i dont think anybody really cares that strongly about the system anymore to be bothered. Blizzard has been slanting games for too long for anybody to give a crap.
1 Like
your MMR is not a simple number, it’s a matrix and nobody could do anything with it
yeah, so how many games per season do you play on that account? I think 75 wins per season is a good number. You don’t have to play too much and you’ll get an accurate rating.
One thing about your experiment. You have to assume that the player already has an established SR. And what that means is each time you flip heads, tails is more likely to come up the next time. If you flip 10 heads in a row, there’s going to be a 60/40 chance that the next flip will come up tails, because the rank of the player has already been established.
did you see Kaaumba’s paper about poor match quality? I think it shows everything. In ideal circumstances you will have a 16.6% impact on the game, that’s 1 divided by 6. but let’s say all 11 people are SR manipulators, well then you basically have a 50/50 shot. So under normal conditions you probably have a 10% or more impact on the game. That’s why we need guilds and a robust LFG system.
Just think if Overwatch was 100v100, how little impact you’d have on the game. And that’s an aspect of Overwatch that they left out. Team formation. 100v100 solo queue would be dumb, but if we had guilds and a robust LFG system and scoreboards…then 100v100 would be doable. Because half the battle takes place off of the field.
1 Like
ROFL as I write this there are FOUR throwing groups in LFG.
Instead of nagging can you tell me what’s the answer? Because it seems you are into the habit of nagging I said I think I didn’t say I know
If I had all this QoL for the LFG, I would never use matchmaker again. QoL for the LFG
furthermore, I think everyone should start at zero SR and be able to shed SR with the click of a button, you can’t stop sandbaggers, so why try?
I know they have the new team that finds and punishes sandbaggers, but with my method, nothing will be left to question.
What I’m saying is that 100 straight games on an account where you’re 56% favoured to win [not by Elo just by player ability] and win 25 every time and lose 22 every time (so gain steadily even from a 50% win rate). And that for 1/5th of people like that who both the system (through pbsr) and by their own likelihood of impact (56%) are outperforming their tier… 1/5th of those will not even move after one hundred games. Over half the case moved less than half a tier. To me this is extremely tedious progress for a normal person with a normal life. Like this person hasn’t plateaud. They are playing consistently and notably better than the average of their tier but could take a few hundred games to realize a tier difference even if they continued to improve (and so kept that 56%).
If they started at their correct tier they will have 50% expected win rate and variation is even greater. On average half a tier on average in either direction randomly over 100 games. To me that is a big yikes.
My assessment of 56% as a decent expected win is based on being solidly but not remarkably better than the average of your tier. Of course normally as you rank up that would go down and as you rank down it would go up but because that’s modeling beyond what I want to try and touch atm I’m using a simple criteria. Still it’s a remarkable amount of variation.
The 56% is out of thin air. If smurfing, trolls, boosters, etc effect is significant than maybe even playing 300SR above you only gives you 53% Sr or something. It might be 57% in noiseless circumstances but with how many incorrect MMRs these issues create (noise) the result is pushed closer to 50% which means slower progression both negative and positive. In which case the progression to then start averaging around that new point 300SR might be better recorded in hundreds of hours than hundreds of games.
It’s depressing when put in those terms and probably why MMR can feel so inaccurate and tiers so variable. Exceedingly slow progression with lots of out of place mmr. It really does mean if youre a bit better than your current tier (maybe up to 500 better) it could easily be faster to start a new account and hit that new average rather than try to bring your old one up there.
I haven’t read that particular paper. I was competitive in a 3v3 and 5v5 game and I experienced how slow progression was in 5v5 mode comparatively even as a top .1% player. It was still fast of course but were talking picking up like 2.5x as many losses along the way. The more I look at outcomes when you’re close to the right skill tier the more believable maintaining multiple accounts at moderately different skill tiers despite trying your best on all of them seems. The cynic in me does realize they make more money both directly through smurfs and by people wanting a fresh role at their mmr after improving at the game.
1 Like
how did you come up with this number?
this is way too long to read, would you rather talk about it over voice comms?
I could at some point later. Just killing time. Overtime atm lol
56 is basically me saying “what would the likelihood of someone who is averaging a performance decently above their tier be to win any particular game in their tier”. This would be much higher in a 1v1 game. It would be much higher without smurfs, trolls, fresh players on their way down, derankers, and all the people those accounts affect. Basically I’m describing “noise” in the system as what slows down the lowering or raising of SR in this mmr system when a player is not at the mmr they perform at on average. So if you’re undervalued or overvalued for your mmr you will eventually approach your actual rating but noise slows this down.
So if someone’s average performance is 2200 and they’re 1900 what could we expect them to win as a % of their games given there’s all these noise factors? Even in a noiseless situation it clearly wouldn’t be 100% or close. They’re not THAT much better. Maybe 70%? But every noise factor pushes this closer to 50% since every noise factor can benefit or hinder the player (we could make an adjustment for 5 allies vs 6 enemies here but I’m too lazy).
So 56% is just a best guess at what might be our expected win rate for such a player in the current system. If it’s actually 58% they progress faster though it’s still slow. If it’s 53% they progress like molasses. In any case I’m saying unless we think a player like that has very high odds to win each game their rank progression will be extremely slow. And this is the problem with smurfs and to a lesser degree starting new accounts by default at 2300 mmr.
When I made a new account I was winning 80sr losing 80sr. Progression was lightning by comparison. To see those same gains on an account winning and losing 24 would take hundreds of games longer. And that’s what I’ve experienced I think.
TLDR. Another big bunch of walls for ya. No worries if you don’t wanna read it. We can talk in a couple days if you want tho. Feel free to add me.
2 Likes
sent, let me know when you’ll have some time
1 Like
1)people would probably not know what to do with mmr represented in his true form.
- people who want to know are usualy the ones who think system is unfair to them.
You just need to know how to climb, you dont need to see your mmr to do that.
3 Likes
They should just remove the whole MMR. Especially if it’s within ±3 range. No point at all.
We basically had it season 1.
I was rank 48
It went from rank 1 to rank 100.
You only moved when you had a certain amount of wins or losses.
Ppl complained cause you could play 10 games & not move.
So the changed it to sr
Um, no. It was changed because the system was brutal, you’d lose nearly a whole rank if you lost 1 match, and a minuscule amount when you won a match
I think the system is totally fair. I want to see the real number.
“Hello Mr. NFL commissioner how do you get to the super bowl?” “um, I’m not going to tell you just try to win as many games as possible.”
But you can see real number after your placement 
Thats basicly your MMR. I mean if thats not first initial placement. Every other placement is placing you where you belong based on your skill so SR after placement = your actual MMR. If you play enough for sure, and not 1 comp game per month.
is it the real number or is it basically the number, because it can’t be both
I have 100% blind faith in the matchmaker/mmr system.
I want the matchmaker/mmr system to be 100% transparent.
These two statements are not necessarily in opposition to each other.
Well, I am pretty sure that your SR number after placement is accurate representation of your MMR in that moment. Thats why you can do big jumps in placements if you didnt belong in your rank before placement.
I am not sure if MMR is actual simple number as SR tho. I do believe i did hear someone say that it is. Was it Jeff on Seagull stream?
MMR is a floating point, scaled to standard deviations. I can get the exact quotes (and why I read them this way) later if you care.
1 Like