Thank for making a version of TAS of this game mode that I call complete, including the most updated strategies! And thanks for giving me credit for the strategy used in Goomba board. The main reason I wanted to create my TAS of this was implementing that specific strategy of having one fewer CPU turn. I specifically stopped before Shy Guy's board because there I didn't find any significant improvements since that point, and it's not worth me finishing it. I also just wanted to see if I could get it to generate a King Boo map for ideal conditions for TAS but that's the best I could get. I kinda just found the first board (also manually explored) that has a star within 3 rooms and has a DK space, and stopped right there. You explored many more boards than I did, so that rewarded you with a better layout. The biggest challenge to further improving this game mode would definitely be trying to reconstruct how a King Boo map is generated. I actually have the RNG address (idk how I found it or obtain it from a source) which 802ce150 , I also tried to find out how the board is laid out in memory, but it's too complicated for me. The space index for a board can represent multiple spaces of different rooms on the board?? It started to become nonsense.
Me from the following day:
I just observed the RNG algorithm being 2^32 - (x*16807) mod (2^31-1) , but the game sometimes return an invalid value that doesnt follow the algorithm with its highest bit set to 0, and it forces a different value when its original state is 0, so I have no idea. With the algorithm revealed, one should be able to monitor the RNG by index.
Me after giving a go at decompiling: it actually similar to what
this post describes, should fix the failed cases of my original attempt. The function returns (x * -16807 +
floor(x / 127773) * (2 ^ 31 - 1))
mod (2 ^ 32) . Small text to indicate that they are int32 operations. I then didn't find much stuff looking at the functions that call the RNG advancing functions.
