I think I’ve had reasonable discussion in this thread. The people who say MM isn’t perfect but it’s fine the way it is can discuss why it’s not perfect with themselves. I do appreciate anybody who has a willingness to discuss. All I want is good games in my ELO. I don’t want stomps on either side. Being a Master Overwatch player or above does nothing to enhance my life. What good games do is let me have more fun. And that’s what I want. So… cool guys, anyone who argues against me… if it was a good discussion. Thanks. If it was an argument that had to go to extremes, may you meet someone who cares as passionately for a video game as you do.
I don’t think anyone is saying stomps are good or that we shouldn’t want good games. My main point has always been in these discussions that all it takes is one player having a good or poor game compared to their average to turn an otherwise balanced match into a stomp. And there isn’t really a way to account for that.
The only real way to make “close” matches consistently is to join a scrim or PUG group.
I would 100% accept stomps and streaks if I knew the matchmaking was just random for my rank. No hidden analysis, interventions, or cherry-picked teams.
The way it was explained to me by people on these forums (and this thread in particular), is that random games (SR only, no MMR matchmaking) would regress to a rank average. So you are totally accounting for variance and corrupted matches, which will be smoothed out by the randomness. It’s also good for player retention, because by just spamming games you are promised to overcome the random effects and being better, worse, or average for your rank. Thus, moving to your proper SR label instead of getting rigged on by hidden mmr.
Most likely business reasons. Lower ranked players getting repeatedly stomped by better players, while a fairer system, is worse for business from Blizzard’s perspective than the alternative: weaker players getting some wins off the backs of their more skilled teammates. This keeps those players playing longer, more interested, more positive about Overwatch, keeps them buying lootboxes etc. There are very strong incentives and benefits for “rigged” matchmaking that add up to revenue and increased usage metrics for Blizzard.
You are certainly in the minority of people who have considered whether the 50% odds thing make sense. You may be someone who doesn’t care too much about their rank in Overwatch, or you may be in a rank above Platinum where MMR and SBSR are not factors. I am in the subset of Platinum-and-below players who want the same rules to apply across the whole competitive board.
I don’t believe the MMR system is trying to disrupt individual players winning/losing streaks. I think the disruption is a byproduct of handicapping.
Aha, well your findings support my theory of Match Making Rating. MMR guarantees that the winning team does not have all of the best players, or even a large proportion of them. MMR creates an even distribution of skill across teams in each match. Depending on the relative skill of participants, that distribution can be done at anyone’s expense, with the outcome of the match hinging on a single participant in many cases.
Definitely. This becomes apparent when you read the patents of the game holding company, Activision Publishing Incorporated (featured in my video).
I have been as low as 1500 and as high as 3200. I have two accounts that I played basically the same and ended up in roughly the same spot after playing for a bit (high plat/low diamond). I played less after I hit diamond because I never thought I was that good. I view myself as rather average.
Took a break from the game, and I’ve recently returned as a rust bucket that would be lucky to maintain plat if I played comp.I mostly just use QP in small bursts when I play at the current time as a matter of convenience.
I would agree with this statement. I don’t think the system care who wins or loses individual matches - those come naturally.
What I was specifically referring to on that part is how the confidence factor in MMR calculations likely influences SR gains.
For example: If Joe is a 2,000 SR player for 100 games but goes on a 6 game losing streak while performing statistically strong, it is reasonable to think he may lose 20 SR for the first two but then lose 15-16 SR for the last game. This would happen because the system believes Joe to be capable of 2,000 SR and he is performing the same, but he must incur an SR loss due to being on the losing side of a match. He may gain 21-25 the next win he gets as the system tries to more quickly get him back to the 2,000 SR range sooner rather than later.
My stance on the topic is that not all of the best players should be on the winning side. This is for 2 reasons:
It makes the match a better and more fair competition.
The matchmaker learns more valuable information about the players by systematically putting individuals that it thinks are equal against each other. In theory, this helps match quality in the longer term for the most people.
The matchmaker ignoring SR and trying to be the smartest person in the room with its MMR is what ruins the game. The fact that two players can have an identical SR but the matchmaker thinks one is better than the other for reasons is just silly. That means one player has to work harder than the other in order to gain the same reward just because the MM likes their stats more.
It also means that someone that is naturally 1700 but on a win streak isn’t weighted the same as someone that is naturally 2300 but on a losing streak for a 2000 SR game.
