What I don't get about "forced losses"

Competitive Discussion
57

I think the real question is: Why would the matchmaker build such an obviously imbalanced team comp?
In my opinion its because we have two metrics (SR and MMR) and the MM uses both to create matches:

  1. All players need to be within the same SR range, as close as possible to 2000
  2. The average of all player’s MMR on each team needs to be as close to equal as possible

I, say, we only had MMR used to creating matches, the 2450 player would be teamed up with other 2450 players, against a team with 2450 players. There wouldnt be any “carry outlier” that needs to get hit with a bucket filled with cold water in order to get sober because he has been put on an important mission to save the world.

I know that what im describing (both SR and MMR being used at the same time) has been heavily disputed in the past. But, reading this thread now and the forum recently (ive taken 4 months break), it seems that the “MMR outlier” theory has become more popular and less disputed. Correct me if im wrong, but isnt this what Cuthberts famous/notorious “Handicapping” theory has been about all along?

What i can add to this discussion is my recent experience from just one week of play. When i came back, i placed 2200. I had been up to 2700 about 4 months ago, played in low Plat stably for about 150 games before i took a heavy streak of losses and went back to 2200, where i placed now.

When i decided to give OW another chance, i did so with a free and open mind. I told myself i had to work on improving my gamesense, especially tracking the important Ults on both teams to judge how tempo currently is in the game and what the chances are for the next team fight. In the first few days, this seemed to work pretty well, as i saw good improvements and my increased efforts to improve gamesense paying out.

But then i noticed that i wouldnt really move away from the 2100-2000 range as i would expect. So i looked into SR gains again and PBSR became a higher focus again. I experimented a little with what stats i should focus on (playing Moira), e.g. more healing/def assists vs. damage/elims and what would yield better SR gains.

I also started checking Overbuff/Oversumo/c0derwatch again to see how my personal performance would fare. All of these 3 source indicate that my personal performance was way overperforming those of my peers at around 2200. c0derwatch sees me at around 2500. Oversumo sais im overperforming 80% of my peers. And Overbuff had my recent Moira sessions display stats (except damage) in the 90ish percentile.

Yet i could not climb up to where these statistic-based tools see me. Also dispite my efforts to improve gamesense as well. The day i quit the game, after just one week of playing again, i climbed from high silver back to 2200, only to go on an (almost) straight loss streak of 10 losses back down to silver.

That was the point where i decided that all of my efforts on this game were wasted. Just as wasted as theyve always been before. And that giving the game a chance and trying to improve even more than ever before would not get me anywhere. Its also when i started to check this forum again and see how far the discussion about “forced losses” and hidden MMR would be.

Just as a side note: I would prefer to not equate the “forced losses” discussion with “conspiracy theories”. There might be people who actually believe that Blizzard is conspiring against them or a group they belong to, but i think that most people just believe that there is a flaw, an error, in the system. Or maybe if not an error per se, at least an implementation that leads to unproductive results.

1 Like
58

This is not true. I play in Gold. I’m constantly getting “platinum+silver” groups in my matches. If you will tell me about “average MMR”, I would like to hear it.

Regarding MMR difference. I keep track of team SR difference at the start of the match. This SR difference changes from negative to positive values in a cycle. The only randomness there - where is baseline for this change. So, in the end - matchmaker DOES give you higher, THEN lower MMR opponents (and teammates). In a cycle.

59

“Forced losses” is literally a conspiracy theory. If you believe it’s just an error or flaw, it’s not “forced”. The forcing is the line over which the rational critique turns into conspiracy.

I speak a lot of educating people on what is a reasonable expectation of what the MM and ranking system can do. I get that it’s frustrating, but if you’re only 300 SR below where an app thinks you should be, you’re well within the range where it’s nearly impossible to tell the difference. It’ll be a long slog up from there even assuming that’s where you actually belong This goes to my 'over precise" argument above. 10 losses sucks, but if you’re 500 SR from where you really should be those games should be pretty easy wins.

