puzzles strung together like beads onto a necklace, offers the considerable advantage of lending coherency. But it is also liable to make long, linear sequences of puzzles which must be completed in an exact order. “I've found it incredibly hard to keep the puzzles from leading the whole story” (David M. Baggett, 1994).

Mazes.   In the Traité des systèmes of 1749, Condillac wrote: “What could be more ridiculous than that men awakening from a profound sleep, and finding themselves in the middle of a labyrinth, should lay down general principles for discovering the way out?” Ridiculous, but very human, because the dogged exploration of a maze is dull indeed, repetitious and irritatingly drawn-out: far more enjoyment is to be found in the working-out of its general principles.

Be it clearly said: it is designers who like mazes (“Concocting such mazes is one of my delights” – Peter Killworth); players do not like mazes. In the original puzzle, a tangled set of rooms have indistinguishable descriptions so that mapping becomes impossible: the original solution is to make the rooms distinguishable again by littering them with objects as markers. This solution is easily obstructed (the ‘Advent’ pirate and the ‘Zork I’ thief wander around picking the objects up again), but this only makes the experience more tiresome. When David Baggett was asked “How do I make my maze so that it doesn't have the standard solution?”, his entire reply was: “Omit it.”

Nevertheless, like the writing of locked-room mysteries, the devising of new solutions for mazes – usually involving guides, or hidden signs, or ways to see – is a modest art form in its own right: novelty being the essential point, though it is equally important to signal to the player that a novel solution exists. The unique The Adventure Gamer's Manual (1992), by the Cornish vicar and eccentric writer Bob Redrup, devotes all of Chapters 7 and 8 to solving maze variants: a faintly weary tone is maintained throughout. Redrup was an aficionado of the Topologika, and thus of the formerly Cambridge University games, which are simply riddled with mazes. Here, in ‘Crobe’ (1986) by the indefatigable maze-maker Jonathan Partington, the player is evidently not expected to explore haphazardly:

You are in an underground marsh, a treacherous place where everything looks alike and water and slime lap around your feet. One false move would mean death, but you do at least have the choice of 8 horizontal directions to wander in.

Because lethal unless solved utterly, this is a benign sort of maze. Likewise, in ‘Kingdom of Hamil’ (1982):

You are in the Maze of Hamil. Light streams in through many gaps in the rocks. There is the constant sound of rockfalls, distant and not-so-distant.
There is a small nickel hexagon here, with the inscription “1 PFENTIME”.

Whatever is going on here, it doesn't look like a simple matter of dropping marker-objects. Partington also has a (thoroughly unfashionable) penchant for elements of randomness which a player can overcome with great difficulty by careful planning. The caryatid maze from ‘Fyleet’ (1985) makes the novel twist of imposing random obstacles on a determined layout, and in ‘The Quest for the Sangraal’ (1987): “There are exits in various directions, but, since the island is rotating, these directions change continually.” Infocom's output has its share of mazes, too, one per game: those in ‘Starcross’ and the ‘Enchanter’, ‘Sorcerer’, ‘Spellbreaker’ trilogy are the most satisfying.

Light source puzzles.   Almost as disliked, but offering a little more scope to the designer, is the “bringing light to darkness” puzzle. The two standards reduce to refilling a lamp with limited oil and bringing light to a dark room which can apparently only be reached by a player who hasn't got a light source. (‘Advent’ includes both. The lake and aqueduct areas of ‘Zork III’ have an elegant light puzzle, probably the best thing in an otherwise so-so game.) Darkness need not be a problem to be solved, though: it might be a fact of life. Though few games have tried this (but visit the secret passage in ‘Christminster’), a large permanently dark area might still be explored with the other senses.

