So you are saying that there are multiple ratings in the system for the actual puzzles?
So a puzzle rated 2600 as shown to the users may be rated only 1800 to the system?
Puzzling.
Correct. The puzzles have basically been re-rated on the back end and those ratings are now being used for the puzzle selection algorithm and changing member's ratings. Some puzzles may have similar ratings to what's in the UI but many have changed a lot, so using the listed rating to gauge how the system is reacting isn't going to be accurate until the UI starts using the new ratings.
That would be ideal, but it doesn’t align with what I’m observing - or at least, it raises some questions.
Let’s say the backend treats low-rated puzzles as being close to our actual level. Then we should see rating changes that reflect that - similar to what we’d get from puzzles at our level.
But that’s not what I see. Consistently, about two-thirds of my puzzles match my rating, while the remaining third are about 1k points lower. For that third, I always gain 5 points regardless of speed and lose significantly more than I do on the higher-rated ones.
If rating changes are calculated in the backend, as you’ve described, then the UI should reflect those updated values. So why isn’t it? And if it’s not displaying them correctly now, won’t all past rating changes eventually need to be corrected?
It all seems very odd to me.
So you are saying that there are multiple ratings in the system for the actual puzzles?
So a puzzle rated 2600 as shown to the users may be rated only 1800 to the system?
Puzzling.
Correct. The puzzles have basically been re-rated on the back end and those ratings are now being used for the puzzle selection algorithm and changing member's ratings. Some puzzles may have similar ratings to what's in the UI but many have changed a lot, so using the listed rating to gauge how the system is reacting isn't going to be accurate until the UI starts using the new ratings.