Something the "just git gud supports" people neglect

You’re saying I shouldn’t edit out spelling mistakes if I see em… reaching and desperate buddy.

First, you made a false presumption of 50% WR at all skill levels and talked about how MMR is an issue. Both factual untrue, and any conclusion drawn before is flawed as these things show your lack of understanding.

correlationis implied by “goes up with”.

  1. a mutual relationship or connection between two or more things.

back to name-calling.


At bronze 45-47%
At GM 55-60% (maybe even more at top 500)
Also isn’t at infinte games this happens but the infinie is the model you use to measure skill vs wr.
Never said it was rigged. This shows you don’t understand the sentence.
MMR can have a detrimental or additive effect on your game simply because it has an effect.

These are very easy questions don’t know why’d you ask em.

The problem about you calling it out. Is you used as a point to back your arguement up. I made no claims my formatting or w/e is better than anyones, basically cause I’m happy to stick to the point. I’m happy to call out how egotistical and desperate it is to bring it up though.

In conculsion you couldn’t answer my questions at all and just spammed more insults and assumputions.

Nice counter point.

I mean you were claiming it was at a professional standard. Like if you don’t want to be held to that standard then why claim that it was? :thinking:

so then do explain what you were referring to with “I’d call this a professional way of communicating” when you said the following:

because it really seemed like you were referring to your own paragraph/section immediately before that line (that included the questionable line) and regarding that as “a professional way of communicating”.

idk where to even begin…

ok first with overbuff:

addressed before when I said:

if you don’t understand what’s said there esp in regards to the 2nd point with 2 subpoints, ask. otherwise no, that’s not how WR works esp in context of how you want to use it for you “model”.

it’s not a false presumption that “after” an infinite number of games, one gets a 50%WR because matchmakers has stabalised where everyone is for their skill level.
like what do you think matchmaker and MMR/SR are doing all throughout that infinite number of games?

They will still be active all throughout those infinite number of games. And if they’re not rigged, then players will reach that stabalised state mentioned above.

This was all explained in the previous reply and in the bit before where you quoted… Like this isn’t a case of my deflecting or not understanding what was said. It was literally addressed. If anything, it seems you were the one that didn’t understand.

inb4 you call this an insult and then go on and on about “omg you have no arguments so you insult, misquote, defect, etc.” :eyes:

not really because you followed it up with %s which entirely threw me off of what you were trying to say. like were you trying to say “WR goes up with rank linearly 47-55% of the time”? or were you saying “you can rank up even with 47-55% WR”?

A much clearer way of saying it would’ve been “WR is positively linearly correlated with rank.”


  1. lacking knowledge or awareness in general; uneducated or unsophisticated.

I mean considering you failed to see things in my previous reply the did in fact address what you raised in your current reply, and quite clearly you didn’t understand the explanation before despite being laid out for you, how does that description not fit?

1 Like

that’s a strawman. Please quote where I said that. At least this is proof that you making assumptions based on no facts and stating them as fact.

Direct communication is a professional way of communicating imo. I showed where you aren’t direct.

You made an assumption that was untrue.

Enough games have passed to support the model. The proof is that you can see a trend in winrate.
No scenario has an infinite number… don’t know why you’d even point that out.

In GM you get less SR for a win and more for a loss so it is impossible to stay in GM or any rank above diamond with a 50% win rate it’s the same for bronze to close to median rank.

As I side note, I could go on forever I mean the fact you think I thought that I was claiming rigging when I said MMR has a detrimental and a positive effect on a match. Just shows your reaching and not drawing logical conclusions.

Oh, another support blame thr…OH, wait a second :relaxed:

The game still requires teamplay and coordination. More individual impact CAN happen, but a coordinated team of 5 will mop the floor with 5 people who take the “it’s a deathmatch game now” sentiment literally.

Still a team game.

1 Like

The quote was literally provided there…

Not really. If anything it showed how your poor wording and poor punctuation leads to misundertandings of what you mean.
I.e. if yoy wanted this bit :point_down:

To actually be talking about direct explanations in general as opposed to something else, you should’ve worded it like say
“Lastly, I’d call direct explanations as a professional way of communicating. And by your own words, …”

The way you had it originally, you finishing off the sentence with a period when you weren’t actually intending to, plus the wording of “on the point of direct explaination I’d call this” made me believe that you were referring to something else i.e. something you had said earlier.