Capacity and exhaustion puzzles.   Again, unpopular because their solution is normally tiresomely repetitive, forcing the player to keep putting things down and picking them up again. It can seem ridiculous that the protagonist can carry hundreds of bulky and fiddly things around all the time, so many designers impose a limit for realism's sake, typically of seven objects. It is bad form to set puzzles making life difficult because the limit is four and not five (after all, in emergency anyone can always carry one more item). In some games the limit is instead on total weight. Taking realism further, some games measure a state of health or even numerical levels of “strength” and “constitution” during play. The protagonist grows hungry and needs food, tired and needs sleep (in ‘Enchanter’ he is positively narcoleptic), wounded and needs recuperation. ‘Planetfall’ simulated a progressive illness whose symptoms are increasing need for food and sleep, and put many players off this kind of puzzle for life. Exhaustion rules are difficult to make fair. A rule requiring a return to an orchard for fruit should be watched carefully, as it will irritate a player to have to do this for a second, a third or a tenth time.

Timed puzzles.   Completing this round-up of unpopular but still sometimes justified puzzles are those which involve timed events, running along a script which requires the player to do something specific at one particular moment. In the prologue to ‘The Hitchhiker's Guide To The Galaxy’, why would any player buy the cheese sandwich in the pub and then feed it to the dog in the lane, on the one and only turn in which this is possible? Admittedly, an alternative exists later on, but this is not evident at the time. Mike Roberts (Usenet posting, 1999):

Aside from the annoyance, the reason I try to avoid timed puzzles is that they make you acutely aware that you're playing a game. As soon as I get into a save-try-restore loop, any sense of immersion is destroyed for me; I instead feel like I'm debugging a program, since I'm running through a series of inputs to figure out how the game responds.

This “sense of immersion” can partly be restored by keying events not to game turns but to the time of day, provided this fits the scenario, and making each stage last for a great many more turns than are strictly needed to solve the puzzles. Events which come only to he who waits are also problematic. In the Land of Shadow of ‘Zork III’, only the player who decides for some reason to wait on a fairly uninteresting Ledge will be rewarded with a visitor. (This case is defensible on grounds of context, but only just.)

Utility objects.   A designer who wants players to think of some items as useful needs to provide many situations – more than one, anyway – in which they can be used. A hallmark of better-designed games is that the player accumulates a few useful tools during play and wants to keep them to hand thereafter. (Cf. the crowbar and gloves in ‘The Lurking Horror’.)

Keys and doors.   Almost all games close off segments of the map on a temporary basis by putting them behind locked doors. Many variations on this theme are extant in the literature: coded messages on the door, illusory defences, gate-keepers, the key being in the lock on the wrong side and so on. More usually a locked door signals to the player that a different puzzle (i.e., finding the key) has to be solved before this one can be, so that a designer uses it to impose a chronology on events. Questions to ask here include: if there are people just inside, do they react when the protagonist knocks on the door, or tries to break it down or ram it? Can the door be opened, closed, locked or unlocked from both sides? Are there skeleton or master-keys capable of opening many different doors? Are the keys which do open different doors sufficiently distinctive in appearance? Roger Giner-Sorolla commented that keys are the most naked kind of Get-X-Use-X puzzle:

One can only find so many keys inside fishes' bellies, lost in the wainscotting, dropped at random in corridors, or hanging around guard dogs' necks before the artifice of the puzzle structure becomes painfully clear. By contrast, all six of the keys in ‘Christminster’ are hidden in places where one might actually keep a key, and all their locks are guarding places that one would expect to be locked; moreover, we end the game with a pretty clear idea of who normally uses each key and why.

Machinery and vehicles.   Machines are among the easiest puzzles to design: they have levers or ropes to pull, switches to press, cogs to turn. They need not make conversation or respond to anything beyond their function. They often require specialised tools, which brings in objects. They can transform the game in a semi-magical way; time travel or transforming coal to diamond being the clichés. They can also connect together different locations: chains, swinging arms and chutes may run across the map, and help to glue it together. Writing in the TADS manual, Mike Roberts makes the useful point that machines assist interactivity:

For example, you might design a machine, described as “a small metal box with a button, a short plastic hose on one side, and a large metal pipe on the other side.” When the button is pushed, “a loud hissing comes from the plastic hose for a moment, then a large drop of clear liquid drops out of the pipe, which hits the floor and quickly evaporates into a white cloud of vapor.” If the player puts the plastic hose in a glass of water and the button is pushed, “the water is sucked into the plastic tube, and few moments later a block of ice is dropped out of the pipe.” This allows the player to learn by experimentation what the machine does, which is more fun for the player than if you had labelled the machine “a freezer” or some such.

