This week I launched Rank the Things, a project in which people can vote to form a crowdsourced ranking of anything from pizza toppings to Pixar movies. I wanted to make it as simple to use as possible, so one is given a choice between just two options and has to pick which they prefer (or skip if they really can’t make a decision). To get a decent sample size, they’re given a bunch of opportunities to vote, but not too many to prevent one person from having too much influence on the final results.
Of course, there is the challenge of how to take all these votes and turn them into a ranking. The simplest way is to divide the number of votes for something over its total matchups, but that runs into the problem that there isn’t necessarily going to be a balanced strength of schedule. For instance, as of this writing, about three times as many people have weighed in on pepperoni vs. banana peppers as they have banana peppers vs. black olives. Therefore, I opted to go for a maximum likelihood estimation instead, which adds some complexity but isn’t excessively fancy.Continue reading Rank the Things: How the ratings work (feat. math)