The harsh reality is that there is reason to believe that 1700 Steve isn’t as good as 2300 Tom even though they both have 2000 SR at one point in time.
You might not like the notion, but there is purpose behind it that is trying to make things fair. It’s meant to reduce volatility in match quality.
Edit: Whether or not it should take that approach or succeeds in that goal is obviously up for discussion.
FWIW - I was trying to add to your statement and not offer a blanket rebuttal.
Not really, in fact I would personally rather have Steve on my team in that situation. But as long as their SR’s are the same let them play it out and see who deserves to climb and who doesn’t. The matchmaker shouldn’t be trying to predict where it thinks you belong on the ladder. Don’t give Tom worse teammates just because you think he plays like a 2300 and Steve is boosted. If that is true then Tom will climb back up and Steve will fall back down naturally without the system trying to force it. Too many Toms getting screwed by the MM while the Steves get carried is called handicapping and makes it harder to get back to where you belong.
I’m cool with testing out our SR games. It wouldn’t bother me a bit. I do think people would be disappointed in the results, though.
I don’t understand where this notion comes from. In a pure SR matchmaker or a completely random system, somebody will get worse teammates. It’s literally always going to happen.
The current system in action rewards 2300 Tom more than 1700 Steve for the scenario I presented. Tom would win more and lose less while Steve would win less and lose more if all things are equal. The assumed better player gets the better end of the stick for the scenario.
I feel like there has to be some part of this handicapping theory I can’t get through my thick skull, cause I can’t make sense of it outside of caricaturing it to the point it makes the MMR system seem like a boogeyman that is actively trying to hold good players back.
Beyond that, just on the math, I don’t see how the question of matchmaking being rigged is even a debate. I don’t know how or why people are arguing against this. If a player in the 95th percentile has a lower chance of winning a match because they’ve been placed into a match where their odds of winning are “between 40-60%”, then yes, it’s a FACT, the match has been rigged against that player if their odds in a completely random grouping fall outside of the range of 40-60%. So yes, matches are rigged for any players who otherwise would not have had between a 40-60% chance of winning that match. These are facts. This is not opinion. This is not open to debate.
Yes but it shouldn’t always be the same people getting the short end of the stick. The ones who are perceived as better players than their actual SR because of a hidden unknown rating.
How does Tom get rewarded more? Both Tom and Steve are 2000 SR. But the game expects Tom to play like a 2300 and Steve to play like a 1700 based on their MMR. So to compensate in order to make the game “fair” it will always give Steve slightly better teammates, despite the fact that they are equal SR. The assumed better player does not get a fair deal in this situation.
Handicapping just keeps players in a slightly lower or higher rank than they belong for longer than they should. It will not stop a GM player from climbing out of gold or hold back anyone for eternity. But it is obviously a thing for anyone who has played the game long enough to witness for themselves. It makes you grind more which is in Blizzard’s interest to increase player activity and to artificially adjust your rank to prevent players from feeling stuck forever. But overall it is unnatural and unhealthy for the game.
Tom is valued higher than Steve, as he should be. His track record is proof it’s reasonable to assume he actually is better. His SR gains are adjusted to recognize the fact he is, indeed, a better player.
So, when putting the team’s together, the teammates for Tom are worse relative to himself, and Steve’s are better relative to himself, but the teams as a whole are equal. Tom has an equal to Steve, and Steve has an equal to Tom. It’s an identical problem to running matches based purely off SR for anyone, the relative differences within a balanced team mean the best in any match has to carry more weight compared to the weakest.
That is an unavoidable reality of having fair matches - even conceptually. The only difference to a pure SR system that balances out the numbers is that SR is visible at the cost of being a more volatile metric.
It makes you grind more because the goal of high individual match quality is valued over letting ranks bounce around more freely.
Whether or not it accomplishes that or feels good to participate in is a debatable topic.
This might be reversed. One part of the the MatchMaker™ discussion had to do with the player who had a slight advantage (in a 51/49 scenario) would lose more on a loss and the other side would gain more for a win. Not sure if this directly connects but it could cause confusion.
Tom’s teammates will be better than him in some games making Tom the “bad” player.
Likewise, Steve can also be the high MMR player in his SR pool and will get lower MMR teammates.
It reduces the volatility of the player’s rank. When the player demonstrates consistently improved performance, they will climb. Having some good games but stagnant average per 10 stats will keep you in the same place. When your average stats move, so does your MMR.