In machine puzzles, then, the player experiments with the controls and forms a theory of what is happening. With larger machines this involves visualising the physical construction of the components and how they affect each other: ‘Hollywood Hijinx’ is a tour de force of such puzzles, with a see-saw and a suspended safe. But the literature also includes highly complex self-contained machines presenting something of a black box whose internals must be deduced, such as a B-52 bomber and an Enigma cipher machine (‘Jigsaw’) and a computer which is programmable in the language Scheme (‘Lists and Lists’, Andrew Plotkin, 1996). Vehicles in games to date have included cars, tractors, fork-lift trucks, boats, hot-air balloons, log flumes, punts and elephant rides. Vehicles increase the realism of a landscape, by making it more than a set of rules about walking. They nevertheless need a little care to code: for instance, to disallow driving up ladders or through a narrow crevice.

Fire.   The elements all tangle up code but add to the illusion. Fire has many useful properties – it makes light, it destroys things, it can cause explosions and chemical reactions, it cooks food, it softens materials, it can be passed from one object to another – but in the end it spreads, whereas the game's understanding doesn't. If the player is allowed to carry a naked flame, then the game is likely to need a rule to tell it whether or not every other item is flammable, and so on.

Water.   In any location where water is available, players will try drinking, swimming, washing, diving. They will try to walk away with, indeed on, the water. Liquids make poor objects, because they need to be carried in some container yet can be poured from one to another, and because they are endlessly divisible. “Some water” can easily be made into “some water” and “some water”. If there's more than one liquid in the game, can they be mixed? Pouring liquid over something is likely to make a mess of it: yet why should it be impossible? And so on. The compromise solution is usually to have a bottle with a capacity of, say, 5 units of water, which can be refilled in any room where there is water and so that 1 unit is drunk at a time. The player who tries to pour water over most things is simply admonished and told not to. Implementing swimming, or being underwater, is a different order of difficulty again, and many games agree with ‘Parc’ (John Rennie, 1983) that “since you cannot free yourself, and since you are by nature an air-breathing mammal I'm afraid you drown!”. (Level 9's game ‘Adventure Quest’ is rare in containing a coherently worked out underwater section, though many games have the odd turn's-worth of diving. ‘Jinxter’ (Georgina Sinclair and Michael Bywater, 1987), also has an elaborate underwater section, with a seldom-discovered shipwreck to boot.) What happens to objects being held? Can the protagonist swim while wearing heavy clothes, or carrying many things? Is it possible to dive?

Air.   Smoke and fog can obscure the scene, but puzzles involving air are mainly about its absence. The lack of oxygen to breathe has featured in many games, not always through being underwater: ‘Zork I’ and ‘Sorcerer’ share a mine with poor air, ‘Starcross’ and ‘Trinity’ include locations in the vacuum of space. A scuba mask, space helmet or some other kind of breathing apparatus is called for. (Other gases simulated include helium, explosive hydrogen and laughing gas.) Can the protagonist speak, or eat, or listen, or taste while wearing this apparatus?

Earth.   Digging for buried treasure… the shovel can be found in just about every traditional-style game and a good many others which ought to know better besides. The problem is that the player may want to dig anywhere and everywhere, which the game will probably not want to implement: to dig may artificially create a new location, or a new map connection, or a new container – the hole left in the ground, that is. (The prologue to ‘Infidel’, though the least interesting part of the game from the point of view of playing, has a good implementation of digging through sand.)

Plants.   Vegetation fits into almost any landscape, and on the grounds of interactivity generally plays some part in the game, which is good for variety, because people deal differently with plants from machines and people. Undergrowth can be pulled away from something obscured, or useful plants picked. Trees and creeping plants ought to be climbable: players nearly always try.

