02.05.2021, 11:36
Some suggestions have been made to alter the statistical evaluation in order to filter out the "troll ratings". Before such changes are implemented, a few more thoughts.
First, it is not clear to me how to define what "troll ratings" are. I take it that people using this expression mean ratings which are given solely for the purpose of changing the overall appearance of the puzzle, without any reference to the actual contents and the quality of the puzzle. That alone is a serious accusation. After all, any such vote can only be given after the voter has invested some time and energy solving the puzzle. (One might argue that the puzzles in question come with a link that allows automatic solving, so there is no actual investment. But that is the author's business, isn't it?)
And second, even if there are any such votes, the filtering process cannot distinguish between troll ratings and genuine ratings, i.e. people who have solved the puzzle in question and sincerely judge that it is a bad puzzle. I repeat, such votes have their rightful place in the rating statistics and should not be removed, else it would be censorship. Solvers have the right to disagree. The filtering process would serve to eliminate even genuine disagreement, and I am strongly opposed to that.
A filtering process in the above sense requires guessing why a person has given a specific rating. I disapprove with Pjotr's approach (checking the rating average each time another solver has voted), because it thwarts the anonymity idea. In the end, any attempts at mind-reading come down to a claim that solvers justify their votes. As I see it, the rating process was originally implemented as it is so that exactly this justification is not demanded. We should seriously reconsider whether a change in that direction is what we really want.
First, it is not clear to me how to define what "troll ratings" are. I take it that people using this expression mean ratings which are given solely for the purpose of changing the overall appearance of the puzzle, without any reference to the actual contents and the quality of the puzzle. That alone is a serious accusation. After all, any such vote can only be given after the voter has invested some time and energy solving the puzzle. (One might argue that the puzzles in question come with a link that allows automatic solving, so there is no actual investment. But that is the author's business, isn't it?)
And second, even if there are any such votes, the filtering process cannot distinguish between troll ratings and genuine ratings, i.e. people who have solved the puzzle in question and sincerely judge that it is a bad puzzle. I repeat, such votes have their rightful place in the rating statistics and should not be removed, else it would be censorship. Solvers have the right to disagree. The filtering process would serve to eliminate even genuine disagreement, and I am strongly opposed to that.
A filtering process in the above sense requires guessing why a person has given a specific rating. I disapprove with Pjotr's approach (checking the rating average each time another solver has voted), because it thwarts the anonymity idea. In the end, any attempts at mind-reading come down to a claim that solvers justify their votes. As I see it, the rating process was originally implemented as it is so that exactly this justification is not demanded. We should seriously reconsider whether a change in that direction is what we really want.