Cauldron II: The Bouncing Pumpkin's Cruelty

Almost every platform game ever made begins by giving you a body that obeys you. The contract is so old it is invisible: press left, go left; press jump, leave the ground; release, land. Everything the genre does afterwards — the enemies, the gaps, the timing — is built on top of that obedience, and the difficulty comes from what the designer puts in front of you.

Palace Software’s Cauldron II, released for the Commodore 64 in 1986, tears the contract up on the title screen. You are a pumpkin. Pumpkins cannot walk. What a pumpkin can do is bounce, and so bouncing is the whole of your vocabulary, and the hardest thing in the game is crossing an empty room.

The inversion

Steve Brown had made the first Cauldron the year before, and it was a decent horror-flavoured thing in which you played a witch, flew a broomstick over a side-scrolling landscape, went underground into flick-screen caverns to gather what you needed, and eventually dealt with the Pumpking. It sold. It reviewed well. It was, structurally, a game of its moment.

Cauldron II: The Pumpkin Strikes Back does the thing sequels almost never do. It keeps the world and swaps the chairs. You are the pumpkin now, rolling into the witch’s castle with revenge on your mind, and the witch — your player character twelve months ago — is the boss. That subtitle is doing a lot of work and it knows exactly what it is doing.

The joke would be enough for a marketing line. Brown went further and made the inversion mechanical. A witch has hands, a broomstick and a spell book. A pumpkin has none of those things. It has mass and it has a floor to hit. So the sequel’s design problem was never “what does the pumpkin do” — it was “what happens to a platform game when you delete the legs”, and the answer turned out to be a genre Palace effectively had to invent on the spot.

Locomotion as the antagonist

Here is what it feels like to move.

Your pumpkin bounces constantly, on its own, whether you like it or not. The joystick lets you push it left or right, and the push takes effect over time, so you accumulate horizontal drift rather than choosing a direction. You can bounce higher. You cannot stop. You cannot stand still and think. Every screen you enter, you enter already in motion at a speed you set several seconds ago in a different room, and the momentum you brought with you is now a fact you have to solve.

That is a physics game wearing a platformer’s costume, and the closest relative on the machine is Thrust, which came out the same year from Firebird and asked precisely the same question: what if the player’s input governed acceleration and everything else followed? Jeremy Smith’s answer was a spaceship in a gravity well, and it reads as fair because a spaceship is supposed to be hard to fly. Brown’s answer is a vegetable in a corridor, and it reads as an insult, which is better comedy and considerably worse for your blood pressure.

Wizball works the third variant of the trick — the ball that rolls when you want it to stop is the entire opening of that game, and Sensible’s joke is that you buy your way out of the handicap with the power-ups you earn. Brown offers no such mercy. The pumpkin bounces on screen one and it is still bouncing on the last screen you will ever see.

The cruelty, itemised

The castle is a flick-screen maze: each room is a static screen, and crossing an edge redraws the next one whole. Palace took a technical shortcut and turned it into a weapon. Because the screen flicks, you cannot see what is coming. Because you cannot stop, you cannot approach an unknown room slowly. Therefore every new room is entered blind, at speed, mid-bounce, and the room may contain something that kills you on contact positioned exactly where a pumpkin arriving at that velocity would be.

Learn the room and it becomes trivial. That is the loop: die, memorise, die, memorise, until the castle is a sequence of bounces you perform from muscle memory. The game is asking you to write a program in your hands and then execute it perfectly for twenty minutes. One badly judged push near the start and every subsequent room arrives at the wrong height.

Then there is the geometry. Ceilings are the real enemy. A corridor with a low roof will not let you bounce high enough to clear the gap at the end of it, so you have to arrive with the right combination of height and drift, which was determined two rooms ago. Water kills you. Contact with the castle’s inhabitants drains an energy bar that you cannot easily refill. And you are collecting things — scattered items to be gathered and carried to where they do their work — which means you are backtracking through the same murderous rooms with your patience already spent.

Zzap!64’s reviewers of the day liked it and said it was brutal, and the brutality was the point of the coverage. It is one of the small number of games from the era that people describe with genuine grievance forty years on. That is a design achievement of a sort. Nobody bears a grudge against a game they merely lost.

The house that Palace built

Palace Software is one of the more revealing British studios of the period, and Cauldron II sits exactly in the middle of its short, strange run. The company came out of the Palace Pictures orbit — a film distributor with a taste for horror and the arthouse — and the games carry the fingerprint. The label’s identity was drawn art, a slightly nasty sense of humour, and a willingness to let a designer’s personality show through the pixels. Antiriad arrived the same year with Dan Malone’s comic-book presentation attached. Barbarian followed in 1987 and became infamous for its decapitation animation and a Page Three model on the cover, which tells you how the British industry marketed itself and how young its audience actually was.

Brown was the through-line on the Cauldron games and he was a graphic artist making design decisions, which is the interesting part. An artist-led game from 1986 tends to look extraordinary and play like a slideshow. Cauldron II looks extraordinary and plays like a mathematics problem, and that combination is rare enough that I can count the era’s examples on one hand. Whoever worked out that a bouncing sprite could carry a whole game was thinking about systems, whatever their job title said.

The comparison I keep reaching for is Head Over Heels, which arrived the following year with its own answer to the same question: give the player a body that is deliberately incomplete and make the game about the gap. Ritman and Drummond split you into two characters with half the abilities each. Brown gave you a vegetable. Both are British, both are cruel, and both understand that the most interesting constraint you can hand a player is the one attached to their own hands.

Is it fair?

I want to argue this properly, because “hard” and “unfair” get used interchangeably and they are different accusations.

Cauldron II is fair in the way that matters: the physics are deterministic. The same push at the same moment produces the same arc every time. Nothing is random. Every death is legible, in the sense that you can reconstruct exactly which decision three rooms ago put you into that spike. A game with those properties is teachable, and teachable is the whole of the defence.

Where it fails is the cost of the lesson. The flick-screen means the first encounter with any hazard is unavoidably lethal, and the checkpointing is merciless, so the price of learning room forty-five is replaying rooms one through forty-four with your hands doing the same memorised sequence they did an hour ago. That is attrition wearing difficulty’s coat. Modern design solved this — the reason Dead Cells can be savage without being resented is that it makes the retry cheap and the run different — and the solution was available in 1986 in the form of a generous respawn. Palace chose the grudge instead.

And the graphics carry it further than they should. Brown was an artist first, and the castle is beautifully drawn: the sprite work is crisp, the pumpkin has a frankly excellent face, the rooms have a mordant Hammer-horror wit to them that makes each new one a small reward for surviving the last. It is a lovely thing to be killed by, which is more than most of its contemporaries could offer.

Where to play it, and how to survive it

The C64 version is the one; Spectrum and Amstrad conversions exist and the C64 has the visuals. It emulates trivially and it is short in the way that early British games are short — the content would take a competent player under half an hour, and it will take you weeks.

My honest advice is to use save states, and I say that as someone who normally regards them as cheating yourself out of the game. Cauldron II’s actual design — the momentum puzzle, the room-shaped problem, the arc you have to earn — is excellent and lives in the individual screen. The bit that has aged into pure attrition is the replay. Save at each new room, learn the bounce, move on, and what you are left with is one of the sharpest pieces of physical design on the machine and one of the funniest revenge stories anyone put on a cassette.

The pumpkin, after all, has been waiting a year.