I believe that leaving is not changing your MMR, because if you leave all your placement games on new account, you will place at your starting point in placement, in mid gold. I experinced 3x leaving in row resulting in -150 SR. After I won next match I received a lot more SR than I would normaly do. Thats why I think it doesnt affect MMR at all.
I once tested this too on new account. I played 4 games and left the last match to see my final after placement SR. Then I went back to match and when I finished the match my SR was actualy lower because I dint play well. It was like game placed me at my final MMR from match 4 and match 5 didnt count because I left it (at moment when I left the game before I joned back) But because I went back and continued playing, losing match and played bad I placed lower.
So after all placement games I had actual SR different than my season high even when I didnt play one single game after placement. I placed lower with 5 placement games played than I would place with 4 placement game played and 1 left match
I dont have soruce post, I use logical thinking. How can communication affect your SR but not MMR? I will present example.
You are in match and it is the last fight in overtime, you have chance to affect outcome by telling your team mate to go ball to stall the point. He would switch from zarya for example to ball. Now it depends how you comunicatite it, you can say " Zarya, take ball please so we can keep overtime" or you say nothing. First option will lead into Zarya going ball to stall the point, second option where you wont say nothing will lead into zarya keep playing zarya and game will be lost because nobody can make it to payload in time. Little complicated example, but as you can see your communication skill have direct impact on result of match as it will be either +SR or - SR. But if your commucation skill saved the game while you wasnt even there on payload in the time when team won, your MMR is not changed, because you didnt change any of your stats. Thats how I see it.
Idk maybe I am overcomplicating things
I think so, definitely for majority of player base which has MMR at SR or super close to it. When I was smurfing, I saw SR gains in high numbers, massive amounts of SR gains actualy, rarely it went above 100 tho, so I think even smurfs far away from their true elo dont have MMR too far from SR. My prediction is that SR cannot go more further than like 200 points from MMR.
Well just had a match where there was a leaver within the 30 second to 60 second window and the match did not cancel, I am thinking if there is any intentional change, its based on whether the leaver had any activity or not (dealt damage, etc.).
Unfortunately no, I decided to leave by option of forfeit (choosing to leave after two minutes from the original leaver), so the replay is invalid (should have remember that would have happened). If I see the behavior in the match again, I am going to play out the match for the replay.
Summary insufficient. I want a precise, word-for-word definition of “skill” for the purposes of this game, and I want to know how it is measured. I also want to know what “skill” is being measured, and how, precisely, it is being measured?
What I’ve written (the whole article, not just the summary), including references, is a compendium of everything that Blizzard has said on the topic. They have not published their code, or equivalent, so I can’t give you that.
However, you should read http://www.moserware.com/2010/03/computing-your-skill.html which gives a semi-quantitative description of how skill is measured in Microsoft’s TrueSkill, a similar rating system. Code for TrueSkill is available here: https://trueskill.org/, https://github.com/sublee/trueskill.
MMR is not skills. Sorry dude. We had this convo before, Jeff said MMR is like SR with a slight difference. SR is super based on win/loss. MMR must also be based on Win/Loss. When I read “slight difference” I dont see “major difference” to SR.
MMR appears to be very condense, and only affected if you beat the odds of winning or lose when you were favored to win. - Thats not skill thats based on Winning / Losing a match against the predicted outcome. You know what that is, thats machine learning. If it factors in stats - that Neural network is going to take more time to compute.
What I like about this post, it gets all stuff together and points out there are some BIG holes. Mainly how does a team get put together. I would rather them tell me its random than not tell me at all.
Another glare and a realization. They built this to classify you and NOT to climb. I think that realization others need to make. Because we all know to “climb” you have to be face with people who are better. If MMR spectrum is based on brackets then people I’m against and with all have the same MMR. How will I grow? I rather play against people with a higher MMR, and not “prove” their system is wrong.
Also - stats / MMR have to differ per CHARACTER. If we’re using this system that chess players use and such, the glaring difference is in chess you cant switch to widow. Some players are going to be naturally gifted at certain playstyles.
I’ve had this happen a couple of times as well. If I catch it on a game while recording then I’ll make sure to send it in. I stupidly didn’t record it at the time because I thought it was just a new feature (All of the times I’ve seen this the player moved at least once before leaving the game, so I thought they made cancellation due to early leavers only happen if it’s in the first 2 minutes and they didn’t move. In all these cases they left before the first 2 minutes of the game and it did not cancel).
There are no ‘slightly above +3s’ because MMR is a range from -3 to +3 and is measured in Standard Deviations. I’ll have to get out my Statistics textbook and see about posting a graph of 0-5000 with the standard deviations overlayed.
