Competitive matchmaking's MMR system is VERY good

Competitive Discussion
63

This is 50% correct to what I’m saying.
I gained 800SR in a single night due to the anti-smurf system but you need to realize the matchmaker is not TRYING to get you out of your rank, it can only give you more SR for a stellar performance. When I go 25/0 30/2 game after game you would think the matchmaker would not want me pooping all over people - but instead it eventually starts giving me nothing but smurf players - which has the result of getting throwers, derankers, leavers, and opponents of a higher caliber. It’s quite obvious what is going on when the matchmaker has to place me against and with groups constantly just to balance me out. It’s weird being the only person in a game with stars and the only person who has played prior to season 8.

Interestingly it does seem to allow me to get wins if I take a significant break. Perhaps the matchmaker is “unsure” of my skill at that point and I’m actually given oppenents to beat.

It seems people who play often are significantly handicapped due to stats smoothing.

1 Like
64

I wrote a similar topic, based on personal experience:

(remove space after “https”)
https ://us.forums.blizzard.com/en/overwatch/t/competitive-am-i-not-supposed-to-win-when-i-am-good/63490

1 Like
65

This seems to be everyone’s experience.

I started writing every querk / reason / observation / win / loss / sr gain/loss over baseline, and what character I was playing and how many hours I had on that character and how many days break I took between games, and used statistical tools to suss out a few things about the matchmaker that appear to be universally true.

Since last november there was a subtle change to matchmaker that takes some measure of SR spread into account on your teammates, but I don’t think it takes MMR into account.

There is a “smoothing” that seems to occur after 10 days in a row of playing.

Logging out after a few wins appears to be detrimental to your matchmaking if you’re doing quite well and your K/D ratios are high on dps (or your heal to damage, etc). Being “On Fire” or the percentage of that meter is what the system is looking for - as the actions that lead to being on fire are probably logically the stats the matchmaker looks for to see “how good is this player compared to others at this MMR and map”.

Some heros, when you’re playing to the objective, simply won’t get on fire very often or they’ll lose it fairly quickly, even though without that hero you’ll get stomped. Maybe your dps are doing work and you don’t need to do much. There’s tons of variables as to why your stats might be low in a round. Maybe the game was over before the 10 minute mark - we still don’t have clear indication if quick games get factored in as hard - let’s say you had a 5 minute game where the attacking team rolled you twice and you couldn’t take the first point. They happen, and no matter how legendary you are if you can’t break the point (and live) without support your stats will look awful. What will that do to the matchmaker’s idea of your MMR? How largely will that statistical outlier affect your overall computed skill level? If those outlier effects are not downplayed this could be causing a significant detriment to a player’s computed “true skill.”

Interestingly it’s been confirmed in blueposts that the matchmaker is “unsure” of your skill after you take a break. I can accurately reproduce this effect by taking a 4 day break when I’m starting to get insanely difficult games even without ranking up - the quality of games changes quite rapidly once you start winning - my first game in placement absolute steamrolls with the exception of the first one, which was our team getting steamrolled in the other direction. Every game after was me taking gold damage, objective time, kills, objective kills and with no more than 4 deaths, everyone on my team and the opposing team was solo queue. After those 4 cakewalks the game immediately slammed me with smurf accounts on both sides and groups of players: it had cleared figured out my “forced smurf” status again and “caught up”. In fact, this happened IMMEDIATELY after what the server considers a “next day”: 12am GMT.

Clearly there is a delay in the calculation of your stats curve on a day by day basis or at least on some kind of server tick.

This can be gamed, btw. If you play either during low peak times you’ll get matches where the matchmaker has you severely imbalanced. If you are truly good and the matchmaker hasn’t figured you out, it’ll put you with players close in SR you can walk all over. I’ve been in silver and gotten the “very long time to find a fair match” message at 7am on a weekday - that’s pretty telling.

And, after 7pm EST, when the server resets and recalculates you, you should play in 3 stacks to throw off the matchmaker and improve the quality of your games after that time. Not only is it primetime, and the matchmaker will have less trouble finding people to place you against, but it’s just now recalculated your ability and MMR: blizzard have also admitted in blueposts that it’s hard for the matchmaker to figure out mmr in “large” groups - logically that means 4+. Anything less then 4 couldn’t logically be “large”. - This seems to be the most effective way to fall above the bell curve. My suspicion is since the matchmaker probably THINKS it can balance a group of three, except that now 1/2 of the team is enabling your best heroes, allowing you to pop off and look godlike compared to before.

