This has probably been talked to dead, so if someone with better search skills can provide the relevant link I'd be very grateful.

I used to play the old ladder rating system, but with this new one I haven't make much experience yet.


From my short experience it seems that no matter if someone has similiar rating or is 150 above or below, the loss/gain is always around 14/15 rating points? Surely that can't be correct, the skill difference should be taken in account..

Also people claiming the rating system is quite bad for the lower ranks, which seems to be quite true around r450. At what point does
the rating system become a real good indicator of skill? Just the top10-20?

It’s TrueSkill with some slight modifications: https://en.m.wikipedia.org/wiki/TrueSkill

150 rating points is an absolutely massive difference implying a 68% win rate for the higher player. If the lower rated player wins, they will gain roughly 2x as much assuming a similar number of games played

https://www.warzone.com/Forum/697546-new-ladder-rating-system-working

you might be interested in reading this if you haven't already

Ok, I must have looked at the numbers wrong, or don't understand the rating system quite correctly. Although I'm usually okay'sih at math and formulas, my math brain is not quite awake right now.

Would you be so kind to explain a bit more? I'll be sure to drink some coffee before I read your answer.

e.g. I played a win against Neet 502 vs me (406). So only 96 rating difference.
24 games played for me, about 1070 for him.

Seemed the rating increase for me was only 15 or so.

Simliar playing a los against Sloppyfatfinger 534 vs me (436)
27 games for me, about 1030 for him.

Lost about 14 skill points?

Edit: posted at the same time as Stefano, will read through that link again first.

Edited 2/7/2025 16:03:08

I'm usually not good at math, from what I gather fizzer capped the rating changes so much that it becomes meaningless, but I might be wrong.

Edited 2/7/2025 16:03:58

Either way it's a good link you shared. I had seen it before, but not read it in depth yet. It's very good reading.

Edited 2/7/2025 16:43:22

After reading/glancing through both the linked post and the linked pdf's I've decided my math days are behind me. I love that people are so passionate about the topic that they go too such depth to analyse it. For me though, I'm a simple guy, I was just looking for a simple set of 1-2 formulas.

Edited 2/7/2025 19:50:16

Realistic range is 11-17. It all comes down to your rating and the opponents rating at the end of the game. The largest gain for me has been 16.2, and smallest 12.1.

As you saw in the great post by Beep that Stefano posted the system is TrueSkill with tweaked parameters, so no simple formula to use. If you like computers and want to simulate it, there is a python package for that.
As a rule of thumb, it does take into account ratings and the effects are more visible the more you play. As exemple once you reach 800 ratings you'll get 7 to 10 points for a win and will lose 8 to 13 for a loss. This effect will amplify for higher differences in ratings.

Cowboy, that's a really remarkably small range then. Very small pay off for someone beating buns then. (Except bragging rights I guess.) Although I guess his ranking is so out there that the gain might be a bit bigger.

Edited 2/7/2025 20:35:57

Cowboy, that's a really remarkably small range then

That's the complaint in the linked thread. The ladder's version of TrueSkill is set up to be extremely slow to learn, which is why it sucks.