Streaks in Overwatch, Simulation and Analysis and Data!

Competitive Discussion
1 2 3 7 Next
Too long, didn't read

If your SR over time looks like "Skill Rating vs Game, CDF Statistics" at https://docs.google.com/spreadsheets/d/1AnHSkLQz098SI1Z3VDhtdWBmBgQ1ehoG2Z4uI3h0UQg/edit?usp=sharing, that does not mean that Blizzard is using broken matchmaking. Random variables, on the other hand, are not your friend.

Introduction

There has been a great deal of posts lately implying that because there are win/loss streaks, and win/loss streak reversals, there must be some sort of illicit Blizzard meddling going on. In general, these posts show a lack of understanding of how random numbers work, which links such as https://wizardofodds.com/image/ask-the-wizard/streaks.pdf show. However, that link and others are not explicitly focused on what goes on in Overwatch, so I did a more relevant simulation. The graphs are at https://docs.google.com/spreadsheets/d/1AnHSkLQz098SI1Z3VDhtdWBmBgQ1ehoG2Z4uI3h0UQg/edit?usp=sharing and the text description is below.

Coin-flip statistics

First, consider a case where a player has a 50% chance of winning, and a 50% change of losing each game, and has played 1000 games (starting at 2500, +25 for a win, -25 for a loss). This is coin-flip statistics. The player’s SR over time will look like the plot “Skill Rating vs Game, Coin-flip Statistics”. Note that even though the player’s starting rating is 2500, it dives all the way down to 1200 before working its way up. If we attempt to ascribe narrative to this, we would say that there was a huge weight on the player for ~500 games, which was then somewhat lifted. However, since we can look at the code, we can verify that there was no such effect. It was only random coin flip chance. We also know that there is really no such thing as “true rating” because each game is independent from the past. Running the simulation many times leads to random final SRs.

Not only does the overall trend line go through large swings, but there are also many streaks, as “Streak Frequency, Coin-Flip Statistics” shows. Streaks of up to 6 are common. Streaks of up to 15 occurred in this simulation.

The final plot shows the autocorrelation function, which measures "Assuming a player won game 0, what is the probability that a player won games in the past and the future?" As determined by the model, we get the expected result: There is no influence on future or past games based on the current game. (If a win at game 0 guaranteed that you lost game 1, then there would be a spike to -1 at x=1.)

Cumulative normal distribution function statistics

Of course, it isn’t really coin flip statistics. As a player goes up in rating, wins become harder. As he goes down in rating, wins become easier. I did a second simulation, in which win probability is determined by a cumulative normal distribution (mu = 2500, sigma = 500), as shown in “Win Probability vs Skill Rating, CDF Statistics”. Put simply, if the player’s rating is 2500, his win rate is 50/50. If the player’s rating is 5000, his win rate is zero, and if the player’s rating is 0, his win rate is 100%. There is a smooth s-curve between 0% and 100%. This modification fixes the problem of random SR drift, and the trend over time averages at 2500. However, there still are large trends down and up. From game 425 to 575, the player gains more than 400 SR. He then falls all the way back down in about 125 games. Throughout, there are many large and small swings. However, the underlying math has never changed, and his “true rating” remains 2500. Streaks without interruptions are slightly shorter, but still frequent and long, as the plot “Streak Frequency, CDF statistics” shows.

Here the autocorrelation looks basically the same, indicating no correlation between games. In fact, though, if I run a simulation with 100000 games instead of 1000, there is a very slight negative correlation. This corresponds to the increased difficulty in winning as rating goes up (and decreased difficulty in winning as rating goes down).

Real Game Data

Thanks to Porkypine, I was able to analyze the game data at https://docs.google.com/spreadsheets/d/1O9mprevYPN11aBP9eLP7DOty_Bs9kmKmz4Wmv0hNZRs/edit?usp=sharing and compare it to my model. This is shown in the third row of charts. All of the charts look the same as the CDF plots, within the limits given by the error bars. Win Probability vs Skill Rating is consistent with a cumulative distribution (or a straight line, for that matter). Skill rating vs game shows the same sort of motions. Streak frequency has the same fall off. The win/loss autocorrelation function is one for the same game, and zero (with the errors of the measurement) for other games, which means that win probability is not based on past games, and blizzard is not taking into account past games (including streaks) when matchmaking in any measurable way (outside of wins leading to slight difficulty increase and losses leading to a slight difficulty decrease because of rating change).

If anyone has more data that they would like me to look at for weirdness or otherwise, I would be happy to do so.

If anyone would like to play with the code, it is at https://pastebin.com/7NJLUbK2 and https://pastebin.com/YFUsUshd
You will need Matlab with the statistics toolbox and curve fitting toolbox to run.

Addendum: Application to reroll experiments

You may remember me from (one of the few) systematic reroll experiments. https://us.battle.net/forums/en/overwatch/topic/20758807010 With this new data, we can now definitively say what is going on. With the model above, it takes roughly 500 games for a middle ranked player to fully explore their possible range of skill ratings. Therefore, if you reroll and your new account is placed higher, it is likely that it will take hundreds of games (on both accounts) for your two accounts to converge. If your true skill is changing (because you are getting better) it may take even longer.

