I had no idea it was this complicated. I thought it was only "roll a random piece, if it's a repeat reroll once." Crazy that there's so much more to it and it explains some of the weird statistical behaviour I notice sometimes. Good video.
If someone was doing an implementation of NES Tetris on modern hardware, that IS pretty much the algorithm intended. I guess you could keep the mod 8 check where there's an unconditional 1/8 chance of reroll to keep repeats slightly more frequent, but that's just a coding oddity. The bit that makes it funky is that "get a random value" on modern hardware is easy, but on a NES you're stuck with stuff like linear shift registers which means the random component isn't as random as we'd like. But that's really the cause here IMO, NES "RNG" style.
What about an "anydrought" mode where instead of a 1/28 chance of repeating the last piece, there is a 1/28 chance of reintroducing the longest-ago piece
4:01 Doesn't every piece spawn with a canonical orientation? Can you force a piece to spawn in with a different orientation? If not, I don't understand how this is different from just having 7 different IDs for the 7 different tetrominoes.
If you hack the game, yeah you can change the orientation in which it spawns. Each orientation has its own ID, rather than having a piece ID and an orientation ID.
I'm curious if this gets cross referenced with the actual possible outputs of the RNG table if there's more information to be learned, like yeah O's could hypothetically always result in a repeat, but is there a seed that will even do that?
Anything that is theoretically possible is also practically possible in this game because you can pause and wait for the PRNG to be the correct number. Here's a tas of 1400 O pieces: ua-cam.com/video/-_qJec90LQ8/v-deo.html
tournament? no. high score attempts? the world record had pauses in it once (the first time eric got wr at 3ish million) but people hated it and I don't think you can get away with it anymore.
I had no idea it was this complicated. I thought it was only "roll a random piece, if it's a repeat reroll once." Crazy that there's so much more to it and it explains some of the weird statistical behaviour I notice sometimes. Good video.
If someone was doing an implementation of NES Tetris on modern hardware, that IS pretty much the algorithm intended. I guess you could keep the mod 8 check where there's an unconditional 1/8 chance of reroll to keep repeats slightly more frequent, but that's just a coding oddity. The bit that makes it funky is that "get a random value" on modern hardware is easy, but on a NES you're stuck with stuff like linear shift registers which means the random component isn't as random as we'd like. But that's really the cause here IMO, NES "RNG" style.
In SM64, a coin with high vertical speed (>75th percentile) and horizontal speed (>87.5th percentile) cannot spawn within 4 of the 8 angular octants
Imagine being so good that you can predict in real time.
well I assume AI learns it ...
What about an "anydrought" mode where instead of a 1/28 chance of repeating the last piece, there is a 1/28 chance of reintroducing the longest-ago piece
You turned into a math channel :p
Clear enunciation and explanation and very interesting video. :)
4:01 Doesn't every piece spawn with a canonical orientation? Can you force a piece to spawn in with a different orientation? If not, I don't understand how this is different from just having 7 different IDs for the 7 different tetrominoes.
If you hack the game, yeah you can change the orientation in which it spawns. Each orientation has its own ID, rather than having a piece ID and an orientation ID.
I'm curious if this gets cross referenced with the actual possible outputs of the RNG table if there's more information to be learned, like yeah O's could hypothetically always result in a repeat, but is there a seed that will even do that?
Anything that is theoretically possible is also practically possible in this game because you can pause and wait for the PRNG to be the correct number. Here's a tas of 1400 O pieces: ua-cam.com/video/-_qJec90LQ8/v-deo.html
@@hydrantdude3642 Interesting. Is pausing legal in a tournament setting? (Not that that would ever actually result in an advantage)
tournament? no. high score attempts? the world record had pauses in it once (the first time eric got wr at 3ish million) but people hated it and I don't think you can get away with it anymore.
That's super interesting
Don't see the link.
i forgor
it's there now