News for hackers. Stuff that matters.
Re:A New Algorithm for Controlled Randomness by dahart on Saturday August 31, 2019 @07:43PM (#20849508)
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.

Re:A New Algorithm for Controlled Randomness by klipt on Saturday August 31, 2019 @07:46PM (#20849520)
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#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_...)

Re:A New Algorithm for Controlled Randomness by dahart on Saturday August 31, 2019 @07:55PM (#20849562)
> 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.

Re:A New Algorithm for Controlled Randomness by kzrdude on Saturday August 31, 2019 @07:54PM (#20849555)
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.