This is true under normal circumstances, but doesn’t smoothly apply to my specific example of looking at two people on streaks that perform at the level they’re intended to in a 50/50 example. It’s a related, but different can of worms all together.
Your post does illustrate exactly how nuanced and complicated this topic can be, though. I’m glad you pointed it out.
You are making an argument for a pick-yourself-up-by-the-boot-straps, individualist philosophy. I don’t see how any of this connects with my argument against algorithmic handicapping.
It’s not.
The Matchmaker doesn’t control your actions in the game, it only controls the selection of players for matches and the arrangement of teams. But I have argued that such controls are profoundly inappropriate for ranked matches. Let’s circle back to this first for a concrete example:
Supposing Ana is the most skilled player in the match and you (in this scenario) are the second-most skilled. In that case, the Matchmaker virtually guarantees that you will be on opposing teams. If the disparity in skill between Ana and the other players in the match is large, then you may have also have the third, fourth, fifth, sixth, and seventh-most skilled players on your team to offset the difference. Consider the lopsidedness of that, and the disproportionate burden being placed on one player (Ana) to artificially create 50% odds for a match with 11 other players.
That is absolutely not true. The PBSR system does not even award rank when the match is lost. And if you appreciate the nuances of gameplay as you say you do, you should understand that the system is blind to the real turning points of battle. Victory and defeat hinge on the actions that each of us take in our matches. But they are also probabilistically related to the statistics of each player.
By maintaining a perfect state of knowledge about every player’s gameplay statistics, MMR does inform the Matchmaker about how to effect a handicapped match effectively. But that does not put the PBSR system in a position to decide each player’s merit. Players merit should be proven in the outcome of the match, and nothing else. Players should be awarded and deducted SR based on the outcome of the match, and nothing else. These automated reward systems are not just imperfect, they are perversions.
Mmmmmmm…. It connects to the over arching philosophy that I and the rest of us are pointing out that, consistency is what determines abs separates truly great players from the average ones.
Are you determining skilled through SR or MMR? Because algorithmically these are very different (as I’m sure you’re aware).
You’d have to work very hard to prove this based off of all information available (from the patents AND player experience; and I don’t mean the hard stucks with big egos).
Indeed…. So referencing my earlier question, it seems like you’re discussing MMR based matching.
If I (a masters DPS smurf) go into a game in say, 2100 SR (bare with me and assume I’ve already played 2 or 3 competitive games please), the system clearly knows my MMR is not that of a low gold player.
Okay…. Well now I have a player playing in gold who doesn’t belong here. But I’m constrained to find players from 2100 to 1100 or 3100. However, I need to do it in a reasonable time.
Mr. Smurfs MMR is so high, I cannot find another individual player to offset his MMR. But if I put this tank and these two supports on the enemy team, the likelihood of the enemy winning is now above 40% (so the match is made).
If the tank and 2 supports dig deep (with their teammates) and defeat Mr. Smurf, Mr. Smurf only loses (I’ve seen as low as) 10 to 18 SR, depending on (oh idk), dmg/10, elims/10 compared to deaths/10 etc. etc.
That above scenario is the PBSR system preventing the smurf from just camping in low gold and rolling kids.
Conversely (before I stopped playing Overwatch competitively) same season I hit Masters my first time, I climbed a gold bordered account from 13xx to 2371 and it took exactly 20 games averaging 50 SR per game. Some games were 45 SR, some games I got 65 SR.
Is that not the PBSR system saying “okay dude you’re not a Bronze/Silver/Gold player, get the hello out of this Elo….”
So coming back to this discussion, assuming no smurfs in a game, the SR and MMR variance is honestly relatively average. IE there’s rarely exceptional players playing in these games.
The 2nd/3rd/5th best players have the potential to be the best players in the match if they work together as a team.
What if the 5th best player is a Symmetra main and the 3rd best player (now on the enemy team) is playing Mercy and keeps getting tasered to death by the turrets because that player (who is usually a good player) is fatigued/emotionally exhausted and generally just not completely focused that game?
Well the third highest MMR player isn’t contributing the third highest performance in the match.
I think these breakdowns would make more sense if Overwatchs after action reports were as good as valorants
But it definitely does!!
If you put me on a low Elo account, I will gain more SR for winning a game and lose less SR for losing a game. I’ve tested this and I gave you an example above.