The only way to fight against this unfair system is to game the matchmaker, honestly. You’ll find that many highly ranked players play very little compared to people like me, and if you watch them stream, you’ll see that they play during the time of day when the servers are quite empty.

66

This is exactly what has happened to me when I go on a big win streak. Suddenly I get noobs on my team and smurfs on the enemy where they are the best in the ELO to balance. Unfortunately, my team are getting destroyed by the smurf Widow, Hanzo, Tracer or McRee. They only DPS and they are either a smurf, cheating or being boosted.

Nobody is that good in Gold or Plat but I get matched agaist these type of players when I perform too well in too many matches. My wait times to que up in Solo Que also is a much longer wait as the matchmaker puts together an enemy team with the best players in the ELO. Since I main tanks and support you can image what happens to the DPS on my team. They get utterly destroyed having only 9 elims or even less than 5 elims at the end of the match. If we win, it was because the tanks and support did a lot of carrying but usually the smurfs or booster service accounts on the enemy team win.

67

I looked again today and I absolutely dominate you and most others pound for point at my sr. My soldier and mcree stats are absolutely ridiculous for my rank. The only metric you need really is eliminations per life and accuracy when it comes to dps.

Perhaps you were looking at the EU player that shares my name. I’m Lhun-1133 Give it a look when I’m done placements.

68

Dude. It’s the second day of the season. Those stats aren’t up yet and they’re not large enough to be accurate if they were. You haven’t even PLAYED McCree this season. What on earth do you even think you looked at? There’s nothing to compare at this particular point.

I posted the photos. https://imgur.com/a/1Jkkr It’s definitely you.

We suck. Both of us. The difference is that I’m not delusional about my abilities.

If you find data otherwise, post the photos. I dare you. Good luck finding comp data to compare to other people of our rank from last season. Seriously, I’d love to have this resource.

69

It depends if the system spotted you out and thinks you deserve fair matches. Matches will take a bit longer to find.

Hope I don’t eat my words later after more comp this season.

70

masteroverwatch dot com has all seasonal data from last season. Look at sombra, widow, soldier and mcree - all played over an hour last season so there’s major stats smoothing. I would like to see how you would explain those numbers for a silver player?

Widowmaker is objectively the purist representation of mechanical skill there is in the game. Across 7 games, I maintain 53% scoped accuracy, maintain 2.5+ kd, and some games exceed 4 K/D ratio - in competitive. This is between top 13% and 2% percentile of all players who use the site.

Would you say that someone who is objectively silver should be able to maintain those stats? There is no argument that that level of mechanical skill is below average, at all. If you’re looking at player numbers that puts me in at least the top 100k players of the game.

Let’s look at a hero with less stats smoothing from 9 previous seasons since I just picked her up, sombra.

Stats over 8 games:

Elminations: 91st percentile
Damage/min: 91st percentile
EMP: 81st percentile (emp speed is a good measure of effectiveness)
Final blows: 74th percentile

Cough.

How about mcree?
Accuracy: 88th percentile
Elims/min: 78th percentile

And that’s even with a few garbage 1 life rounds mucking up the stats after switching off a counter pick - I get 2.0-3.31/min and 1000-1778/ dps a minute on most days. I track this stuff obsessively, because I’ve noticed a problem with the system after dropping from platinum to <500 between season 1 and 3. There IS a bug or some major oversight on old accounts with tons of playtime but it affects such a small group of players that it can’t possibly be blizzard’s focus. We’re vocal but we’re a massive minority.

Funny enough some of the same problems people like me face happen in the upper tiers. When the system as no idea what to do with someone who’s skill seems to massively mismatch their SR the quality of games you get are downright bizarre.

There was a time where almost every game I got was filled with small groups of plat+ 3 stacks with a bronze in it, which made absolutely no sense to me. I found games EASIER on diamond then I do during primetime on silver for various reasons.

Mokugeki#1389 is my alt. Unfortunately last season we used the spare account to let friends play in the house, but you’ll have to look at season 6 and 7 in game when I was tryharding on it.

The ultimate fix is pretty simple. Create matches based on SR alone, and use the quite good system MMR has for figuring out similarly skilled players to match teammates. You teammates should be close in MMR, your opponents should be close in collective SR, and that’s it. Then the cream rises to the top.

