Playing TETERI

The CPU Opponent and Its Three Difficulties

How the computer player decides where to put things, and what actually changes between rookie and nightmare.

The computer opponent is a heuristic search rather than a scripted difficulty curve, which means its three settings differ in genuinely different ways.

How it thinks

For every piece it considers every rotation in every column, for both the current piece and the held one. Each candidate placement is simulated: the piece is dropped, complete rows are removed, and the resulting board is scored.

The score weighs four things: total stack height, the number of covered holes, how uneven the surface is, and how deep the wells are. Holes are weighted heavily, because they are the most expensive object in the game.

It then plays the best-scoring placement, one input at a time, so you can watch it move rather than seeing pieces teleport.

What changes between difficulties

Rookie thinks slowly, acts slowly, does not use hold, and deliberately picks a worse-than-best placement about a third of the time.

Pro thinks quickly, uses hold, and blunders occasionally.

Nightmare thinks almost instantly, uses hold, and never deliberately blunders.

Notice that the difficulty is not a speed multiplier applied to a single behaviour. A rookie CPU genuinely makes worse decisions, which is why it is beatable in a way that a fast-but-perfect opponent would not be.

What it does not do

It has no multi-piece plan, so it does not build T-spin setups or openers. It does not look at your board at all. It does not choose when to attack.

Those gaps are where a human beats it: by out-damaging it with setups it never attempts, rather than out-stacking it, which is a contest you will lose.

How it holds up

Given garbage, nightmare digs out efficiently and keeps playing. In testing it clears lines steadily for tens of thousands of frames without topping out on its own. If you beat it, you beat it on attack.

PLAY TETERI ALL ARTICLES