Okay okay…. When I say it rewards rank, I mean SR gains/losses. It won’t promote you an entire rank and it shouldn’t…. It also does Demote entire ranks and it shouldn’t.
OMEGALUL…. I very much understand this, babe. I was hardstuck platinum for 5 seasons before I finally hit Diamond.
Yes!! We agree!!
Yes!! We still agree!!
But think of a Gold hitscan player statistically compared to me (a Masters hitscan player).
If you look at my stats in Masters games, I’m putting out (across an hour of gameplay) 2.5 to 3 kills/death.
If you look at the gold player at 2100 SR, he’ll also be putting out 2.5 to 3 kills/death.
But what happens if we switch places?
Well now I’m being matched against players who aren’t used to playing correctly and I punish them.
Now the gold player is playing with teammates who know what they’re doing and will manage to pull through as long as he does “just enough” not to negatively tip the scales. Does that really mean he belongs in Masters?
He can’t consistently get solo-elims or final blows without extreme help from his teammates.
But now, you have me in gold killing off the supports/DPS and damaging the tanks, and my tanks/supports and DPS counterpart clean up the mess on aisle 5.
So I may be dropping 7-8 elims/10 and only dying once or twice per game (in gold) where as the gold player gets 2-3 elims/10 and maintains a 2.5-3 K/D.
Statistically stating, these players are clearly not the same (as you know).
Now…. Being top fragger in a game like Overwatch doesn’t matter if you don’t help your team achieve the objective.
Just because a player is good at their role, doesn’t mean they’re good at Overwatch, right?
I mean…. The variance in skill of a gold junkrat, compared to a gold Hanzo, compared to a gold Mei…. They’re all so different.
So MMR is in place to try and give each player (hero preference enabling) a chance to have a decently matched game, which is why MMR and PBSR is also linked between every player (in whatever rank) that plays the same heroes. This part and some of the other parts I agree with you that it gets a little…. Odd…. But it’s clearly in the interests of allowing people who wanna play their precious doomfist/Genji a chance to succeed.
Honestly I think this would lead to more unbalanced games bottom to top.
In a game like CSGO (also uses SR and MMR BTW) where everyone has access to the EXACT SAME equipment and tools…. A pure win/loss meter would make a lot more sense (and I can’t tell you why they don’t do it that way in CSGO because I’ve only played it a couple times)
In games like Overwatch and Valorant (if you haven’t sat down and tried to enjoy Val you’re missing out) where each team has COMPLETELY different heroes, each hero has different abilities, each ability affects the game differently, not having MMR would literally destroy match quality.
———
———
It’s a good (and valid) analogy, really.
Though pretty much hero balance destroys a lot of our discussion tbh.
And the Matchmaker knows which players are consistent based on their MMR. Individual players do not have consistent odds to rank up, based on skill relative to their peers. Because of MMR, every player has 50% odds in every match, regardless of consistency and skill.
SR and MMR are both supposed to be measures of player skill. Of course they are different, but they are connected systems.
No I don’t. This is proven by an explicit statement about MMR from Lead Overwatch designer Scott Mercer. You are either ignoring or failing to understand Mercer’s statement.
Wrong. Your MMR is rated as that of a low gold player, because your performance can only be rated relative to that of your peers. Your MMR in gold can be relatively high or relatively low, according to your performance. By the time you reach a higher or lower rank, your performance has already been reassessed.
No such assumption is necessary. The law of averages applies to everything, including the performance of smurfs.
That’s not true. I see huge differences in skill between players from Bronze to Platinum. The MMR system accounts for those differences, and the Matchmaker schedules matches accordingly.
You’re leaving out the 4th, 6th and 7th best players, who are also probabilistically guaranteed to oppose the best player (or sometimes the 1st and 2nd are together against the 3rd, 4th, 5th, and 6th, depending on the presence of groups, or wide skill variance, or any other combination that will produce near 50% odds for the match).
Anyway, the question about individual players’ (or even groups of players’) chance to “be the best players in the match.” It is about players’ chance to win the match. Their performance and individual stats do not really matter, with regards the match’s effect on each player’s Skill Rating. These measurements only matter to each players’ Match Making Rating, and they are being scored on performance individually, not as a team.
No, PBSR does not compensate for the skewed match results that are caused by Match Making Rating.
We agree on many things, but not the workings of the system. I think your perspective on this is different because you play above the rank of Platinum, where PBSR and MMR do not exist. I have only played Platinum and below.