71

Easy. This is all against silver players. You wouldn’t have those stats against gold players (or above).

Also, you don’t seem to play these heros that you’re good at in comp very much, so I don’t know what you’re complaining about.

If you really believe that your Widow is better than Silver, then play Widow to get out of Silver. Not Zen where you firmly belong in Silver.

You’re looking at 8 games against Silver players and thinking your top 22% of ALL players? What’s the new phrase? 4head?

Plus, stat padding doesn’t win games. If you want to play McCree, great. Make a VOD and have a higher level player rip you a new one, then take that knowledge to go out with your (seemingly) decent skills for a Silver and get yourself out of Silver.

Or, you know, you could just sit here and blame your teammates and Blizzard for not granting you the score you feel you deserve without having to work for it.

72

I hear this argument every single time, but I already gave you an example that proves that’s not the case, on an alt account.

I don’t blame teammates. I blame the matchmaker’s detection. In a group, we would dominate if the matchmaker didn’t skew our opponents heavily.

Make a VOD and have a higher level player rip you a new one, then take that knowledge to go out with your (seemingly) decent skills for a Silver and get yourself out of Silver.

Done that tons of times, they always remark and say “what the heck is going on, why are you silver?” And it devolves into, “looks like you need to buy another account”. So I eventually did. We play in local annual tournaments. Last year we took 2nd place with a combined team average SR of ~1600 against team averages of 3000+. Diamond, Masters and GrandMasters players get quite confused.

Statistically speaking filling in for a critical role such as tank or healer will give a disjointed team a greater likelihood of winning than 5dps 1 tank. It’s why I picked up Zen again. All I do is fill fill fill, unfortunately. I don’t have to fill in my static, but then we get HEROICALLY hard matches against people’s low ranked alt accounts.

I’ve climbed a combined total of at least 1675SR since season 3, with massive 800SR falls on weekends and prime time that we had to climb back out of.

I’ve got more then enough evidence to support this and dozens of people have witnessed it themselves.

But I have a day job, performing statistical quality analysis of experimental software and hardware, and I’m busy. It’s part of the reason this bothers me so much, as I can see the flaws in the way it’s detecting players, from a pseudocode level almost at this point

When it really comes down to it I already proved my point by reaching diamond on my alt. I would like my “main” account with all my loot and history to reflect my actual skill, but in the grand scheme of things I’m prone to dismiss the whole endeavor of getting blizzard to fix the algorithm because it’s not worth the frustration.

I play this game to be entertained and for the joy of semi-professional competition. Would I like my 17 years of FPS experience to be reflected in a shiny number next to my character and get a better quality experience in the games I play, and to not be judged as someone suffering from dunning kruger? Sure, but it doesn’t look like that’s going to happen anytime soon.

Folks like you come out of the woodwork all the time. I just laugh. Custom game deathmatch separates true skill quite nicely, and I’ve never had an issue making top 10 in ana paintball or widow hs only. shrug.

73

I mean, does it though? It seems either a) you’re full of it or b) this strategy isn’t working well for you at all.

I had a VOD done once as Zen where my team actually was bad and I asked what a GM would have done to win. Response: Pick DPS and carry.

As to the rest of your post I’ll have to wait till I get home. Nice job allowing your friends to ruin games on the drop from Diamond to Silver on your alt, assuming you’re telling the truth.

74

You’re right, it isn’t, but I can’t bring myself to make games poop for other people. When I instalocked dps I managed to climb 300+SR in one day (a holiday in my country, and not in america) and I was lucky enough to be playing during a time of day I could truly carry. I’m quite good at a variety of heros, and I care about the feelings of the people I’m playing with. During primetime, getting others to comp properly is hard. Unfortunately some characters require more synergy than people can muster, Zen for example cannot be left uncontested or unprotected.

Response: Pick DPS and carry.

Yep. This is unfortunately VERY true. However, I don’t like getting yelled at, and I don’t like forcing other people into roles they are bad at, IF they switch, which may have a further detrimental effect even if I’m getting a 5.0K/D the entire time I switch to genji or soldier or whatever though. It doesn’t matter how good you are, put 1 single top 500 dps player vs an entire balanced team of silvers and the silvers will win every single time. Hell, a balanced group of 3 silvers usually does vs 1 top 500.

When I have to switch to widow and carry, to deal with a pharah for example whom our soldier cannot shoot: I’m forced to mute all chat, including voice.