Animals.   In many ways preferable to people, animals add a splash of colour, and what would the Garden of Eden have been without elephants, rabbits, leopards and guinea pigs? They move and behave in curious and obsessive ways, displaying what human characteristics they like but not needing to react to conversation or to show human curiosity or surprise at what happens. This makes them much easier to design, but it doesn't exempt them from characterisation. It's a little predictable to make the player feed an animal into obedience and then get it do something. (The bird in ‘Advent’ is nicely characterised, in that it is frightened by the rusty iron rod with a star on one end. ‘Trinity’ is positively overrun with animal life, with some critics having called its roadrunner the most important character.)

Monsters.   Many of the early adventure games included trolls, orcs and dragons, or else Frankenstein's Monster, Dracula and vampire bats: some, like ‘Zork I’, allow hack-and-slay combat in the style of a role-playing game like Dungeons and Dragons. Others, like the heavily infested and therefore somewhat repetitive ‘Murdac’ (Jonathan Partington, 1982), base all their puzzles on getting past or getting rid of things. “Getting past” occurs often because most monster puzzles are no more than doors with the decoration of slavering fangs. Even when monsters wander, they are generally dull because – being monsters – they have no unpredictable behaviour. Whereas the capacious underworld of the same author's ‘Kingdom of Hamil’ houses a baby hexapod (a what?) and a Conan Doyle-like Lost World of dinosaurs, which is much more the thing.

People.   So dawns the sixth day of creation: we have the mountains, rivers, plants and animals, but as yet no people. The nightmare of coding real characters is illustrated well by one of Dave Lebling's example bugs from ‘Suspect’ (1984):

>show corpse to michael
Michael doesn't appear interested.

The body is that of Veronica, Michael's wife. Objects representing people often take extensive code, perhaps five times that needed for even a complicated room, because they may need to react to events, follow instructions (“robot, go south”), wander around the map and make conversation of sorts (the woman selling bread-crumbs in ‘Trinity’, who plays only a minor role, can say over 50 different things). Games with strongly-defined protagonists tend to have a stronger supporting cast, too:

‘Christminster’ does an exceptionally good job of outlining Christabel's role as a woman by limiting her actions (she can't enter chapel bareheaded) and through … dialogue (the villains and the Master are condescending, while young Edward sees her as a confidante).

(Roger Giner-Sorolla.) What distinguishes a character from, say, a spanner is that it has an attitude to the protagonist. One model of this has the current attitude as a position in a “mood maze”, with different moods being like locations and stimuli applied by the protagonist being like directions:

Hungry feed Grateful

(Setting out in the “feed” direction from Suspicious leads nowhere, as the food is not accepted.) Such a person is no more than a plaything, entirely reactive and without memory, so most designers would want to conceal this fact by adding spontaneous or even random behaviour, startling the player from time to time. Or, of course, by simulating some memory of events. Adam Cadre's ‘Varicella’ (1999) handles conversation using around 450 flag variables to remember which questions have been asked and answered before. It also reformulates conversation. Here, the protagonist Varicella – as bad as the rest of them, Cadre having a rare gift for the amoral – encounters Miss Sierra, the King's mistress:

>ask sierra about king
“What can you tell me about the King?” you ask. ”You seem to have known him better than anyone else…”
Miss Sierra scowls at you. “That had better not be an attempt at a personal question,” she says. “If you're expecting a rhapsody about how he won my girlish heart, think again.…”
>ask sierra about king
“Is there anything else you care to say about the King?” you ask.

Elsewhere, “ask guard about rico” can come out as “Is Rico in?”, “ask queen about prince” as “How is Prince Charles?” and so on according to what would make sense in context. Such programming is exhausting but fruitful in a game placing great reliance on conversation.

