20.07.2012, 01:27
Hi Berni,
In the Liga, people will usually choose the 2 or 3 puzzles that they are good at, they don't choose at random. So for each puzzle the contestants will have a high average rating, maybe a median of 2000 or 2100. If you give rating scores using the normal formula with median=1500, then most ratings will fall. Maybe you can see this from your Testbetrieb data?
Some ideas?:
1) Have both a Überraschungsrätsel-rating and also a Liga-rating.
2) Or each user can choose if they want their Liga-puzzles to be rated or not-rated.
3) Or change the formula, so the median solver gets [the median-rating of the contestants] instead of 1500.
In the Liga, people will usually choose the 2 or 3 puzzles that they are good at, they don't choose at random. So for each puzzle the contestants will have a high average rating, maybe a median of 2000 or 2100. If you give rating scores using the normal formula with median=1500, then most ratings will fall. Maybe you can see this from your Testbetrieb data?
Some ideas?:
1) Have both a Überraschungsrätsel-rating and also a Liga-rating.
2) Or each user can choose if they want their Liga-puzzles to be rated or not-rated.
3) Or change the formula, so the median solver gets [the median-rating of the contestants] instead of 1500.