Wow, like did you not read how there was another point raised with 2 subpoints showing how overbuff winrates aren’t all the reliable with their win counting logic?

uuuh no not really.

it’s quite consistent at about +/-24SR unless you’re >=4200.

i.e. if you truly cap out at <4200SR then you will have 50% WR “after” an infinite number of games regardless of playing a high individual impact hero or not. hence the point I’m making about how your “model” doesn’t actually work. you can’t look at an individual’s WR “after” an infinite number of games to judge whether the hero they played has high individual impact or not.

how? because there simply isn’t a correlation between “WR after infinite games” vs how much individual impact a hero used to achieved that has.

  • with a high individual impact hero you will get that 50%WR state if you cap out at <4200SR as explained before
  • with a 0 individual impact hero, if we follow the logic of your “the dps skill diff between allies and enemies average out over infinite games” —> assumption of even likelihood of good and bad teammates (the “averages” can’t cancel out if they weren’t equally likely) —> also converge to 50%WR if you cap out at <4200SR
  • with a low individual impact hero (i.e. not 0 or negligible but not high either), when you take into account the aforementioned even likelihood of good vs bad teammates that you assumed, they will still eventually get to that 50% WR “after” an infinite number of games because you stochastically still have a drift to your true SR place, just with a smaller drift.

and how this all changes for people that actually are >=4200SR is that

  • the high individual hero case plays out the same for all the same reasons, just that they will have some WR>50% for how high SR they’re supposed to cap at to combat the SR gain debuff
  • the 0 individual impact hero case simply has 0 hope of maintaining SR above 4200 and will just cap out at 4200 instead before SR gain decay applies
  • the low individual impact hero case depends on wher the player was supposed to cap out i.e.
    • if their drift rate/factor >= SR gain debuff factor at where they were supposed to cap out then they will reach and stay where they’re supposed to cap out as if the SR debuff wasn’t there “after” an infinite number of games, with whatever WR% just like the high individual hero case.
    • if drift rate/factor < SR gain debuff factor at where they were supposed to cap out then they’ll just fall and cap out at 4200 with 50% WR

and if you were observing this for the >4200SR edge case, you can’t really tell because

  1. you don’t know where people are supposed to cap out (before considering the SR gain debuff) “after” an infinite number of games to differentiate between

    • someone that was supposed to cap out >4200 but played a too low individual impact hero and didn’t vs someone supposed to cap at <=4200SR; nor
    • whether they played a high or low individual impact hero where they’re supposed to cap out >=4200SR
  2. you don’t know really know the necessary WR per SR gain debuff factor

  3. and most importantly, when the distinct cases can lead to the same results between each other, obviously you can’t “just look at WR after infinite games have passed to judge whether the hero has high/low individual impact”.

no. in the <4200SR cap out case, you literally cannot get a SR that’s far off from 50% without the assumption of a rigged matchmaker. like matchmaker is still active and doing it’s job in your theoretical infinite number of games. and if it’s doing it’s job properly, one will converge considering between diamond and 4200SR, the SR gains are deterministic and not favored for gains or loses.

just because you fail to see the logical implications of your own logic, doesn’t mean I’m reaching or “making things up”.

You won’t believe this but

win 1v1’s against dps by realizing which 1v1’s to take (because when a dps kills you in a 1v1 it’s you won’t believe it - your fault - be it via positioning, cs managment or just aim)
keep someone alive well by realizing that their utility does it better than heals

Your claim…

None of that is me claiming I was writing at a professional standard. I was pointing out your not and you choose not to and therefore don’t communicate in a direct way.

You’re connecting dots that aren’t there and stating your assumption as fact.

“about” – eyeballing at it’s finest.

Overbuff data for WR for rank = Sum of (Each hero’s ( hero pickrate * winrate))
GM = 53.71%
M = 51.24%
D = 52.09%
Plat = 51.27%
G = 50.24%
S = 49.48%
B = 47.67%

Observable trend.


zzzz… misinterpret more stuff you just proved you draw illogical conclusions about the things you read.

Your claim…

Unless the rank population is on a bell curve and the adjusting of SR per win/loss is how you distribute the population.

Rank pop - 2018

  • Bronze – 8%
  • Silver – 21%
  • Gold – 32%
  • Platinum – 25%
  • Diamond – 10%
  • Masters – 3%
  • Grandmaster – 1%

This is a bell curve. Is it technically impossible to have 10% people in diamond and 25% people in plat if you don’t give less SR for a win and more SR for a loss.

There are other factors that determine SR gains and losses… like you get more SR for a hard game and less for an easy game.
That difference when added and minimized over infinite games… guess what… cancels out.