Ropes and chains.   Are notoriously troublesome to implement consistently:

[Someone will] say “Well, I've got this rope… how do I do a rope? It can be in two rooms at once if you tie it to something and take the end with you, and can you tie things up with it and drag them around with you?”
Then we'll stop and think and say, “You don't want to have a rope in your game,” and that makes it much easier for the new writers, you see.
My new game [‘The Lurking Horror’] has a chain in it, and it's even worse than a rope in almost every respect you can imagine and it's caused me no end of horror… the number of bugs that have come in on this chain alone would stack from here to there and back again.

(Dave Lebling again. But the chain puzzle in ‘Lurking’ is a masterstroke.) ‘The Meteor, The Stone and a Long Glass of Sherbet’ has a rope solving several puzzles whose source code runs to 300 lines of Inform, which is more than the whole “lily pond” region took: put another way, 5% of the entire code is occupied describing the rope. There is also a long ladder which is nearly as bad, and which the player is not allowed to tie the rope to.

Riddles.   Numerous games (‘Beyond Zork’, ‘The Path to Fortune’ (Jeff Cassidy and C. E. Forman, 1995)) include sphinxes or talking doors which pose riddles to passing strangers, and the writing of good riddles is an art form in itself. But who put these puzzle-obsessed doorkeepers there, and why? The knock-knock joke door in Irene Callaci's ‘Mother Loose’ (1998) sits much more happily, as the game is a wry mingling of nursery-rhyme stories for her six-year-old granddaughter (and indeed for the rest of us).

Decipherment.   Perhaps the most abstract and, if done well, the most satisfying of puzzles are those which present a system of coded messages, clues to the meaning of which are scattered across the game. ‘Infidel’ has hieroglyphics. ‘Edifice’ (Lucian Smith, 1997) requires the player to learn the language of Nalian, a puzzle which won considerable plaudits from players. But there are non-linguistic decipherments, too: in a sense the map of ‘Spellbreaker’ is itself a cipher. On a smaller scale, several Cambridge University games contain tricky cipher puzzles not unrelated to recreational mathematics. ‘Avon’, for instance, has a substitution code which is insoluble, but which it is possible to make partial deductions about: just enough to solve the problem at hand.

Clues.   At least in one view of interactive fiction, clues are essential and the principle should be that an ideally perceptive player could win on his or her first attempt, without recourse to saved games: in particular, without knowledge of past lives or of future events. (The exact opposite, in fact, of what ‘Brand X’ does. The player begins in a shop containing an aqualung, a cushion, a bunch of keys, a piece of sausage, a teabag and a sign declaring that “only two implements may be removed from this shop under penalty of death, so choose carefully”.) Here are three clues which did not carry:

  1. In ‘Dungeon Adventure’, a pride of lions is carved over a doorway. Any player walking through falls into a lethal pit. Did you miss the clue?
  2. The diamond maze in ‘Zork II’ is almost impossible to fathom unless (or even if) you are familiar with a certain multiple-innings team sport played only in America. In the words of even its designer: “always annoyed me… pretty lame.”
  3. Almost every player of ‘Advent’ has considered the rock marked Y2 to be a decoy, emblematic of the mysterious cave. But it was meant as a clue: on the cave maps used by Will Crowther's group, “Y2” denoted a secondary cave entrance, which in a certain sense is what this location is.

(1) is a bad pun, (2) an unconscious assumption and (3) an in-joke. Games that are entirely in-jokes, like the subgenre of college campus simulations (‘The Lurking Horror’ features MIT, Infocom's alma mater) are at least deliberately so, but it is all too easy for designers to include familiar objects from their own lives into any game, and unconsciously assume that the player shares this familiarity. When that familiarity is needed to solve a puzzle, the game may become unplayable.

Luck and accidental solutions.   Small chance variations add to the fun, but only small ones. The thief in ‘Zork I’ seems to me to be just about right in this respect, and similarly the spinning room in ‘Zork II’, but a ten-ton weight which falls down and kills the player at a certain point in half of all games would simply irritate. A particular danger occurs with low-probability events, one or a combination of which might destroy the player's chances. For instance, in the earliest edition of ‘Adventureland’, the bees have an 8% chance of suffocation each turn carried in the bottle: one needs to carry them for 10 or 11 turns, giving the bees only a 40% chance of surviving to their destination.

