Hey, do you have discord?
I would really like to discuss some things with you if you dont mind.
1 Like
Not necessarily. Optimally all players would have equal MMR, but for every variation the system tries to have a counter variation in the opposite team. This means that the opposing tank can be higher/lower MMR and any of their teammates can be higher/lower on the contrary sense regards to that. However the threshold, for a match to meet the requirements for the match to be able to even be initiated, is 40%/60% estimated chances at a victory. In other words the opponents can have relatively higher MMR up to a degree where their estimated chances at a victory is 60%. As long as that requirement is met the game can be initiated regardless of the variance in players
3 Likes
I feel that you may have misunderstood fundamentally how the MMR system works. Sure the goal is to give you a “handicapped” as opposed to random match but that’s exactly what you want it to do.
It’s really not some shadowy process as you imply, obviously it’s a specific implementation for Overwatch but it’s based on some solid principles as a derivative of the elo system for team play. They talked about it like it was no shadowy secret because it really wasn’t and some insight into it didn’t exactly break any of the fundamentals, it just showed us their specific implementation.
The main difference between SR and MMR is that SR is a scaled version of your MMR designed to make convenient brackets for rewards and provide a punishment for leavers,etc without actually affecting your match rating. If you punished MMR for leavers like you do SR then you’d lose some accuracy in a system that can only ever be an estimate anyway. MMR in overwatch is a small floating point unit that is not very human readable by the account of Jeff Kaplan.
Ultimately, the only way to create a system that gives teams approximately a 50% chance of winning is to accurately assess the skill of those players so the skill rating you have is approximately accurate. If it was not accurate then you would constantly defy the odds until it became more accurate - literally by design.
I think it would probably be good to show people their MMR but if anyone is expecting this to change their skill percentile by any significant margin then I would have some bad news.
3 Likes
Allright, but that implies that my MMR keeps more or less the same, while my SR gets tanked by leavers, throwers and smurfs.
At least this explains, why I don’t get easier matches when I lose.
1 Like
Well throwers and smurfs will calculate your changes as though they played normally. I would assume however that leavers would affect your MMR less, if at all, than a normal loss.
SR will effectively “magnetise” back toward your MMR so the further your SR from your MMR, the greater the change until they get very close. It usually doesn’t get that far away though. The biggest time this happened was when they had SR decay into players above Diamond, you’d have 3K SR and win like 200 a match because your MMR was so far away from your SR.
This is why I’m trying to explain that your MMR is really not that hidden, ultimately your SR is very close to your MMR at all times. Without decay in the game now the main things that can desync it would be leavers/disconnects.
Ultimately, MMR (and by extension SR) is trying to track your progress over many many games. I think Jeff said that the system really isn’t confident until at least 50 games, even then it’s always an approximate rating so it can flux quite a bit and still be functioning.
In my experience the main problem players have when assessing their rank is they look directly at mechanical skill with a hint of positioning. If they go down 800SR, they are going to run into plenty of people with the same mechanical skill as 500SR+ above but that player is missing a different aspect of skill required to climb. They get into a situation where that player does the mechanical thing fine and think “oh clearly he’s misplaced” but neglect the fact that this same player maybe consistently rages at their team or doesn’t understand ult economy.
MMR doesn’t care if you are terrible at aiming; if you consistently win against people that can aim because you shotcall or always backcap then you are equally valid at that rank and if you can aim super well against people who can’t but keep letting them get the objective then you’re equally less valid to climb.
2 Likes
I do, but please reach out to me by email first to give me an idea of what you want to discuss. Address is listed in my YouTube channel’s contact info.
No, I do not want Competitive Play to be artificially, automatically balanced. I have played games where I regularly and voluntarily took a handicap, by joining losing teams or opposing other players who I knew to be skilled. But when rank and reputation are on the line, fun is not the objective. When I’m being rated on my skill for all to see, I want victory, and I want a fair chance to win according to my skill.
I have no interest in the match being handicapped, because I know it is being handicapped at my expense. I am vastly better than the majority of players I get on my teams, in every class and character. I want a random match and a fair chance prove my merit. And then mark my words, I will drag Blizzard’s precious “E-Sports professionals” down from their pedestals, together with other true competitors who are being suppressed by MMR.
