it is explained here — https://www.evanmiller.org/how-not-to-sort-by-average-rating.html
basically, a small number of games played by a person would give a less confident estimate of their skill, than a large number of games played by that person.
Ratio’s don’t factor that in. For example, a player with 100 wins and 2 losses gives a ratio of 50. While another player with 1 win and 0 losses gives a ratio of infinity. And from ratios, we are supposed to believe the second player has higher skill.
the link includes an excel example. equating win/loss with up-vote/down-vote.