Even in a puzzle with no element of luck, many problems are solved by accident or trial and error. (The notorious Bank of Zork puzzle in ‘Zork II’ has been understood by almost nobody who solved it.) This is unsatisfying for both player and designer, and some games take steps to try to avoid it. The gold-assaying puzzle in ‘Spellbreaker’ is such that a shrewd strategy will always succeed, but in principle even a random strategy might succeed. The game rigs the odds to ensure that it never does.

Optional, partial and multiple solutions.   Most designers like to give two or more different solutions to a few puzzles in a game: it seems more real, it means that even a winning player hasn't found all of the game's secrets and it makes a difficult puzzle easier. (There are seven ways to open the child-proof medicine bottle in ‘Curses’.) Multiple solutions to the same puzzle need to be equally valid. The designer should not think of one solution as the “real” one, or allow another “short cut” one to skip critical plot events – this would short-change the player. On the other hand the designer must be relaxed about the inevitability that some part of his golden prose will never be seen whichever path the player takes. Most additional solutions are added in play-testing, but here is Brian Moriarty on ‘Wishbringer’ (1985):

Most of the problems in the story have two or more solutions. The easy way out is to use Wishbringer. If a beginner gets frustrated, he can whip out the magic stone, mumble a wish and keep on playing. Experienced players can search for one of the logical solutions – a bit harder, perhaps, but more satisfying. It's possible to complete the story without using any of the stone's seven wishes. In fact, that's the only way to earn the full 100 points.
The puzzles are highly interconnected. Once you start wishing your problems away, it's very hard to continue playing without relying more and more on the magic stone. The impotence of idle wishing – that's the moral of ‘Wishbringer’. All really good stories have a moral.

Analogous perhaps to the Wishbringer stone, ‘Enchanter’ has a one-use-only anti-magic spell. Although this solves one in particular of the more difficult puzzles, to use it up so early forfeits the game, since it is needed later. If you do fall into this trap, one of the ingenious dream sequences offers an oblique warning:

You dream of climbing in an unfamiliar place. You seem to climb forever, beyond reason. A fleeting hope arises in you, and you search furiously in your spell book and possessions for something. After a moment, you become frantic as you realize that you don't have it! You bolt awake in a cold sweat.

Rewards.   What reward for solving a puzzle? One is obvious: the game state advances a little towards its completion. But the player at the keyboard needs a reward as well: that the game should offer something new to look at. The white cubes in ‘Spellbreaker’, with the power to teleport the protagonist to new areas, are far more alluring than, say, the “platinum pyramid” of ‘Advent’, which is only a noun with a few points attached and opens up no further map.

“[A puzzle] should be logical, according to the logic of the game's universe. In a fantasy game, a puzzle can rely on magic, but the magic must be consistent throughout the game. A puzzle should be original in some way, not just a rehash of an earlier puzzle with different objects.” (Steve Meretzky). “My basic principle of designing puzzles is that the player should always know what he's trying to accomplish. Metaphorically, a player should always be able to find a locked door before he finds the key” (Mike Roberts in the TADS manual). “In all cases, after a particularly arduous puzzle, reward the player with a few simpler ones” (C. E. Forman, XYZZYnews 1). “Err on the side of easy. (He said, waiting to be struck dead for hypocrisy.)” (Andrew Plotkin). “There's definitely a difference between ‘satisfying’ and ‘pertinent’.” (Lucian Smith. These last quotations from the round-table discussion on puzzles in XYZZYnews 14.)   For more on the invented language in ‘Edifice’, see ‘Parlez-Vous Nalian’ in XYZZYnews 16.

  Pride comes before a fall.

  Several Cambridge University games, written by mathematicians, refer to “J4”: ‘Acheton’, for instance, has a “J4 room” rather like the “Y2 rock room”. The then-recent construction of the previously only hypothetical group J4 had completed the classification theorem for finite simple groups, and was a departmental triumph. The sign “∃J4” remained above a doorframe for some years.