And while it might remove one threat from decimating my team, it doesn’t win the objective. You can only guarantee a win as a team. No matter how godlike your tracer jive and weave is, there are CC heroes that can get a lucky shutdown or ult on you ONCE (junk tire, whatever) and your entire team falls apart. I’ve watched it happen way too many times. Carrying is possible, but incredibly stressful. You have to play at a level FAR ABOVE even your normal skill, because now you’re playing without the support of a competent team.

And as for the deranking thing ruining games? It actually didn’t do so bad. Because the account is played so little, both times it played it was in placements. (and it was always grouped with my main and friends in near full 6 stacks anyway)

So, it dropped to around high silver mid gold without touching more then a couple full competitive matches, since placements can swing around 500SR at the extremes. No teammates were harmed since the account was played in full stacks with friends only, if anything our opponents got a few mulligans. :wink:

The only reason I was convinced to buy a new account was because it had a true purpose in my household beyond stroking my ego, otherwise I would not have done it or bothered despite repeated pleading from people who knew me to do so to “fix my rank”. We have a whole separate computer here for friends and my young daughter to play on, and that’s one of the game options now that they can play with us.

75

Hi it’s lhun. Posting here so that others can easily find my alt.

76

I just checked again, looks like it was season 5 on moku. I can’t believe it’s been that long, jeez.

77

Sure man. Yep. And my Silver level Widow stats are better than yours in all respects…except accuracy where it’s the exact same. My crit accuracy is actually better, which is a key stat.

Seriously dude. Start posting pics or something. Your delusions are too easy counter. Honestly for your sake I hope I’m looking at the wrong thing, but here’s what I found:

https://imgur.com/a/1Jkkr

You know. You almost had me. What you write SEEMS so reasonable. I’m glad I waited to post anything until I came home and checked the data myself rather than taking your word for it. You’ve proven to be an unreliable narrator.

AT BEST…if your skills have improved recently then stop tying one arm behind your back and play the hero you’re best at. I’d rather win a game with 5 dps than lose with 2-2-2. You’re not doing anyone any favors by “playing nice”.

But you’re no longer convincing that your stats show that you should be better when your stats are as good as mine.

80

I’m not sure if you’re just blind or dishonest. I hope it’s just blindness, or an inability to read or interpret stats correctly.

The thing you posted simply shows greater gametime and overall compiled totals, and shows absolutely nothing of game to game averages. You should learn how to use the tools properly so you can evaluate actual performance over a number of games.

i .imgur.com/RUTffnX.png <- this is my last 30 or so games.

Even with a massively negative win rate on zen my stats are in the top 10-15% of all players across all ranks that use this site. And, to top it off the site is biased towards gold/plat players, btw - since the people who care are generally higher ranked.

My kill death ratios are currently obscene on dps, on the heroes I’ve played more than one game/part of a game on - in the upper 16% even after 35 minutes of play. That’s the average that matters as it’s a more representative metric of overall play.

It’s a little tough to get a good idea at the beginning of a season until I’ve played every hero I play (practically all of them) but you should get the idea.

You are clicking the “competitive” tab, right?

overbuff .com/players/pc/Lhun-1133/heroes/zenyatta?mode=competitive

I even have a lower winrate then you and a higher overall hero score. o.o You’re honestly underranked too in many ways and you would benefit from the flaws in the system I’m trying to expose. You belong at least in high gold to platinum on some heroes.

Here, this is a lot easier. Switch heroes around, play with the comparisons.

overbuff. com/compare/pc-OzoneOOO-1681,pc-Lhun-1133?mode=competitive&hero=zenyatta

You die almost twice as much as me no matter what hero it is and we have the exact same number of Games. You’ve actually managed to win 1 more than I have - Again lending credence to the issue being supportive teammates or not. Last season I was on a ridiculous winrate with Zen, amazing how brigetta changes things up.

Anyway, this shows 1 of two things. Either your positioning is way off compared to mine or you’ve had less supportive teammates (or you couldn’t force them to be supportive).

We’ve actually both played a few games as solider. Play two more and compare our stats again.

Even though you happened to have more kills than me in the one game you played as S76, look at your deaths.

Your issue is NOT your mechanical aim, it’s your mapsense. You’re either rushing in and getting blendered in the meat grinder, or you’re not staying near the correct supports to be effective.