More later, thanks everyone for your interest in the discussion.
1 Like
I think this parallels my experiences with competitive match making. For me, it seems imbalance comes from using a team-wide MMR versus a per-role MMR. Having a Gold tank playing against an opposing Diamond tank with the purported compensation tucked into one of the supports feels unfair at times.
I wonder if per-role matching closer to 50% would result in games that feel more balanced on average. That would likely exasperate queue times though.
1 Like
I agree with this whole heartedly. I also think the que time thing is why it hasn’t been done.
You actively have this across the aggregate of games. While in one instance you may have a game that can’t be won, you ability to display your personal skill and climb on merit is present entirely by the exact system you are criticizing.
MMR is rating where it thinks you are and across multiple games would effectively “handicap” just as many of your opponents as you.
If you truly believe that they somehow decided that your account was designated to not be the correct rank then you could just create new accounts until you got a “lucky” seed and got into the high ranks. By your own admission you expect it to favour new players over experienced players, should be easy.
Ultimately, and I get that maybe you may not wish to acknowledge this - If you’re struggling with your rank and the OWL players are simply reliably getting to the high ranks on multiple accounts then they are clearly overpowering the “handicap” in a way that you are not skilled enough to do so.
I’ll try to point it out again, MMR is a rating of your skill, if it is too low it will put you into a game where it assumes you have a 50% chance to win but you actually have a higher chance - you will win games consistently until it is approximately accurate. (which is why highly skilled players consistently get to GM+)
If you want to climb, you have to look at how you can improve the macro outcome of your matches, not just pull off a nice play but be consistent across every game.
3 Likes
But it does this. I have another account that has consistently stayed high silver and gold… nothing great, exactly where I should’ve been… it is much newer. This account has dipped into bronze on multiple roles, while I have been able to climb this season into silver… It’s still unreliable. Am I Silver/Bronze? or am I silver gold… why would I be on opposite ends of the borders? I don’t play any different on the two accounts. And I do realize that it isn’t that big of difference, but they do play a lot different. Is it because I played more DPS on that account? and way more Support on this one? WHY do I have a better average kill rate on that one… compared to this one when I started and was complete feeding trash?
I currently have 4 different account, two are exactly where I am on this account, one is maintaining high gold placements only, and one that is on the boarder of diamond (brand new level 28 currently), if I was to play the two higher ranked account for a significant amount of time, they would surely fall to where I am now due to my skill level, nothing to do with the matchmaker, I truly believe it’s all based on a certain mindset playing on different accounts.
1 Like
I’d argue that the matchmaking is working here but the problem is how they have displayed SR to us. The idea of a gold or silver player, it gives you an identity so you gain attachment to it. Problem is, the backend doesn’t really care for the concept of gold/silver/bronze.
Another issue is that you may be considering that maybe there is a correct accurate number for your skill rating. This however is just not the case in reality, players don’t play consistently in every match so really everyone has a skill range.
For some players this range can be very wide and those players tend to see larger swings in rating.
Because players don’t actually have a specific skill rating, there are a few things done to help them get to approximately the right area. If you don’t play for some time then the MMR change will be slightly bigger (in either direction), this is the same for new accounts to help them match quicker.
There is also for players under Diamond another system called Performance based SR. In this you get additional adjustments based on how you performed that game compared to others of your rank, on that hero, on that map under that victory condition. This is turned off over Diamond and everything is based entirely off W/L to avoid people metagaming the system.
Don’t get me wrong, the MMR can be wrong for some time by chance however over extended playing it trends toward the right location. If you goto bronze and are not a bronze player then after playing a reasonable amount of games you should climb.
The further in skill you are from where the system currently thinks you are, the faster the change.
It’s totally reasonable for someone to have a 500SR skill range, I think if you get into a skill range of around 1000SR then you would be in a very small minority but it’s still valid.
You are considering things such as your average kill rate,etc - However these stats are basically not part of the equation to work out your SR. Because of Performance based SR you get a minor boost /hit (~5 SR) from your performance in a game but the real value is winning/losing.
If you get double the kills but lose the point/game then you’re still taking a bigger hit to your MMR than getting half the kills but winning the objective/game.