I do know that this doesn’t disprove “WR that’s significantly deviated from 50%”
Your claim was tho…

If you’re always in diamond and about isn’t true. (I’d say plat do but it is possible to gain enough SR through performance-based SR to negate a negative win rate)
If you’re always in silver and below isn’t true.

Proving you’re understanding of the issue is flawed, under the presumption you wrote that correctly.

Edit: Used over buff data over 6 months and to 2 decimals. (
P-D-M do not follow the trend but does overall. This is likely down to performance-based SR and new accounts affecting data.

when I said “The quote WAS iterally provided there” it was pointing to this: :point_down:

where the innermost quote was where I was referring to that I did include where you said what I claimed you said. And yes, there was a misunderstanding from your poor wording + punctuation that I acknowledged but apparently you didn’t understand nor read my acknowledgement which also explained how your wording mislead me. :roll_eyes:

evidently someone didn’t open the linked posts :roll_eyes:
literal graphs from the linked posts:



but sure, “it’s just me eyeballing” /s

it doesn’t because very evidently you’ve either not read or not understood my points about the flaws of overbuffs winrates :point_down:

:question: :thinking:
you’re going to have to explain furher how that part of the explanations is somehow misinterpreting what you’ve said or an illogical conclusion.

uuh no you can absolutely get a bell curve distribution of people across the ranks without inherently making SR losses are of larger magnitude than SR gains.

like what do you think an inherent distribution of the playerbase’s skill level (i.e. not in terms of SR, but simply overall how good people are) looks like? Answer: very likely a bell curve itself.
If someone wins, someone else loses. i.e if someone gains SR, someone else loses SR (not necessarily by the exact same amounts obvs, but the point of “someone’s gain is someone else’s loss” stands) —> you can’t really have everyone or even an askewed portion of the population just gain plenty of SR to jump ranks without dragging another portion down.

So yes, you can have 10% in diamond and 25% in plat without making SR losses larger in magnitude compared to SR gains, because the thing SR measures inherently has a bell curve distribution.

it’s especially true for people who cap out between diamond and 4200SR because SR changes are more consistently +/-24SR for even team SR matchups as shown by the graphs + linked forum posts before.

people who cap out at a certain SR in silver or a certain SR in bronze will be in that 50%WR state because despite the SR changes there aren’t as consistent as between diamond and 4200SR due to performance based SR, there’s still the core matter of whether such an individual wins or loses is very much largely about how well their team plays vs how well their opponents play.

i.e. as everyone converges to their “true” SR/MMR, the likelihood of someone winning when being matched against players of higher SR/MMR is lower, and conversely someone of losing when being matched against players of lower SR/MMR is also lower.

Not to mention as everyone converges to their “true” SR/MMR, the performance based factor would be less of a factor in matches around their “true” SR/MMR (as they are playing at the level that they’re actually at as should others in the match), but have more of an impact further away from the “true” SR/MMR to push them back towards their “true” SR/MMR as per their design.

idk what PDM abbreviates… at least for OW context anyways…

Zen and Moira can easily rack up more damage than most DPS while still healing their team. Not sure what this weak support thing is all about.

For example, I’d even say Zen “the most vulnerable support” has advantage against Tracer 1v1s now. Since his 25hp buff and her 10% dmg nerf adds up to 21% longer time to kill on Zen. Meanwhile, he can 1-2 tap her.

Yes I read and understood it. He mentioned it here.

The eyeballing you wrote “about” 24 which isn’t analysis of data in anyway.

Over buff data flaws you claim don’t apply in this case as it’s the sum of all heroes pick rate * winrate. This factors out win/loss dumping you get on certain heroes eg sym.

Saying direct explaination is a professional way of communicating. Isn’t claiming I or anyone is.

This is Strawmanning.

Do also keep in mind while I do enjoy the rabbit hole we’re going down this still wouldn’t change that your skill relative to your SR isn’t effected the variance players you get in your game as there effect averages out.

the average of the absolute SR changes of all the data points they had for SR between 3600-4200 was 24.50607287.
And even if you were to take averges of gains and losses separately you get:

  • average SR gain in (3600SR,4200SR) = 24.53642384SR
  • average SR loss in (3600SR,4200SR) = -24.45833333SR

so idk what you’re on. the data + plots + analysis was in their forum post. if you actually understood it, idk what you’re questioning here.

And in terms of my arguments, I even gave you a detailed explanation exploring both scenarios of the individual capping out in SR for [3000SR,4200SR), [4200SR, infinity), and [0SR, 3000SR), i.e. all scenarios were covered and explained.

it does because if the logic of counting wins is messed up, obviously any WR that is calculated from such a count is also messed up. And obviously, it doesn’t matter how you do weighted averages via pickrate, it’s not going to fix the fact that those WRs are messed up.

