Repository files navigation

To rank the best games of the year, a Bayesian Rating is computedon RPGCodex, which provides a solid ranking. However, this methodology is not appliedon SteamGAF/ERA, and a sum of scores is applied instead (think Eurovision). Our goal is to use the data shown on SteamGAF/ERA and re-rank games with other methods:

• a Bayesian Rating,
• Schulze method

Our prior is computed as follows:

• regarding the average score, the prior is the median of average scores,
• regarding the number of votes, the prior is the average of numbers of votes.

The original data consists in sets of GOTY candidates, which happen to be ranked. As pointed out byDurante, interpreting the ranks as a score leads to undesirable properties.

Full result ishere.

Schulze method is an alternative toBorda count, which is the Eurovision system. SeeRanked-choice voting.

The original dataset, which consists of the rankings submitted by each user, was used.It cannot be shared for privacy concerns.

Full result ishere.

• https://fulmicoton.com/posts/bayesian_rating/

• https://en.wikipedia.org/wiki/Additive_smoothing

• https://planspace.org/2014/08/17/how-to-sort-by-average-rating/

• http://www.dcs.bbk.ac.uk/~dell/publications/dellzhang_ictir2011.pdf

• https://en.wikipedia.org/wiki/Ranked_voting

• https://en.wikipedia.org/wiki/Schulze_method