Today, I was playing the Pokemon TCG for the GBC on my GBSP. An amusing thing happened, the game started with both of us benchless and both of us had the exact same Mew as our active Pokemon. That was amusing.
To continue the story I should describe that Mew’s attack does 10 damage multiplied by the number of energies on the active Pokemon, and this requires on energy. The game then went the following way:
I started, but did nowt.
The NPC put an energy on his Mew and attacked me for zero damage.
I put an energy and attacked the NPC Mew for 20 damage (due to weakness).
The NPC used a trainer card to remove the energy on my Pokemon. Put an additional energy on its own. Then attack me again for zero damage.
I then attacked again knocking out my opponent’s only Pokemon and winning the game.
Quite possibly, you had to be there. Quite, possibly I only find it so amusing since it is personal to me. The key thing is that as AI gets more sophisticated that can’t happen; a series of usually good moves that created a mess for itself. The game designer could try and replicate this effect with a more sophisticated AI, but – as I’ve mentioned previously – that feels fake. I want serendipity not stage management.
It behooves me to mention that the idiotic AI is not the sufficient condition for these things. What it requires i for the game to be complex – although simple enough that you can hold the AI in contempt for not understanding it – including these odd special cases. An example of this thing with the idiotic player would be my repeatedly using an attack that devolved the opponent’s active pokemon getting annoyed when the opponent immediate reevolved it, not having noticed the move returned the relevant cards to the player’s hand rather than discard card.
So when I say the idiotic AI, I mean an AI that doesn’t fully understand all the complexities of the game that it plays, relying on generally good heuristics. However, the capabilities of AIs to understand complexity is going up. We remain generally static.