AI Framework

Two-player games are fun, but sometimes you want to play solo. This unit adds an AI opponent — a computer player that takes Player 2’s turns automatically.

The AI starts simple: it picks a random empty cell. No strategy, no intelligence, just legal moves. But this framework is the foundation for smarter opponents in later units.

Run It

pasmonext --sna inkwar.asm inkwar.sna

The title screen now offers two options: “1 - TWO PLAYER” or “2 - VS COMPUTER”. Select vs Computer and watch the AI claim cells automatically after each of your moves.

Game Mode Selection

The title screen needs to distinguish between two modes:

We add a game_mode variable and an ai_timer for the delay before AI moves. The delay makes the game feel more natural — instant AI moves would be jarring.

Reading the Mode Selection

The ZX Spectrum keyboard is read in half-rows. Keys 1-5 share row $F7:

The BIT instruction tests individual bits. Bit 0 is key 1, bit 1 is key 2. When pressed, the bit reads as 0 (active low).

AI Turn Handling

The game loop checks if it’s the AI’s turn:

Three conditions for AI to move:

1. Game mode is VS_AI
2. Current player is Player 2
3. AI timer has expired

The timer creates a half-second pause before each AI move, giving the human time to see what’s happening.

Random Number Generation

The Z80 has a built-in source of randomness — the R register:

The R register increments with every instruction fetch. We mix it with a seed value using addition, rotation, and XOR to produce varied results. Not cryptographically secure, but fine for a game.

Finding an Empty Cell

The AI needs to find a valid move. We start at a random position and scan until we find an empty cell:

Starting at a random position adds variety — without it, the AI would always claim the first available cell. The wrap-around ensures we check every cell even if we start near the end.

Why Random AI First?

Random AI is the simplest possible opponent:

• Always makes legal moves
• Requires minimal code
• Easy to test and debug
• Provides a baseline to improve upon

In upcoming units, we’ll make the AI smarter by prioritising adjacent cells, defending against the human, and targeting strategic positions.

The Complete Code

Try This: Faster AI

AI_DELAY    equ     10              ; Shorter delay (~0.2 sec)

Try This: Instant AI

AI_DELAY    equ     1               ; Nearly instant

Be warned — instant AI feels unnatural and can be disorienting.

What You’ve Learnt

• Game modes — Single variable controls two-player vs AI behaviour
• Keyboard row reading — Individual key detection with BIT instruction
• Random numbers — Using R register for pseudo-randomness
• AI framework — Turn detection, delay timer, move execution

What’s Next

In Unit 10, we’ll make the AI smarter by having it prioritise cells adjacent to its existing territory.

What Changed