The Chaos Engine: The Bitmaps' Steampunk Run-and-Gun

The premise is one paragraph in the manual and it is better than most games manage in forty hours. A Victorian gentleman-scientist called Baron Fortesque builds a machine that meddles with time and space, the machine gets away from him, and the English countryside starts producing things that should not exist. Six mercenaries take the contract to walk into it and turn the thing off.

That is the entire fiction of The Chaos Engine, and the Bitmap Brothers were disciplined enough to stop there. Everything else — 1993, Amiga first, Renegade publishing — is a run-and-gun with two players walking through hostile scenery, and the interesting part is how much design is packed into a genre that was, by 1993, mostly exhausted.

The mercenaries are the difficulty setting

You pick two of six: the Gentleman, the Brigand, the Navvie, the Mercenary, the Preacher, the Thug. In a two-player game you take one each. In solo you pick both and the computer drives the second.

They are genuinely different, and the differences are structural rather than cosmetic. The Gentleman has a long-range weapon and poor durability, which makes him a player who must read the room before entering it. The Thug is fast and fragile and shoots a short distance, which makes him a player who solves problems by arriving before the problem has finished loading. The Navvie is slow and hard to kill. Each carries a distinct special ability and each upgrades along its own path, so the character you pick determines what kind of level this is going to be well before you reach it.

This matters more than the usual character-select shrug because The Chaos Engine is stingy. Your upgrade currency comes out of the levels themselves and there is never enough of it, so committing to a mercenary is committing to a build for the whole run. Pick the Brigand and you are playing a game about spacing. Pick the Thug and you are playing a game about panic. Same maps, different sport.

And in a rare bit of design generosity, the upgrade screen lets you spend your money on your partner. In two-player that produces an argument at the end of every level, which is clearly the point. In solo it produces a real decision: your AI companion is a resource, and a resource you have invested in is one that survives.

The AI partner is the achievement

Here is the thing The Chaos Engine deserves to be remembered for, and it gets buried under the steampunk chrome every time.

Co-op games with a solo mode almost always solve the partner problem by lying. Either the second character vanishes and the game is quietly rebalanced, or the partner exists as a following object that shoots occasionally and functions as a walking morale hazard. The Bitmaps wrote an AI that plays the game. It moves at a sensible distance, it engages what it can see, it does not wander into water, it picks up what it walks over, and it hits things. It contributes.

Getting this right in 1993 on a 68000 is a nontrivial engineering result, and the reason it works is that the Bitmaps kept the problem small. The partner does not need to solve the level. Level progress is entirely yours — the switches, the routes, the decisions. The AI’s job is combat presence and mobile firepower, and that is a task narrow enough to write competently in the CPU budget left over after the sprites. The design and the constraint agree with each other, which is what good work under a tight machine looks like.

The consequence is that the solo campaign is the real campaign. You are not playing a reduced version. You have a partner who can be levelled up, who dies if you are careless, and whose loss you feel in the next firefight, and the game becomes quietly about protecting an asset you paid for. That emotional loop happens by accident of the economy, and it is the single most modern-feeling thing in the game.

Nodes: the level structure that stops it being a corridor

The Chaos Engine runs across four worlds of four levels each, and each level is built on the same skeleton: find and activate the node terminals scattered around the map, which opens the gate that lets you leave.

That sounds like a fetch quest and it is not one, for a reason worth spelling out. The nodes gate the exit, and the map is laid out so that finding them requires you to cover ground you would otherwise skip. A pure run-and-gun is a conveyor belt — the level pushes you rightward, you shoot what appears, you arrive. The node rule turns the map into a space with a shape. You have to explore it, which means you have to hold territory you have already cleared, which means enemy spawn points become a strategic concern rather than an inconvenience, and it means the ten per cent of the level you would have run past has your money in it.

The Bitmaps then price that exploration honestly. Every crate and hidden alcove holds currency, and currency is upgrades, and upgrades are survival two worlds from now. So the game gives you a real choice on every map: leave now with what you have, or sweep it properly and pay for the sweep in ammunition and health. That is a risk-reward loop with actual teeth, and it is exactly the same kind of loop the Bitmaps built into Speedball 2 three years earlier — one legible currency, one scarce thing to spend it on, and a scoring table that tells you what the designers consider worth doing. The house style is an economy, and everything else is dressing on top of it.

The maps themselves are dense in a way that rewards a second look. Enemy generators sit at chokepoints. Water and walls carve the space into pockets that favour some mercenaries and punish others. Richard Joseph’s soundtrack sits underneath doing mood rather than melody, and the sprite work is peak Bitmap — brass, rivets, mud, and that refusal of any colour that might cheer you up.

Why the shooting feels right

Underneath the structure there is a moment-to-moment game, and it holds up for reasons worth naming because plenty of its contemporaries got this wrong with the same hardware.

Your mercenary fires in eight directions and moves in eight directions, and crucially the two are locked together — you shoot where you walk. That constraint is doing enormous work. It means every engagement is a positioning problem rather than an aiming problem, so the skill the game asks for is reading a room and choosing where to stand in it. A twin-stick scheme would have made the same maps trivial, because you could retreat and cover simultaneously. Locking the axes forces commitment: to shoot that generator you must face it, and facing it means turning your back on whatever is behind you.

Enemies then exploit exactly that. They come from generators rather than from a script, so they arrive continuously and from the direction you are least able to answer, and the correct play is almost always to kill the generator instead of the enemies pouring out of it. That is a lesson the game teaches by attrition in the first world and expects you to have learned by the second. It is a clean bit of instruction: the game never says it, the ammunition counter says it for them.

The weapons obey the same economy as everything else. Upgrades widen your spread, extend your range, add rate of fire — visible, immediate, permanent for the run. There is no weapon wheel and nothing to manage. You buy the thing, the thing changes how the next map plays, and you carry it forward. Every unit of complexity in The Chaos Engine earns its place by changing a decision, and anything that would only have added texture was left out.

What it gets wrong, and where it sits

The honest criticism is that The Chaos Engine is repetitive across its length. Four worlds of four levels is a lot of the same verbs, and the fourth world does not introduce an idea so much as raise the numbers. The final act leans on volume where earlier levels leaned on layout. If you play it in one sitting the last hour blurs, and that is a real structural fault rather than a period charm.

The password system between levels is the other tell of the era: progress persists, your build persists, and the game assumes a session rather than a campaign. Modern players expect the run to be the unit. This one expects the evening to be.

Set against its contemporaries, though, the discipline is startling. 1993 on the Amiga was still full of games that spent the machine on spectacle and left nothing underneath — the Shadow of the Beast approach, where the demo is the product. The Chaos Engine spends the machine on a rules system that holds up. Its descendants are everywhere in the twin-stick and roguelike space: pick a character whose stats define your whole run, sweep a map for currency, upgrade between stages, repeat. That template got a name twenty years later. The Bitmaps shipped it in 1993 with a Victorian coat on.

Where to play: the Amiga original is the reference, and the Mega Drive and SNES conversions are competent if you can live with a different palette. The 2013 remaster exists and is faithful to a fault. Bring a second player if you can — the AI is good enough that solo is legitimate, which is the compliment, and a human is still better, which is the point of the design.

The sequel in 1996 turned the two players against each other, which is a fascinating idea and a comprehensive misreading of why the first one worked. The whole of The Chaos Engine is two mercenaries who cannot cover the same ground alone. Taking that away leaves the brass and the mud and nothing to do in them.