@Jean-Louis: When you post a comment on an article there appears a 'Rating' selector above the edit window, by default set for 'none'. You can also select 'poor', 'good' or 'excellent'. I always thought that this was determining the number of stars that would appear in the top-right of the comment. (But is a mistery to me how 0-5 stars could be derived from just 4 chocies...)
I agree that it is probably not a good thing to make this too complex. But if we want to involve the rating system in the favorites scoring, the logical way to do it would be this:
We now attach a weight of 60/(N+50) to a member with N (>= 10) votes. So the total weight of the votes (60*N/(N+50)) saturates at 60 when N approaches infinity. The rationalization for the discount is that there is no indication that any of the favorites indicated by this person are really close to the top of his list. But if that person would have rated, say, 15 of the games he favorited with 5 stars, and the rest with fewer, we do know that these 15 are the top 15 of his list, and there would be no reason to discount their weight more than those for a person that has 15 favorites in total, and never rated any of those. So these games should get weight 50/(15+50) = 0.923. If in total that person voted for (say) 70 games, which would have given him a total weight of 70*0.5 = 35, the 15*0.923 for his 5-star games could be taken out of this total budget, to leave a 35 - 15*0.923 = 21.15 for the other 70 - 15 = 55 games he favorited, which means a weight 0.384 for each of those.
This system could be applied 'bottom to top', where you would first divide up the total budget (35 here) between the unrated favorites and those with 1-5 stars, to see how much budget he should get for the rated group based on the 60/(N+50) formula and the number of rated favorites, and what that would leave (of the 35) for dividing over the unrated favorites. This would then be repeated for the 2-5 star group for determining how much would go to the 1-star favorites, etc.
This would be complex, but there should never be any reason for the users of the website to understand how the scoring works. If they use the rating system and favoriting in the intended way, the page would calculate a fair 'figure of merit' from this, which they should simply trust as a measure of how much the crowd here appreciates the variant. Detailed knowledge of how the scoring is calculated is only useful for those who are looking for flaws or weaknesses through which they could subvert the system to get variants of their choice higher on the list than they should be. And why would we facilitate that?
@Jean-Louis: When you post a comment on an article there appears a 'Rating' selector above the edit window, by default set for 'none'. You can also select 'poor', 'good' or 'excellent'. I always thought that this was determining the number of stars that would appear in the top-right of the comment. (But is a mistery to me how 0-5 stars could be derived from just 4 chocies...)
I agree that it is probably not a good thing to make this too complex. But if we want to involve the rating system in the favorites scoring, the logical way to do it would be this:
We now attach a weight of 60/(N+50) to a member with N (>= 10) votes. So the total weight of the votes (60*N/(N+50)) saturates at 60 when N approaches infinity. The rationalization for the discount is that there is no indication that any of the favorites indicated by this person are really close to the top of his list. But if that person would have rated, say, 15 of the games he favorited with 5 stars, and the rest with fewer, we do know that these 15 are the top 15 of his list, and there would be no reason to discount their weight more than those for a person that has 15 favorites in total, and never rated any of those. So these games should get weight 50/(15+50) = 0.923. If in total that person voted for (say) 70 games, which would have given him a total weight of 70*0.5 = 35, the 15*0.923 for his 5-star games could be taken out of this total budget, to leave a 35 - 15*0.923 = 21.15 for the other 70 - 15 = 55 games he favorited, which means a weight 0.384 for each of those.
This system could be applied 'bottom to top', where you would first divide up the total budget (35 here) between the unrated favorites and those with 1-5 stars, to see how much budget he should get for the rated group based on the 60/(N+50) formula and the number of rated favorites, and what that would leave (of the 35) for dividing over the unrated favorites. This would then be repeated for the 2-5 star group for determining how much would go to the 1-star favorites, etc.
This would be complex, but there should never be any reason for the users of the website to understand how the scoring works. If they use the rating system and favoriting in the intended way, the page would calculate a fair 'figure of merit' from this, which they should simply trust as a measure of how much the crowd here appreciates the variant. Detailed knowledge of how the scoring is calculated is only useful for those who are looking for flaws or weaknesses through which they could subvert the system to get variants of their choice higher on the list than they should be. And why would we facilitate that?