You're all a bunch of anarchists! >> Uniform ratings, what would change?
- Joe Grundy(jgrundy)Australia
Here's another approach you may have been wondering about. (I was.) Make each user's ratings "uniform" (ie evenly distributed) before calculating the average per game. This list is the games whose rankings would be most affected by switching to this approach.
As usual, now I'm gonna waffle a bit and most of you are just going to skip to the list.
DISCLAIMER: Folks, feel free to make your ratings as you see fit! I cast no judgement. I'm just playing with numbers.
But, sorry to say, it's hard to combine all your wierd subjective personally tweaked usages of the number system into any meaningful aggregate. It'd be so much easier and more "accurate" if you all just voted with your gut feelings assigning the values actually following the BGG guidelines. But you don't. You're gamers after all and many many of you play your own ratings games.
So given that hardly any of you follow guidelines anyway, I've decided to take them away from you. Now the all-wise Administration will usurp your attempts at individualism and make each of you vote uniformly across the ratings scale.
- If a user has a rating spread of less than 5, ignore that user
- For each user, scale their votes to be as uniformly distributed as possible.
- Add 'em up for each ranked game!
I've still applied the Bayesian calculation.
Voters: Who's still in?
Only about 25,000 BGG users rate any games at all. Between them they cast a bit over a million votes. (Are you doing your share, comrade?) I didn't want to include voters who give only one value, because all their ratings become "this game is mediocre". I'm sure that's not what they wanted to say. I extended this thought and ignore any users with a ratings spread of less than five levels, eg you are included if your minimum rating is 4 and maximum is 9.
So we end up ignoring half the 25,000 voting population (!) but we still keep 91% of the original million votes on BGG.
For ranked games, this means we still have 12,300 users giving 813,000 ratings instead of the original 24,000 users giving 898,000 ratings.
My data is a few days old now. I don't have the details of Aldie's ShillSeeker Filter (tm, pat pend). So if I just started with the BGG rank and output the "new" rank, those confounding factors would mean you can't see what actually happened. Instead, for each game first I just apply the data I've got as-is and show you the rank I get. Then I show you the rank it gets after making each user's ratings uniform. So hopefully you can see separately how much (or little) effect joining the true breatheren and abandoning your anarchistic ways would have.
I'll be doing a similar exercise in the near future but with a different normalising approach. Please let me know if you find this output format useful.
- [+] Dice rolls