One week though…that’s not enough time. Even if you’re playing 10 hours a day it’s doubtful you’re playing well all 10 of those hours.

This is my Version HC so yes. The idea has several flaws though. First, as Oldphardt mentions below you it’s generally groups that are the issue, not the matchmaking. That’s not a distinction Version HC makes. Second, as I mentioned above if the teams are “even” then they’re even. The idea of someone having to carry is a mistake because everyone has to do their job well no matter their skill level. Third and more specific to Cuthbert in particular, he has a tendency to take any complaint, valid or not, and tell people that it’s due to MMR and that MMR should be eliminated, even when the actual problem is that MMR is being misused in some way, e.g. groups, smurfs, and simply not playing to your skill level.

He also is a conspiracy theorist by definition as he thinks MMR is used by Blizzard to get people addicted to the game to make more money.

Exactly… it would be super stupid to do that. They allow groups of widely varying SR and I think that’s a huge mistake for this very reason. Outside of grouping which it’s understandable why they do it, it’s just not done as a matter of policy. You’ll get games in unpopulated times and SR brackets that will have wider variation, but if you’re soloQ in gold at peak you won’t have a platinum on your team unless they’re grouped with a silver.

You used to get more complaints about this when someone’s season high was displayed on their icon, e.g. having a “Plat” in a silver game when in reality they were either boosted or made it up there very briefly and fell down to silver, but they changed how that works. Most of the imbalance complaints aren’t substantiated in any way and if they are, it’s groups of 800SR differential.

Edit: Someone else asked about this in Cuthberts thread. I posted a more detailed response here: Algorithmic Handicapping (MMR) is Wrong for Overwatch - #562 by OzoneOOO-1681

1 Like
60

I think you missed the “as possible” in what he said. Nothing you say in this post disputes the fact that “the matchmaker always tries to find people with as close MMR as possible”.

Even your “platinum+silver” groups, I don’t like them, but if you’re gold getting groups of 800SR differential that’s not a problem with the matchmaker, that’s a problem with Blizzards stupid grouping policy for comp. A policy that I don’t feel gets enough attention and even when brought up people throw a fit because they “want to play with their friends”.

What they need is a separate group and solo SR. They may even be able to work it into the dynamic queue, but I’m not sure.

61

I’m not sure what you are complaining about. Of course groups of platinum+silver players will have an “average MMR” in or near the gold range, and will usually be put into games around gold. What else would you expect or want?

I guarantee this is just a misunderstanding of probability and statistics. Let’s say I flip a coin repeatedly, and look at the count of heads or tails for the last 10 throws. You’ll see it cycle between having more heads, or more tails.

You have a binary outcome, whether your team or the enemy team is higher. And it’s random. So of course you’ll alternate (or cycle) between the two possible outcomes. I’m not trying to be condescending, but people mistakenly find patterns in completely random events all the time. It’s a well known phenomenon. Actually write down all the SRs for 20 games if you want, and with proper analysis you’ll immediately see there is no non-random “cycle” of harder/easier opponents.

You’ve discovered:

  1. You play better on your mains, and thus tend to win more.
  2. The basic existence of tilting and hot streaks that exist in every video game, and even real sports.

Do you know anything about sports? Do you know how players all refuse to speak to a pitcher who is currently throwing a no-hitter? Do you think that happens because they are given easier batting opponents? Of course not. Humans are complex psychological creatures. Success feeds our confidence and success further.

The same is true in reverse. Professional atheletes have bad games and slumps. In the video game world it’s usually called “tilting”. Nothing makes you tilt more than struggling and losing (like you are likely to do if you flex to heroes you are bad at). This is why EVERY game ever has had people go on huge 20+ game losing streaks, and why the common advice to “take a break” after a few loses is spread around.

A final, more humurous, point against what you are saying. If you read this forum with any regularity I’m sure you’ve seen Cuthbert’s “Handicapped MMR” threads. He, and the people who agree with him, claim the exact OPPOSITE of what you say. You say playing well means the game gives you easy matches to get you up where you should be. He says playing well makes the game “balance you out” by giving you harder opponents. Who is right? Neither. It just shows how neither of these conspiracy theories are formed from any actual testing, and y’all both experience confirmation bias on your own ideas while playing.