I already explained to you like 2 or so replies back how your poor wording + punctuation lead me to believe that you were claiming you yourself was using a “professional way of communicating”. Here’s the explanation again as you obviously missed it:

idk how you still fail to see how I’ve been saying in a proper “infinite game” scenario, you straight up can’t correlate WR with hero individual impact at all because everyone who’d cap out at the same SR will have the same WR regardless of whether they used high or low impact heroes, except for a few edge cases which you can’t realistically distinguish from the non-edge cases anyways as explained before.

Do you understand absolute SR change factors out WR as it takes into account of the SR of the players in the game and alters the SR change?

If you just go by SR changes for a
Win you average = 21SR
For a loss you average = -25SR

The guy said it himself… “In addition to Elo type effects for high rated players (in which they gain less SR/MMR because there are no fair matches available), there is an extra SR debuff for high ranked players to prevent them from getting to the SR limit of 5000. The concern here is that players would start to pile up at the limit of 5000 and break the leaderboard. This debuff starts at 4200 and gets progressively stronger as the player ranks higher”

Overbuff I explained how that’s factored out by factoring in all heroes the way it handles wins cancels out.

There are other things that could mess the data up… growing/shrinking population, meta changes ect… that said the trend is there in 1 week and upto 6 months.

You misread it. Np.

Ps… If you work in stats fields do you want to explain how many significant figures you should go to if a variance is ±10ish and you only have 700 data points?

I have the fear of jumping into this conversation, but I think I should.
ok, so, here is how it works.

The difference in win rates across the ranks is because they have done 2 things.

First of all, they have put a bell curve on the SR people have. There isn’t a lot of people in Bronze, and GM.

There is more in Silver / Masters, but not a huge amount more… etc.

So, if you are in the low ranks, CHANCES are, your team is going to be slightly lower than the other in total SR, simply because you are in it. because SO MUCH more people are playing slighter higher than you in the rank range the matchmaker can get, than below you, because of the bell curve.

The effect is higher as you get to the more extreme parts of the curve.
And yes, the system does take that into account when you gain / lose SR.

Anyway, yes, the win rate doesn’t stabilize to 50%/50% at all ranks because of this. Low ranks are stabilized at less, and higher ranks are stabilized at more.

It isn’t that the matchmaker isn’t trying for 50% win rates, it is that given there is a restriction of people it can get, it just can’t.

Anyway, I don’t know which person this is helping with their argument, but this is how it works.

Weirdly enough the effects would be less, if the matchmaker was told it could use people a certain distance in MMR, rather than SR, but, the restriction is SR based.

I think it is elegant, in that people in low ranks kinda expect to lose more games than they win, and the ones in higher ranks kinda expect to win more games than they lose, and this does that.

Which is very cool (but perhaps not a designed feature, but emergent behavior)

Anyway, the results are not linear, and given SR changes are put in place, it doesn’t really matter that it does this, since it is pretty fair for the players.

The win/loss sr calc is not at ALL obvious how it runs, because it tries to maintain the bell curve, so, your SR wins / loses can be effected by people’s accounts deactivating, new players, people making smurf accounts, etc.

It would be nice if it was just based on what was happening in your games, BUT… people leave from the lower ranks at a higher rate than from the higher ranks, so, it can get pretty impenetrable.

Also remember Blizzard ALSO is want to push peoples SR around by punishing them for leaving games, etc. And if you had a standard SR calc not taking into account the shape of the curve, eventually everyone would end up in bronze, as it pushed the average further and further down.

So really it is complex, and irritating.

I mean, yes everyone DOES get more impact from the smaller amount of players, but to Zephrin’s original point.
Mercy never did have a lot of impact to begin with, so making it scale up linearly, still doesn’t give her that much.
I think the “lets punk the matchmaker by abusing Mercy’s lack of impact” hack would still work for OW2.

When you can LITERALLY abuse the matchmaker because a hero has a massive lack of impact, then you know… there is a problem :slight_smile:

1 Like

aim lab is free yall

First off, absolute SR change was simply talking about the absolute value of SR change i.e. |SR change|, i.e. |+26SR| = 26, or |-39SR| = 39.

Secondly, have you looked at the data they used that they made publicly available in the google sheet? the points in the data included cases where the there is a difference in team SR and whereby the player’s SR isn’t necessarily very close to the team or enemy team sr.

3rdly, where did you get those averages? I literally calculated the averages for 3600-4200 before as per the scope of your earlier point.

