Why AI hasn't conquered Puyo Puyo yet

‌
1. All Puyo fall randomly
- It is difficult to design long-term chains, making it impossible to create a winning formula.
- A guaranteed winning formula cannot exist.
- Repetitive training becomes meaningless.
2. It is difficult for AI to make judgments.
- It is hard to decide whether to trigger a 4-chain immediately to beat the opponent or save up for a 10-chain later.
- While preparing for a 10-chain, the board state can change instantly if the opponent creates garbage Puyo.
- The AI must also judge if the opponent is bluffing—i.e., acting as if they are about to trigger a chain even when they haven't finished building it.
3. You must respond to the opponent's tactics on a daily basis.
- Before the opponent can build a large chain, you must disrupt their stacking and ruin their timing with small 2-3 chains.
- You must decide in real-time whether to trigger a chain first or respond with an even larger chain.
4. In intellectual disciplines like Chess, Go, coding, or law, AI overwhelmingly crushes humans,
and while AI can actually achieve chains very quickly, winning against an opponent is a completely different matter.
To this day, an AI trained in Puyo Puyo cannot beat even a casual expert playing on a mobile phone, let alone a Japanese national-level player.
Tetris has already been conquered by AI, but Puyo Puyo remains in the human domain for now. lol