In the early 1980s, the best Pac-Man players in the world were not the fastest. They were the ones with the best memory, because the original arcade game is deterministic - and a deterministic game can be beaten by a script.

The ghosts follow fixed targeting rules. They do not roll dice. So if you make exactly the same movements, at exactly the same times, the ghosts respond exactly the same way, every single time. Learn one route that clears a level and you can repeat it indefinitely.

Those routes are called patterns, and for a decade they were the entire competitive scene.

Why patterns work

The determinism runs deeper than it first appears. Ghost movement depends on your position, your direction, the global scatter-chase timer, and in Inky's case on another ghost's position. Every one of those is a function of what has happened so far, and nothing consults a random number generator during normal pursuit.

The single exception is frightened mode. After you eat a power pellet the ghosts do choose turns at random - which is precisely why pattern players avoid eating ghosts at certain points, or eat them in a strictly controlled way. Randomness anywhere in the sequence breaks everything downstream.

So a pattern is not just a route. It is a route plus a policy about which pellets to take and which ghosts to eat, designed to keep the whole level inside the deterministic part of the system.

The arcade classic and this version

Worth saying plainly: the patterns that circulated on photocopied sheets in 1982 were specific to the original arcade ROM. Any modern implementation, including the modes here, is a reinterpretation - the maze and the spirit are the same, the exact frame timings and targeting offsets are not necessarily identical.

What carries over completely is the principle. A game built on algorithmic pursuit rewards recognition rather than reaction. You are learning what configurations mean and what they become, not developing faster reflexes.

The practical version of that, on any implementation: play the same maze repeatedly and pay attention to the moments that recur. The scatter windows arrive on a schedule. The ghosts reverse direction together when the mode flips. The tunnels slow pursuers down. None of that requires frame-perfect knowledge to exploit.

The kill screen

The most famous consequence of determinism is the bug at the end.

The original game stores the level number in a single byte, and uses it to decide how many pieces of bonus fruit to draw along the bottom of the screen. At level 256 the counter overflows, the fruit-drawing routine runs wild, and it corrupts the right-hand half of the maze into a mess of garbled characters and letters.

The left half remains playable. The right half is unreadable, and critically the dots there cannot all be eaten, so the level can never be completed. The game simply cannot continue past that point.

That is the famous kill screen, and reaching it - playing a perfect game, clearing 255 levels and eating every dot, fruit and ghost along the way for a maximum of 3,333,360 points - was first accomplished in 1999 by Billy Mitchell, and has been repeated by a handful of players since.

It is a strange kind of ending: not a designed climax but an arithmetic accident, discovered by people who were better at the game than its authors expected anyone to be.

What the kill screen says about games generally

The interesting part is not the bug. It is that the game had a hard ceiling nobody knew about for nineteen years, because reaching it required a level of play the designers never contemplated.

Games with fully deterministic rules tend to accumulate this kind of discovery, because sufficiently determined players eventually exhaust the space. Draughts was proven a draw under perfect play in 2007 after roughly eighteen years of computation - a result that similarly sat outside anything the game's players could have established unaided. Play Checkers and you are playing a game whose final answer is now known, which changes nothing about the experience and quite a lot about how you think of it.

The patience games had their own version: players worked through all 32,000 numbered deals in the Windows version and established that exactly one of them is unwinnable. Anyone who plays FreeCell knows this: that collective exhaustion is the reason you can be confident the deal in front of you is solvable.

Practical pattern play

You do not need frame-perfect routes to benefit from this way of thinking. Three habits:

  • Learn one safe opening. The first thirty seconds of a level are the most predictable. Having a fixed route for that portion removes a large share of early deaths.
  • Clear the dangerous areas during scatter. When all four ghosts turn simultaneously toward the corners, that is a scheduled event, not luck. Use it for the parts of the maze with poor escape routes.
  • Save the power pellets for clustered ghosts. Chain eating is worth 3,000 points; a panicked single is worth 200. Deliberately luring them together before taking a pellet is the largest single scoring decision in the game.

That last one is the same allocation logic every scoring game has - convert a resource when the conversion rate is high rather than when you are uncomfortable. It is the decision behind chasing the bonus rather than banking a safe box in Yahtzee, and the answer is the same: the payoff structure, not your nerves, should decide.

Where to practise

The classic maze is the baseline. Turbo raises the speed, which is the wrong place to learn patterns and the right place to test them. Chill slows things down enough to actually observe the scatter schedule. Survival removes the level structure and turns it into an endurance test where pattern knowledge matters most. Mini shortens the maze so you can run the same route many times quickly, which is genuinely the most efficient way to build recognition.

Run the experiment

Play Mini five times using deliberately the same opening route and note what differs. Then play Chill and time the gaps between the ghosts' synchronised direction reversals. Those two exercises will tell you more about how the game works than an hour of ordinary play.

The rules hub covers the modes, the history FAQ covers the 1980 original, the luck-versus-skill FAQ is a direct companion to this piece, and the glossary names the terms.

For what each of the four ghosts is actually trying to do, see the companion piece.