3 ms·

> I don’t know if this problem has a proper name, but in game development you usually don’t want truly random numbers. I’ve seen the solution named a “pseudo ra

by klipt 7y ago

> I don’t know if this problem has a proper name, but in game development you usually don’t want truly random numbers. I’ve seen the solution named a “pseudo random distribution” but “pseudo random” already has a different meaning outside of game design, so I will just call it controlled randomness for this blog post. The correct term is probably quasirandom or low-discrepancy sequences: https://en.wikipedia.org/wiki/Low-discrepancy_sequence https://en.wikipedia.org/wiki/Low-discrepancy_sequence

dahart 7y ago
While those are good terms for deterministic uniform random number generators, that is not what the author’s talking about. He’s referring to how in games, you sometimes don’t want random patterns. But even when you do, you don’t want the pattern to be truly random, because true randomness sometimes feels unfair and sometimes feels un-fun. It’s a concept with a broader scope than low-discrepancy sequences.
- klipt 7y ago
  They don't have to be deterministic, the wiki article has a section on constructing low-discrepancy sequences from independent random numbers by imposing negative correlation between subsequent numbers: https://en.wikipedia.org/wiki/Low-discrepancy_sequence#Random_numbers https://en.wikipedia.org/wiki/Low-discrepancy_sequence#Rando... Which sounds pretty similar to what the author is doing - removing some of the independence from a sequence of independent random numbers to make them feel "fairer" (basically making them more in line with misconceptions people have about truly independent random sequences https://en.wikipedia.org/wiki/Randomness#Misconceptions_and_logical_fallacies https://en.wikipedia.org/wiki/Randomness#Misconceptions_and_...)
  - kzrdude 7y ago
    Maybe they have misconceptions about randomness, but are they wrong about their expectations for something that models the real world? I'm thinking the space of possible random events might be too big, or the distribution is wrong.
  - dahart 7y ago
    > They don’t have to be deterministic. Right. Agreed. Still, the idea here is about the perception of randomness at the expense of actual randomness, it’s related to but different from quasi-random ideas. Edit: Actually I don’t know, maybe you’re totally right about this article, maybe quasi-random is the same thing in this case. I was projecting a broader idea based on having worked on games and run into this problem... my philosophy on it is that often you don’t want any randomness, but sequences that are random seeming at first but are predictable and learnable by the player. Depending on the game feature, that’s sometimes more fun. In this particular case, they are trying to preserve the intended distribution, and sample a given distribution without shuffling. A low discrepancy sequence does do that, so I’m changing my mind here.
o09rdk 7y ago
Also hyperuniformity? https://www.quantamagazine.org/hyperuniformity-found-in-birds-math-and-physics-20160712/ https://www.quantamagazine.org/hyperuniformity-found-in-bird...
- gwern 7y ago
  Poisson disc sampling! https://www.jasondavies.com/poisson-disc/ https://www.jasondavies.com/poisson-disc/

by klipt 7y ago

dahart 7y ago

While those are good terms for deterministic uniform random number generators, that is not what the author’s talking about. He’s referring to how in games, you sometimes don’t want random patterns. But even when you do, you don’t want the pattern to be truly random, because true randomness sometimes feels unfair and sometimes feels un-fun. It’s a concept with a broader scope than low-discrepancy sequences.

klipt 7y ago
They don't have to be deterministic, the wiki article has a section on constructing low-discrepancy sequences from independent random numbers by imposing negative correlation between subsequent numbers: https://en.wikipedia.org/wiki/Low-discrepancy_sequence#Random_numbers https://en.wikipedia.org/wiki/Low-discrepancy_sequence#Rando... Which sounds pretty similar to what the author is doing - removing some of the independence from a sequence of independent random numbers to make them feel "fairer" (basically making them more in line with misconceptions people have about truly independent random sequences https://en.wikipedia.org/wiki/Randomness#Misconceptions_and_logical_fallacies https://en.wikipedia.org/wiki/Randomness#Misconceptions_and_...)
- kzrdude 7y ago
  Maybe they have misconceptions about randomness, but are they wrong about their expectations for something that models the real world? I'm thinking the space of possible random events might be too big, or the distribution is wrong.
- dahart 7y ago
  > They don’t have to be deterministic. Right. Agreed. Still, the idea here is about the perception of randomness at the expense of actual randomness, it’s related to but different from quasi-random ideas. Edit: Actually I don’t know, maybe you’re totally right about this article, maybe quasi-random is the same thing in this case. I was projecting a broader idea based on having worked on games and run into this problem... my philosophy on it is that often you don’t want any randomness, but sequences that are random seeming at first but are predictable and learnable by the player. Depending on the game feature, that’s sometimes more fun. In this particular case, they are trying to preserve the intended distribution, and sample a given distribution without shuffling. A low discrepancy sequence does do that, so I’m changing my mind here.

o09rdk 7y ago

Also hyperuniformity? https://www.quantamagazine.org/hyperuniformity-found-in-birds-math-and-physics-20160712/ https://www.quantamagazine.org/hyperuniformity-found-in-bird...

gwern 7y ago
Poisson disc sampling! https://www.jasondavies.com/poisson-disc/ https://www.jasondavies.com/poisson-disc/