The current model is linear: If I have 2x someone's points, I'm expected to win four times as much (win = 10, loss = -40). With 4x, I have to win twenty(!) times as much (win = 5, loss = -80). My suggestion is to use the square root of the ratio instead. This would reduce the number of points for large differences in score. The change would not be retroactive, and would apply only to games completed after implementation.
- The current formula is: (loser's score / winner's score) * 20, max of 100.
My formula would be: (loser's score / winner's score)^1/2 * 20, max of 100.
This would change points awarded as follows:
Old points ----> New Points: Win(Loss)
Even: 20 -----> 20 (20)
1.5x Loser's Score: 13.4 (30) -----> 16 (24)
2x Loser's Score: 10 (40) -----> 14 (28)
3x Loser's Score: 6.6 (60) -----> 11.5 (34.6)
How this will benefit the site and/or other comments:
-The current model makes it extremely hard for highly ranked players to play lower ranked ones: given the luck factor, winning 10-20x in a row against someone of even middling skill is nearly impossible. This is even more pronounced in large games, at even 2x the point value it is necessary to win 1/3 of eight player games just to break even.
-It is not possible to restrict games by points, so higher ranked players tend to avoid public games and the massive potential point loss. When was the last time you saw a top 100 player in a public game cooks could join? This would allow top players to play more games without taking such massive score hits if they lose.
-This may alleviate the related fact that top players tend to play only one non-standard map against pre-selected opponents (for example, the current Conqueror only City Mogul 1v1, #2 plays only private games, etc). Hopefully this change will encourage them to use more variety.
-This would be a more accurate rating model, as it accounts for the luck factor: as point disparities increase, luck stays constant, so linear scores will not be a valid prediction of who wins (even scores might split 50/50, but 2x scores are less likely to split 4/1, and 4x scores are very unlikely to split 20/1).
-This would make score resets for cheaters more palatable, as it would reduce the number of points opponents lose while they attempt to regain their ranking./list]