I would state if like this:
“I cannot use a sample population because the population does not reflect the reality. If hotslogs used it’s MMR to predict winrate it would be very inaccurate.”
If your model cannot accurately predict winrate, how can you assume it can track MMR properly?
I got 3 outliers looking at a mere 6 of my HL games from my last season.
I have 35 games total in that season. I’m guessing I would find 10-15 “outliers” examples.
In the game mode that should have the best correlation of hotslogs MMR to real skill, as it is draft and solo queue.
I could be an outlier in this, overall, but I suspect not. Although I cannot of course sanely prove this.
This question gets asked of Blizzard constantly, and every, single, time, they answer it, it’s the most weird, slimy, lame, dodgy non answers you get compared to everything else they answer where they don’t treat us like idiots.
Someone else can jump in and correct me, but the best I ever got was that we can’t handle it.
They openly admit that not all their matches are balanced, and that’s why matchmaking windows increase, why some populations have more trouble.
The problem with you however, is that you create conspiracies based on your most recent memory, and find “evidence” around that confirmation bias, and don’t ever take into account the bigger picture or simple realities, like you know, someone has to lose, or that there are new players, or you can’t magically create players out of thin air to then queue up when you want them to to have the perfect allies and enemies to shape to your extremely sensitive whims. You also almost strictly play QM which is probably intended to have more lax matchmaking because of all the other factors such as grouping and hero choice.
Simply put, the matchmaker can’t ever be perfect because it can only match what is available at any given time. However, they can still show us our freaking MMR.
Nice of you to cut the other part of that comment to be able to make a snarky comment yourself LOL. Dishonesty at its finest.
Yeah, but the MMR spread is all within 400ish MMR. We have been subjected to games with Top MMR being in the 2500 ranges and then a random 1000-1500 gets put on our team and that person is usually the worst on team by FAR!
Your further games prove exactly the same thing (Even when you cherry picked the games that “prove your point”, they actually prove what most of us have been saying).
Just look at how all the games you posted, the Winner had a significant MMR advantage 200-400 and none of the “bad” players who did “good” had MMR that was way lower than the team average.
Also, a Player with a Lower than average MMR on a Team with a higher MMR than the other team will naturally do better so your argument is invalid.
I’ve been playing since Alpha on 3 different accounts (Main, Smurf and Practice). I also strictly play ranked and consider QM to be a terrible game experience.
I’m not asking for the Matchmaker to be perfect, but SOMETHING needs to be done about it. I don’t even care than the MMR system is wrong for most people (To the extent that we can know). But I want games that are not one sided. I want games where everyone is near the same skill level. Most games, there is clearly ONE or TWO players that are weaker than the rest by a HUGE GAP. That is what I want to see gone, that GAP needs to dissapear. I never want to be in a game with someone that has half my MMR.
I propose that all games in ranked must be within 200MMR from the team average. So if team Average is 2000, players are all within 1800 and 2200. Problem solved.
I do not care if games take 20 minutes to find. I would rather wait for a quality match than to have faster but one sided matches (like we have now).
For people who want faster games, they have QM for you guys.
So, in other words, people like Droth will continue to complain regardless even in your dream scenario.
I don’t know, I didn’t play TL that much, but I’ve always found ranked to be pretty good at minimizing the rank spread simply because there was a max one league difference.
Even when I got reseeded to lower ranks due to inactivity, I found that the matchmaker was eerily consistent in the skill level of those players. I was frustrated because they WERE consistently worse, and when I climbed, people were consistently more skilled.
In ranked, I just don’t really see it as a problem. Now, do I get bad games? Sure. I’m not necessarily seeing that as a problem with the matchmaker in ranked though. Some games just are going to be bad even if MMR wise, it says the game shouldn’t be. I don’t curse the matchmaker for it, I just hate the game I played for awhile, and move on to the next game.
Now, for QM, yes there’s a huge barrel of fun in the type of players you get. I’ve always said that there should be hero specific MMR as the biggest wrench thrown in QM is people testing heroes they just aren’t good at.
There are also cases where more seasoned players will play with new players, and although I hate it, I bite my lip, and tolerate it because it’s QM. People should be able to play with who they want to play with out too many barriers.
Yes, I was shocked you actually had a kernel of accuracy in your post.
Okay. So why wasn’t this being a reasonable spread brought up?
according to hotslogs, going 2W-8L in placement last hero league season was enough to adjust my MMR down over 600 points.
I’m looking at your match history on hotslogs for HL, what should be the mode where MMR differences mean the most, and I’m not seeing it.
Could you bring up some example games so I understand when exactly it hits poorly?
thank you!
I spent 5 minutes on my computer, and another 5-10 minutes on my phone while in a car (I was not the one driving) to find these.
I probably looked through about 20 games. And discarded any games which I played in a group of 3 or more people.
You realize the two games you quote (and the third one I mislinked, doing properly now: https://www.hotslogs.com/Player/MatchSummaryContainer?ReplayID=164040169) have one game with a ~225 MMR advantaeg for the winner.
And the other two with super close MMR?.
The other games you didn’t quote had about 100/200/300 MMR difference for winning team. Correct.
I can tell you, being friends with someone in the 300 MMR game, we were winning until I decided to be dumb and threw the game at 17 minutes. Our lowest MMR player, Sonya, played pretty well. And I, our highest MMR player threw the game. =[
These details matter, because it was a close game, with a 300 MMR gap. Until the “best player” on our team per hotslogs threw the whole thing.
context matters. I’m giving it to you, but you’ll probably ignore it.
But here I just described how my “300 MMR game” per hotslogs (which I imagine was within 200 MMR on team average internally) was a very close game until the “best player” on one team made a massive mistake and threw the game.
You don’t seem to get this context.