4thly, you seem to neglect that the effect of SR diff between the individual and the opponents and even the individual’s SR diff with their own MMR actually makes them tread towards their “true” SR more and as they get closer to their “true” SR, their gains/losses are more towards the absolute value of 24SR change. Which treads to that 50% WR equilibrium state “at infinite number of games”.

You may go “but what about people meant to be >=4200SR” in which case:

  1. not everyone is supposedly that high in “true” SR which obvs makes your “model” of “look at high WR to identify high individual impact heroes” highly questionable in itself
  2. it wouldn’t really matter as explained previously because both low and high impact heroes will still reach an equilibrium SR, and no-one can really observe “true” SR of anyone to really say “that person should’ve stabalised at a higher SR with X hero” esp when it’s entirely possible for someone to stabalise where they’re meant to with a low impact hero if the impact of that SR gain debuff isn’t greater than their hero’s drift rate/factor.

uuuh no, if it’s miscounted, it’s miscounted. i.e. if you counted more or less than you should’ve to the point where you can even get negative counts due to an inherently flawed counting logic (i.e. not like it just miscounted a negligible amount in a once off, the inherent logic seemingly is off), averaging regardless of weights doesn’t fix such a flaw in counting logic.

I mean practically speaking how many significant figures or how many decimal points one should use entirely depends on context of what the numbers mean and what they’re for.
if you’re trying to question why I have so many decimal figures in the averages I gave, it’s simply because I was lazy in cutting them down/rounding when I copy-pasted the values into the reply.

I agreed mercy has a lack of impact. Lots of data proving this one case is ML7’s data on supports unranked to GM.

I disagreed there is a disconnect in her skill in pocketing tho as in the context of choosing who, how, and where to pocket there is individual impact.

Oh it is… how did you even get…

  • average SR gain in (3600SR,4200SR) = 24.53642384SR
  • average SR loss in (3600SR,4200SR) = -24.45833333SR

long story short:

  • Rezeak thinks one can measure the individual impact of a hero by taking their WR across an infinite number of games.

  • I’m saying that’s BS because more or less most people who’ll stabilise to the “true” SR “after” an infinite number of games will essentially have a 50% WR (they stabalised so obvs they’re going to hover there; stabalise because they will lose against opponents truly better than them, and vice versa for opponents truly worse than them) **esp when they practically assume equal likelihood of good and bad teammates since he says “WR across infinite number of games will cancel out the effect of teammate’s skill diff”

    • and it’d be that way regardless of high, low or 0 individual impact hero being used esp due to that latter assumption.
  • Rezeak thinks Overbuff data with some higher ranks having generally higher WR% on all heroes disproves that despite me pointing out how ovebuff’s win counting logic is broken with how it has previously reported literal negative values for WR before.

This isn’t what I said. It takes the sum of all heroes (Pick rate*winrate)

It’s not some it’s all heroes and looking at each rank to see a trend.

This is straw-manning.

I have an alt which I do “experiments” with. Try and play with different styles etc.

I pushed it WAY WAY higher than I have ever gotten on my main by abusing the matchmaker and Mercy.

I was trying to come up with a method to measure impact, which Blizzard could apply, and once I came on an idea, I thought… this is actually testable.

The test was, I would in each match try to hook up with the highest ranked DPS player in the match, totally regardless of how well / badly he played, and hard Mercy pocket him next match.

The theory was, I would be QUITE a lot lower in SR than he was, and would drag the match SR down, and because I would be “low impact” and he would be “high impact” then the team would be playing closer to his SR than mine.

Rinse / repeat. No real brain being applied, just boost the highest ranked DPS player, and try to move to the higher ranked DPS player as a duo each match.

I got 2 ENTIRE ranks above my peek SR doing this. The value I got by being lower SR and giving them easier matches was more than Mercy was going to ever give any other way.

It was pretty horrible. But it also gives a way to measure impact :slight_smile:

If a hero’s win rate is boosted if they are the lowest ranked member of the team compared to the boost from normal if they were the highest ranked member of the team, then they had less impact.

And the reverse was true.

Like I think the BASE argument once you get outside of trying to talk SR numbers and the like that Zephrin is making is a pretty good one.

You can argue how SR works, and matchmakers do things, and various calcs, but you know that Zephrin is right that supports tend to be low impact.

Now… is that a problem outside of Mercy? I’m not sure.

But like, I think the problem is, can we get them to queue? If they FEEL like they are useless, then they won’t, and them not queuing is a problem.

But it looks like you’all have moved on to an argument about the mechanics of the SR calc system, so I’ll make my exit, because it is impenetrable. Especially after Sr resets.