Ranking system

[LeeroyJenkins1301]

Hi. I really cannot understand the rating system. Yesterday and today I won all 4 of my games - actually lost a few rating points. Just lost 1 game to a higher rated player and immediately drop down a rank. How can you lose rating points for winning and then go down a whole rank for losing 1 out of 5 games??

[Jerry_R]

Did you win by timeout?

[mark5000]

There’s been a lot of discussion about this circumstance in the rating tweaks announcement:

2020 Rating and rank tweaks and analysis

tl;dr - Ratings and ranks have been tweaked a little, there should be a bit less volatility in ranks now. Ratings and ranks have been adjusted retroactively. Main changes: We were using fixed windows to update ranks, considering up to 15 games at a time to update your rank. When we got to the 16th game, we’d “commit” your rank and start a new block to update your ranks. We now use a sliding window, so we look at your most recent 15 games (up to 90 days), and use that to update your rank, this…

In short, the rating system uses a 15-game sliding window. It so happens that when you won 3 games yesterday and today against 24k, an 18k and 14k players, you also pushed out wins you had 15 games ago against 16k, 19k and 15k players. As a result, you rank increased very little. Your 1 loss today also pushed a win out of the 15-game window (you had a good win streak there), and so your rank fell more than you probably expected.

There’s been a lot of discussion about this system and whether it should change. But perhaps it’s working as expected. This video explains the concept pretty well (we try to measure the blue line rather than the red one):

Trend and variation

[meili_yinhua]

mark’s statement about there having been discussion us a huge understatement, there has been a ton, and it’s not a particularly popular move, with problems ranging from there was a bug in the initial launch (which I still don’t know if it was ever corrected, I was told of concerns about fixing this specific one), to questions of whether or not it makes sense on a theoretical level (I have a little beef with the usage of the “trend and variation” video in this context because it doesn’t make sense in terms of rating: you should never make the opposite conclusion from new data, if there is an indicator that your rating should be higher than the previous best guess, you should never lower it… Particularly when you have no theory saying that it should move in particular ways/patterns… otherwise you weren’t working with the best guess beforehand to begin with), to questions of what the reasoning behind the presented measure behind the accuracy of our system is really any good.

In short: we’ve been up and down many topics on this ratings system, and it’s going to be unintuitive for a while, but I do believe there has been a bit of an initiative to further investigate new alternatives