|
|
|
|
|
by sdenton4
1948 days ago
|
|
Well, there's a lot of forms of randomness... One could use Thompson sampling. Every comment starts with 1 up and 1 down vote. The total number of up/down votes determine a beta distribution. When displaying comments, draw from the beta distributions for each comment, and present them in that order. High quality comments drift reliably to the top over time, but other comments have their own chances at the top to accumulate votes and better determine their place in the stack. |
|