Sure you get a bunch of kills but you’re dying and out of the fight way too much. Nothing about my stats suggests I’m doing much wrong at all. Yours have a glaringly obvious pain point you need to work on, and that’s your death ratio.

You’re simply dying to much. Fix that and you won’t have a reason to be stuck anymore. But, you still might end up being stuck and then start to agree with me - as per the point of this thread.

The matchmaker is not giving you good teammates that can enable your admittedly fairly good mechanical skill. Eliminations Ratio is the most important stat on dps by far - especially highly mobile ones with a winbutton and self-heal like soldier. At the lower ranks soldier is VERY OP, so is Roadhog, if your opponents have AN APPROPRIATE mechanical skill for their SR ranking, which, in the current system, many do not.

Try playing around 3pm eastern on a weekday and your stats will improve as the people in silver probably belong there. I’m physically unable to do so due to work, or I would do the same- so instead I get saddled against players with the same inflated mmr and deflated SR as me.

81

your still the biggest troll on the internet

op acc shared with a gm to get it to diamond
then took it over and threw it into silver

82

I honestly don’t doubt this one bit.

I didn’t notice until after I posted my last post that he used his alt to bump his thread and say he agreed with himself in the process.

I’d still like to know the details of why you think he did the boost/throw thing, but it doesn’t matter.

The guy clearly lacks integrity as evidenced by the alt account bump so it’s not worth taking anything else seriously.

I like how he thinks I’m under-ranked though. I climbed from low bronze by learning and correcting my mistakes with coaching. I’m having trouble at low gold, but it’s my problem, not Blizzard’s.

83

One player (you) does not AT ALL determine the win rate of the match. Your win-rate can fluctuate WILDLY during a match due to disconnects, players tilting, discouragement from an easy or particularly good ult, or someone switching to counter the “carry”. It’s absolutely useless and rediculous to determine a single player’s skill based on the results of a match where 11 OTHER players contributed to the win or loss. The only thing that should adjust mmr is mechanical accuracy and KD ratio, damage over a match, and fire rating, meaning you’re using your ult and abilities far more effectively than other people at the same SR (not mmr!!) …meaning it could actually GO UP in a loss, and SHOULD go up in a loss, if you heroically did your best and outclassed everyone on your team and the opponents team. Games with leavers should NOT adjust mmr and have the minimum SR adjustment possible at best.

You cannot solo carry in overwatch. You can contribute to a win, but even if you’re jjonak or iddqd you WILL lose with a team of true silvers against a team of true golds or high silvers, no matter what role you’re playing. You could be the most godlike widow in the entire NA server and all it takes is one player to go d.va and harrass you, removing your damage from the equation and allowing your opponents to win.

This is the entire problem with personal mmr ratings - the game does not depend on you. It’s more likely that a true GM placed with equally matched teammates and opponents might tilt a game in the other team’s favor, but if that ONE PLAYER is countered by even the most basic of play and their impact is less, if the rest of your team isn’t as good, you WILL lose, I promise you. If you have a leaver and one person counters you, you WILL lose, no matter how good you are, truly.

I know plenty of people who are global elite in CS, or absolutely top tier in Quake, with 60% or better accuracy ratings who often don’t break platinum in overwatch because they screwed around during their first few months and played with friends, now they’re hard stuck. Anyone who’s done a bronze to plat climb has done so on an account with old or no play history. Try doing that on an account that has played a few hours a day, every day (500+xp or something) and maintained 50/50 ish winrates. It’ll take much, much longer unless you group up with people objectively higher rank. Or, try doing it while grouped with people actually AT the rank the current hero is and try to climb. You might manage but it’ll take 60x as long. Also make sure you’re ONLY PLAYING during NA EST prime time or wherever you live. If you do this while playing during OFF HOURS, when the player base is low, it’s not accurate. You must play during the summer and on weekends after 6pm EST. Any other playtime will screw with the matchmaker and give you an easier time.

It’s been demonstrated tons of times that people who are truly diamond and above make it there no problem on a new account.

Now, all of this might no longer be true, but it definitely was before. I haven’t played much at all lately since placement since there are games with better ladder systems now that use the same skills, but the lfg system in overwatch seems to have fixed a lot of this. Once the meta stabilizes I’ll give it a go, but I did nothing but climb last time when I wasn’t getting disconnected.

84

this is gay, how much they pay you? mmr system is just an stupidity based on the avarage of several variables, it lacks of every bit of a complex system