The vast majority of the time a stomped match has nothing to do with any match maker problems. Remember old KotH when it was best of 5? You know how many games I played where one side got STOMPED for two rounds, than reverse swept? Currently you’d just end the match as a loss as say “what an imbalanced match” but they weren’t imbalanced at all.

Small things, sometimes random, can add up to make big differences in outcomes during a game. People don’t realize how close most games are. One of my last games the other team finishes Hanamura with 4 minutes left, pretty good attack right? Except they barely got point A in overtime, after never touching it. If one fight went slightly worse for them, they end without even 1 tick. Instead, my team is tilted over our defense is crumbling, and people started fighting, leading towards terrible play and a loss. The game was easily within our ability to win, but small things caused this disparate outcome.

Phew I’ve rambled a lot. Point still remains. No such things as “forced losses”.

62

He actually did this. Used the moving average with graph smoothing and then called it “cycling”. If you removed the average and the smoothing it’s just a bunch of random jumps to and fro. How Competitive Skill Rating Works (Season 11) - #24 by Oldphardt-2154

No man…it’s the ALIENS!

I’m a bit surprised and saddened that I was able to describe all of those conspiracy theories as quickly as I did.

I should play more.

1 Like
63

You know, why I don’t believe in random nature of team SR difference change? Because there are no long streaks. There are no long streaks in my data, there are no streaks in Kaawumba’s data regarding 3000+ players. It’s obviously controlled in some manner. If it would be totally random, we would see long streaks in one direction. But there are none.
Streaks are present in SR movement (as it should be in case of random events).

If you, or FriendlyFire are proficient in statistics and math, maybe you could help with interpreting the results of spectral analysis I did on my season 11 team SR difference data. I would also appreciate Kaawumba’s opinion.
Here is the spectral analysis result:

https://imgur.com/a/SXq2kgt

64

Do you have more data than the 68 games? He gave his opinion, which was that there wasn’t enough good data to determine anything one way or the other.

I’m simply pointing out that the claim make is just, like, your opinion man. It may look like a cycle to you, but it doesn’t look like one to me, and there’s not enough data to say objectively one way or another.

65
  1. 25 games completed this season. I am 13 to 12. Of the 12 losses, 3 were due to Leavers and 2 due to throwers. So far no enemy Leavers and no enemy throwers I am aware of. So in theory I’ve been robbed of circa 125 SR already this season.

I’ve won games at 2500 before. I’ve regularly held 2350 for 4 seasons. But I crutches on my team’s dps being solid.

When I lose games fairly, is mainly due to being massively out dpsed. Enemy Pharah that no one can hit, etc. Team Sombra that doesn’t get kills and EMPS when we’re all dead at spawn, etc.

As a support I can’t output enough raw damage to carry poor dps. When I climbed from 1700 to 2350 with 65% win rate in season 4, was one tricking Torb.

If I want to dps my way up, need to learn a new carry hero.

  1. My natural Sr is presumably 1800 as that’s where I’m at. I could get a free alt account and likely place 500 SR higher no problem.
    My performance depends on what team mates I get. I can stay alive and support adequately vs. 2.8k to 3k enemies when in qp with my 3.9k dps friend. Conversely I imagine I can lose games at 1.5k Sr if my dps are rotten.
    I played 5 games of 2.4k Sr last season. Won 3 and drew 2.

What is my real Sr? Who knows?

66

It’s 184 games now in season 11

67

I took a quick look…there’s some pretty long streaks in there, red and green. 11 (or 9, you skipped one), 7, 8…seems normal for 189 instances but I’m just looking on my phone real quick.

Was I looking in the wrong place?

68

Look at teams sr difference graph

69

I guess I was looking right then.