That being said, I would like an option to get higher quality matches with slower queues.
What’s something. Based on the video I transcribed a massive portion of, it seems to be working just fine.
That’s what the game does. The problem is there is variance in players skill. The game has a variance factor which it does use, but effects outside of that can happen.
BRING EXAMPLES. Have the examples uploaded somewhere they can be watched. State who the “weaker” players are. Let people watch and decide and try to figure it out.
Put some actual effort into this if you actually think it is an issue.
You won’t be, unless you’re in or against groups.
HotSlogs rated my HL MMR at 1457 when I placed into platinum 2, after not playing for nearly 6 months.
Hotslogs isn’t accurate. Get It Through You Thick Skull. Or Thin Skull.
Now, to be clear, I do think MMR should be shown, if only to get people like you to stop crying, or at least find something else to cry about.
also in regards to various ways to make matchmaking from someone who was at Blizzard at the time, see this post where I timestamp interesting portions and Q&A of the presentation:
please do note, the person presenting is not listed anywhere (including the HotS credits) as having worked on HotS. So while they may have had influence over the people who make/balance/etc the matchmaking system, they might also have no influence.
Regardless, it is a fascinating look into various things involving skill, matchmaking and so forth.
It does have a lot of presenters opinions, of course, but I feel the presenter is very clear what they are presenting is an opinion or not.
Aha! I meant to also tag you here somehow Drothvader. Because this talks about testing by using "fake’ players. But you need to have an accurate match prediction to do so.
I’m saying you cannot get an accurate sample from hotslogs due to it not mirroring the actual population at all.
MMR is a useless number for players to obsess over. It has next to no meaning outside of matching players together. It doesn’t move or provide you with a ranked ladder experience.
Ultimately, MMR doesn’t mean what you think it means and every time a developer breaks down and displays it to the players it’s a disaster as the developers try to warp the number into a value that actually gives players something to look at that doesn’t just sit there making them look stupid.
You think you want to see your MMR but you really don’t.
A good way to start to actually fix the fundamentals of the game a.k.a. the horrid matchmaking we’ve had for years would be to anchor our MMR in QP/UD since plenty of games are just ruined by multiple hundred MMR lower players getting placed in either team and just outright ruining the game. We already use hotslogs which has made it clear that the matchmaking funks up so much that the forced 50% actually has some merit to it so I would also like to have some client side transparency on MMR that is if the next balance patch doesn’t bomb it and I’ll never return to the game
is to give player the tools to self organize into teams where players are queuing up with friends and other vetted teammates, NOT to put any more effort into trying to fix Match Making.
FYI, Forced 50% is no longer a conspiracy, it’s a fact.
start watching from 44:20 for context
45:35 it gets to the point when matches are super even all the time after a while it actually get kinda tiring…so i come up with the opinion that its ok to allow a little bit of balance here and there and have a mix of even hard and easy matches.
Somehow I doubt ANYTHING could convince people forced 50/50 isn’t real when they are determined enough to believe it. They would claim the mmr values are fake, or something similar
False. Teaming up and organizing is not part of the default package so it’s about fine-tuning your gameplay and not fundamentals. For example if you have an organized team you can still end up having miserable time since the fundamental aspect of matchmaking takes priority over that which in bad shape can give you too easy or hard games. This is highlighted when the matchmaking fails to balance teams due to relying on inaccurate averages without proper MMR anchoring. At that point it’s a fundamental issue
But we don’t have enough proof on forced 50/50 one way or another since we don’t see our actual MMR and don’t know how much it effects the gameplay. If the matchmaking is proper then it should estimated that teams have at most 40/60 chances at victory + there’s always some MMR fluctuation between teammates. With these very basic facts there are at least three ways the system could “force” 50% winrate depending on how much these effect the gameplay:
If you’re on a winstreak and your opponent has lost a few games the system might be more prone to put you in the 40% chance at victory team. The system doesn’t just average MMR between players as sometimes the other team has better MMR on average and despite the fact that it could swap 2 players around to even it out it sometimes simply doesn’t. (40% chance is considerably bad btw)
Players with high win%s might be more prone to get the low end of MMR between players as their teammate which can effect your chances to win depending on how bad that player is and in HOTS if that player is considerably below others that team will usually always lose
You might be prone to get more low end 40% chance at victory games instead of just about 50% just for having a higher win%
Also you should check Shaden’s post above that points out that HOTS having forced 50% through these ways is very much plausible:
Incorrect. This is a team game. Team games require teammates. Nowhere else in life where people are playing team games is there an expectation that single players are placed onto teams with different teammate every game they play. This is the part of the paradigm with online gaming that’s broken, not the other way around.
Stop trying to fit a round peg in a square hole. It’s not going to work.
There was a post by blizz just after the release of 2.0 where they were redoing the mm system again and including pbmm. One of the quotes was something like “the mm system before these changes was almost solely based on winning and losing.” Which signals to me that mmr may have never existed until relatively recently and even then, we have no solid evidence that it really exists. Blizzard has given us no real proof it does besides simply telling us. Priests say God is real, but what evidence beyond word of mouth is there really?
I want solid, real proof that mmr actually exists. Just saying it exists doesnt prove it exists.
Yes, because that’s how MMR has been generated for years in a variety of online and real world competitive sports. Your performance determines who you play. This is a no brainer.
The problem is that so many players have this false belief that MMR has direct relationship to real world abilities. Their expectations are what’s in error, not the system itself.
so many players have this false belief that MMR has direct relationship to real world abilities.<
Should it not? I mean if it doesn’t then what the ef are we playing this game for? Games measure skill now, otherwise games would be solely for kids and their grandmas. If this game doesnt have any way to measure real world skill at the game then…should we just stop playing it?
