I played a game today against someone with a rating about 550 lower than mine and won. No real surprise there. But imagine my surprise when their rating went UP. By as much as mine did!
I guess this is one of those weird edge cases the algorithm wasn't really designed for. Any thoughts on what's going on here?
Here's the game in question:
en.lichess.org/HYS1UWz6qomLWithout reading through the documentation, I think it might be a bug
I can't think of a reason why a loss should ever decrease RD no matter the rating differential, nor why a decrease in RD would increase ratings directly
It is always funny to see that but it is something that can happen in the glicko2 system. It is not a bug it is an "inaccuracy" in the rating system.
Sample rating calculation of Loz16, for this single game alone (not ideal).
Loz16
Old Rating: 796.0
Old Rating Deviation: 85.0
Old Volatility: 0.05
Opponent:
rating: [1340]
rd: [64]
Loz16 result: [0]
New Rating: 794.2
New RD: 85.0
Rating gain: -1.8
#2 Agreed; such an outcome is not possible in Glicko-2.
Your opponent had less than 800 rating points so it gives them points to get back to 800. Thats normal since you normally can't have a rating below 800
The implementation is softly bounded at the lower end (basically at 800 but you can temporarily go past it).
Ultimately, it would be preferable to have the ratings above 0 at least. As it is now, 800 means in principle the lowest rating. If new players arrive which needs a lower rating, the current ones with 800 will get bumped closer to 1500 after a while.
LOL. Both sides increase rating. Sorry its funny.
Sorry but the game was so bad both of you guys deserve the extra points
@7
Players may start at 1500 but their true rating could be below 800. But they are somewhat protected by the minimum of 800. Could this then be an explanation of why lichess rating has been observed to be somewhat inflating.