11 is a pretty long streak for 189 games. Even a streak of 9 would only occur in 1 of every 6 simulations of 189 coin flips. Edit…I see now it’s 184…stupid phone.

What exactly are you expecting?

1 Like
70

Can you interpret the spectral analysis results?

71

No. I’ve used spectrum analyzers in chemistry and radio applications…I’m not familiar with any claims to analyze 184 bits of noise. From what I know it looks just like noise but zoomed in to max resolution. Normal spectral analysis has much, much more resolution and any non-noise signal shows a definite peak.

But I want to stress this application isn’t one I’m familiar with.

1 Like
72

Throwers and smurfs are two sides of the same coin. I think that’s the biggest problem with the matchmaker – it doesn’t know what to do with someone who dominates half the matches, and stinks the other half. It pees in the data pool to the point that we’re all swimming in green water.

3 Likes
73

You don’t know how a matchmaker could keep track of someone that has left a large number of games? Or notice when a player is playing well above the skill level at a certain rank? Or notice when a player has an unusually low win percentage? I mean…this isn’t hard to figure out.

This would be true if no one ever improved at the game…lol.

Of course, as people get better, they expect to climb ranks…(duh?)

This is actually what their system does.

They give each player a hidden MMR, which is based more on personal performance. SR gains/losses are adjusted according to where your MMR is. That is to say, if your MMR is higher than your SR, you will gain more for a win and lose less for a loss. (it’s also worth noting that your MMR only goes up on a win and down on a loss).

It’s actually a really well designed system. It encourages team play and winning games, while at the same time recognizing that if a player is performing extremely well, he should rank up.

This is another thing worth noting.

I see a lot of these posts about “10 game lose streaks” etc etc, and I usually respond with, “well, what were you doing during those lose streaks? Playing different characters? Playing characters unfit for certain maps? Could you maybe have switched off your one trick to a better fitting character for your composition? Etc. etc.”

Everyone is so quick to look to their teammates, and not to look at what THEY were doing during those streaks.

2 Likes
74

You mistook my point. I’ll try to clarify.

It COULD have been the case that they started everyone’s SR far below where they actually belonged. They IN FACT used to do this. I forget what season it was but your initial placements would start your SR below what was accurate and be treated as if they were decayed. They admitted this when they stopped doing it. What I’m saying is that it could be designed so that if a person didn’t improve, they would still go up in SR.

Not that it’s built in a way that assumes no one improves. That would be stupid. That’s why I put “failure” in quotation marks and specified that no one actually expects that, it’s just a design philosophy. The idea is that if someone doesn’t improve no movement should happen, not even after one game.

There are people, believe it or not, that feel that they should go up in SR without improvement and some that think it should be based on number of wins rather than being a ranking. The system could be designed that way, but SR would no longer be coupled with skill at all.

1 Like
75

It does all these things. It’s that the response isn’t to put them on your team to make you lose. The response is to move them to a more appropriate rank in the exact same manner someone who is doing it in a non-malicious way would. There’s no discernable difference between someone Alt+F4ing and losing their internet, no discernable difference between a overranked player and a thrower, and no discernable difference between an underranked player and a smurf.

The system treats them all equally because it can’t tell the difference. Even WE can’t really tell the difference. Especially with DCs. You’ve never played horrible in a game? You’ve never just gone nuts for a game?

The statement he’s making is that it can’t track a bad actor vs a normal player being bad. See Version 5 in post 4. If Blizzard isn’t forcing the loss then stop calling it a “forced loss”, that’s not what it means.

1 Like
76

I started a thread on matchmaker. Basic gist was that when I had a 3.9k smurf on my team (On 2.5k alt) then our games at 2.2k got very hard. Reasons include:

  1. Enemy had their own smurf
  2. Enemy team very well organised, never trickling and attacking in wave after wave of 6 .
  3. Terrible team mates that wouldn’t be out of place in bronze (eg, Zen not using his orbs).

System certainly recognises what players have a high MMR vs. their SR and as match maker balances based on MMR, then opposition ramps up, or team mates ramp down to make match as even as possible.