3 Likes
your explanation still boils down to why are we still using at least four different metrics to count one ladder. And it could be more for all we know. If wins are all that matter, great. That’s all that matters. Performance can’t account for strategy. If MMR is used to balance out the game, then why give SR a number at all. Just say gold silver plat master etc. Or don’t use MMR and give the SR total control of the match maker. Only put golds with golds Silvers with Silvers. no need for an average. .
I’m glad to have a discussion with people who are at least willing to give a reason that they think it’s ok… as long as they are willining to hear why I don’t. We probably won’t change each others minds… but it’s worth a shot.
And Sue why would you “surely” fall. You’re being placed with the people who placed there as well. if it would surely fall, then placements are worthless… or you got better because you know what teammates to follow into battle… and how to help them and the team win. If the ladder was running correctly, IMO , if you’re not doing anything drastically different, you should place in the exact area that you’re main is. I don’t think you would “surely” fall. I think there might be a chance, but it seems to me from my limited interactions on this site… that you study the game and how to be better at it. (maybe I’m assuming) so why wouldn’t you be a diamond.
2 Likes
So really MMR is in total control, it’s the real number that’s actually used and is hidden because ultimately Blizzard decided to do that.
It’s also very difficult to read because it’s a long decimal value (imagine being told your rating is -1.3454539 and it just went up by 0.00003014) easier to convert it to a less accurate but user friendly format to read.
I suspect the main reason we have split SR/MMR is literally just so they can punish players without affecting the match making. They want you to have a punishment for leaving a match that is not finished so they take away some of your SR. (Some players respond well to this and never leave as a result)
It also served as a way to decay rating for players above Diamond (this was removed when rolequeue was added) - You had to play every week or your SR would go down but your MMR would remain the same.
SR is basically just a flexible way to intentionally distort your rating without breaking the actual important value (MMR)
When you understand their relationship you realise that SR is basically your MMR for almost every scenario unless you regularly rage quit or have a lot of leaver games (but its still not that far off).
I don’t know if you were around in season 1 but it used to go from 1-100 and you’d get part of a rank and there was no such thing as bronze/silver/gold/etc. Players complained that they didn’t understand the system so they made the numbers between 1-5000 and then added in the brackets so that players could feel like they belonged to a bracket and have a bit of identity about their rank.
All the things about SR are just making it a bit easier for the average player to not need to understand how it works but still roughly know where they are in relation to their friends.
3 Likes
I was around. And I understand the concept of why MMR exists. I don’t think there is some nefarious reason behind it… I think it works as intended, but I think there is a poorly implemented system. you know what I mean? And because in the rank this account currently is in, I get way more throwers… derankers etc. Than in the gold one, so this account is MUCH harder to pull back to where it should be.
PBSR is a great concept but it goes along with the “rigging” idea (I don’t like that word) but if you over perform, and still lose… you lost 5 less points… but you still lost. It’s slower falling instead of tanking. So you go on a bad losing streak… and you end up lower than you started.
With that said, I don’t think it was bad when the game started. I think when they put things on top of the existing match maker… they broke it. And I will continue to argue for a ground up approach to matchmaking for OW2. That I would hope starts all players at the bottom to rise to their correct MMRS. and placements that mean something every season… or at least seasonal SR and MMR resets for COMP only.
1 Like
Smurfs and throwers are the problem.
Smurfs play around a fair estimation of their MMR. They buy new accounts or keep softthrowing to maintain their “low” skilled account.
So MMR can’t recognize a really good player, if he knows how to stay below the radar.
These people come to gold and ruin matches continously, because they are heavily underrated by the matchmaker.
1 Like
These are aspects that are actually outside of the remit of most SBMM systems, I don’t believe Overwatch is an exception to this.
You really have to get through a lot of games without these disruptions for it to be accurate. Ultimately all you can do is report and move on, add to avoid is also a decent idea. If you play enough games without this happening then the system is self correcting and you’ll eventually end up in the right range.
They have definitely made changes that have been to tighten up the MM since launch, I don’t think they broke it though. It’s still objectively working at its primary function of assessing player skill.
I do however think that they have made some unannounced changes with loosening the skill variance. (ie how big the skill gap between the best and worst player in a team)
Now this doesn’t break the match maker but it can make it work way slower for some people, requiring more games to get to a reasonable accuracy.
