The reactions to the ratings are flowing in, and I noticed there are some misunderstandings on how to interpret the ratings, and how to calculate a percentile.  First of all, the ranking is based on the category.  That is, if your game is in “48h compo” it is ranked within a total of 1610 games, if it’s “jam”, within 736 entries.  But that is not the whole story.  A number of games have not been rated because the number of ratings fall below the threshold (which I believe is 12 or 14 ratings or some such).  Without knowing how many games were not rated, you can’t calculate your percentile.  As a first attempt to solve this problem, I took a random sample out of all games, and noted the percentages that were and weren’t rated. Figures calculated from all games would be nice to have, but these aren’t available afaik.

Percentage of 48h games rated: 88.9% (32 out of 36 in my sample)
Percentage of Jam games rated: 64.5% (20 out of 31 in my sample)

Note that the percentage for Jam games is much lower for some reason.  Anyway, this means the total number of rated games in each category is approximately:

48h: 1431 (1610*0.889)
Jam: 474 (736*0.645)

UPDATE: thanks to ashdnazg, we now have accurate numbers!

48h: 1357
Jam: 566

To get your percentile, divide your ranking by one of the numbers above and multiply by 100.

Also it makes sense to look at the absolute ratings (the number of stars) and to know something about the relationship between absolute and relative ratings.  In some cases, a 0.1 star change can drop or raise your ranking by 100 or so.  I’d also like to see a graph of this relationship.  Maybe I’ll make one myself.