There is actually good reason to believe they would have done this as they split the playerbase into more queues - There is clearly a priority to get players into games in a somewhat reasonable time.
If the player base is lower right now as a lot of players wait for OW2 and those that remain are split among more queues than any time before, then it’s entirely likely that the MM will be more flexible with skill variance and more matches closer to the 40/60 limit.
So if you feel match quality has decreased since earlier in OW’s life, you are likely correct. It doesn’t mean it’s not working though, just slower.
MMR/SR is also a somewhat relative value rather than an absolute value so someone in the same rank as before maybe more skilled than that rank a few years back as more players flood the ranks under them.
MMR and most elo derivatives start in the centre rather than the bottom as its much faster to get people to the right place that way. Less of the experienced players demolishing the learning players. (this does mean that it probably feels like you see more throwers if you are under 2500 SR, as new players need to likely get under your rank to be where they belong - should be roughly equal on both sides so it’s not a problem for the MM but can make the match suck for you when you randomly get them)
Honestly that’s why you really don’t want an MMR reset, the game would just be a mess for a week or two with constant one sided matches where one team literally has no chance (there was actually a bug that reset MMR one season - until restored from a backup, you can probably still look up vods. was pretty brutal and hilarious for the like half day it lasted)
3 Likes
I secretly loved that bug. And yes. after an MMR reset things would be a mess. I also forgot to say. LONGER Seasons.
I know people on this forum get mad when I compare to regular sports… but I do it anyway, You don’t start teams at .500 in any competition. I understand the concept, but then you get people who ride the wave. NO competitive thing does that. You start at the bottom and rise. Some will rise quicker than others. You put level 25’s (real, smurf, and actual bots) in to comp where the average is supposed to play. AS well as higher levels ( although they do this slightly better by putting theim in Plat) A progression of a lifetime of being a competitor in terms of population is a downward slope, not a bell curve. The bell curve is the talent stats.
Thanks for the discussion, I think I’m done for the day. when the message board warns me… I listen that I’m probably talking too much.
1 Like
But performance depends on the teammates you have, and the competition you are facing. If you play well and the MMR system places other skilled players on opposing teams to handicap you, then you don’t generate the same stats as you would under impartial matchmaking. PBSR doesn’t compensate you for handicapping. Ultimately, you can only lose SR when you lose a match. You can only gain SR by winning matches, and algorithmic handicapping affects skilled players’ chances to win.
MMR does not oscillate between hard-game/easy-game depending on whether you won or lost the previous match. It’s persistent and fluid; informed by your entire career, with a special focus on your most recent matches. MMR also takes your total experience into account. Someone who plays effortfully and skillfully all of the time is handicapped against their favor all of the time, especially if they are experienced.
2 Likes
It’s a vicious circle. Players have been complaining about the matchmaker since the beginning. Almost every Blizzard game has a very strange way of matching players with the same “skill”. Blizzard is even notorious for this and games like WoW PVP, HOTS and Co. partly failed because of it.
If players get the feeling of “unnatural” matches or series of unwinnable and uncontrollable matches, chances are they will look for a better source of entertainment.
We’re not talking about a hobby we do regularly in a club or with friends, in an environment we can adapt to.
Every single match in Overwatch (and other Blizzard games) is 99% a game of chance and you depend on the goodwill of your fellow players. If even one teammate has sinister intentions or simply dramatically underperforms, the particular match is usually over and you waste 15-20 minutes of your life.
Many “healthy” people understand (consciously or unconsciously) that it is counterproductive to stay in such an environment longer than necessary. That’s why most casual gamers, as well as serious competitive gamers, left relatively quickly when they realized that extremely much of the gameplay was out of their hands.
Fewer players lead to much larger skill differences, which in turn accelerate churn.
Why have people like me stayed so far? Because Overwatch is actually a very good and varied game in terms of mechanics and gameplay. Nevertheless, I also fall into the category of players who have perhaps held out longer than would have been good for their psyche.
Edit: 10-15% of the players are so good according to objective criteria that they can move in all areas of the ladder without any problems. For this minority it is easy to make smart speeches about talent and skill, especially when they themselves start teasing weaker players with smurf accounts out of boredom.
Perfectly fine, but then you also have to accept that the playerbase ist getting smaller and smaller and the match quality decreases due to SR gaps.
2 Likes