We can also say that it takes roughly 500 games to measure your true skill (because that is how many games it takes to fit your win percentage curve with low error). By that point, your skill has hopefully increased, so measuring true skill accurately (within 100 SR) is difficult to impossible.
Blizz should hire this man for their MM system... At least the simulation agrees what I guessed before:

1) SR is not truly skill rating but somehow Lucky Rating.
2) No conspiracy of forced win/lose game intended towards players but from the bad MM system design ideas.
3) long lose/win streak is abnormally high with the algorithm, it hurts players game experiences.

conspiracy chain: Blizz designed MM system like this and doesn't want to reset SR for each season. -> Blizz wants you know little bit of the MM system. -> Blizz wants you finally give up and buy a new copy of game to hard reset your position. -> Blizz can say to others that they had 35 million copies sold out...
10/19/2017 10:26 AMPosted by xsun
Blizz should hire this man for their MM system... At least the simulation agrees what I guessed before:

1) SR is not truly skill rating but somehow Lucky Rating.
2) No conspiracy of forced win/lose game intended towards players but from the bad MM system design ideas.
3) long lose/win streak is abnormally high with the algorithm, it hurts players game experiences.

conspiracy chain: Blizz designed MM system like this and doesn't want to reset SR for each season. -> Blizz wants you know little bit of the MM system. -> Blizz wants you finally give up and buy a new copy of game to hard reset your position. -> Blizz can say to others that they had 35 million copies sold out...


Wut?

This data is simply showing that streaks are a normal phenomena and the SR range that someone can expect is distributed fairly evenly around their "true" rank which in this case was 2500. In a truly 50% "perfect" world you still have streaks that raise and drop SR. Just because its 50% doesn't mean that its always evenly distributed, looking at small windows of time will show fluctuations, but over the long term you will see even distribution. Now take into account that no MM system can ever be perfect as there are tons of factors involved that are outside of raw statistics and this will fluctuate a bit more.

I have always thought 500 SR tier size was a purposeful number and its interesting to see the simulation at +-25 basically showing a 500 range (+250/-250) around the "true" skill number.

People like to complain about bad teams, bad MM, big loss streaks when in reality most of this just means the MM system is probably working pretty much correctly in the sense that it is giving games with fairly even teams where both sides have 50% chances to win.
Except you left out a few things.. Variables.

Cheaters, throwers, and boosters. For starters.
(*Edit* p.s. I have a small theory that a lot of throwers are boosters moving back down to boost someone else.... anyway, moving on)

Then there is the learning curve. How much someone learns from the game how fast. MATH can't calculate skill like you are trying.

The biggest problem with the "skill curve" is is based on the Bell Curve. Which is something a few competitive games have started to move away from. Cause it's a terrible way to calculate skill which ends up in people feeling like they are being screwed, cause the bell curve forces everyone to be in the middle. And through my 20 years of competitive gaming it actually works more like a pyramid. Which also handles variables better. In a perfect world a bell curve would maybe work. But it doesn't. This is why I don't play CS:GO anymore.

Here are some references

https://casualaggro.wordpress.com/2013/11/08/the-noob-elitist-bellcurve/

https://blog.skipfault.com/2017/08/02/my-pubg-stats-and-the-top-monopoly-of-the-dedicated/

The graph on the second is the natural skill ratings people fall into when you start everyone at zero. Because casuals on the left make up the vast majority of those players making way for dedicated players to move right on the scale. He mentions in the article that he thinks that if everyone played equal amounts of time that it would be a bell curve.. and holy HE'S RIGHT. But not everyone plays equal time now do they? Also, they don't play with the same vigour.

And I found this while continuing a thought. And this guy has got the nail on the head. https://www.reddit.com/r/Competitiveoverwatch/comments/515dk5/the_bell_curve_distribution_makes_competitive/

The pubg scalene scale is natural to skill levels as it starts at 0 and graduates people through the scale, and LoL and SC2 use this and people complain a LOT less about comp games etc.

Bell Curve just throws everyone into a well.. Bell curve and says.. Have fun with our computer trying to figure out how to put 80% of you in the middle evenly! Peace!

Forum Mod Edit: This post has been edited by a moderator due to language. https://us.battle.net/forums/en/code-of-conduct/
10/19/2017 11:02 AMPosted by mrjman
This data is simply showing that streaks are a normal phenomena and the SR range that someone can expect is distributed fairly evenly around their "true" rank which in this case was 2500. In a truly 50% "perfect" world you still have streaks that raise and drop SR. Just because its 50% doesn't mean that its always evenly distributed, looking at small windows of time will show fluctuations, but over the long term you will see even distribution. Now take into account that no MM system can ever be perfect as there are tons of factors involved that are outside of raw statistics and this will fluctuate a bit more.

I have always thought 500 SR tier size was a purposeful number and its interesting to see the simulation at +-25 basically showing a 500 range (+250/-250) around the "true" skill number.

People like to complain about bad teams, bad MM, big loss streaks when in reality most of this just means the MM system is probably working pretty much correctly in the sense that it is giving games with fairly even teams where both sides have 50% chances to win.

This is the correct interpretation.

Specifically, the second instance has the most interesting plots.

The range that your system has players vary is about 500 SR, I always thought the the current system varied from 200-300 SR.

I mean this helps back the concept that match making attempts to place you in 50/50 matches and when you under perform you fall. If the system didn't account for skill at all, you fluctuate up to 1400 SR on pure 50/50 whereas assuming you don't improve at all and play at the same skill level you can fluctuate up to 500 SR.

The most interesting thing would be now to run this test multiple times with different seeds and see the results. (Seed as the random number generator that exists.)
The pubg scalene scale is natural to skill levels as it starts at 0 and graduates people through the scale, and LoL and SC2 use this and people complain a LOT less about comp games etc.

Forum Mod Edit: This post has been edited by a moderator due to language. https://us.battle.net/forums/en/code-of-conduct/

Except SC2 uses a bell curve too...

Source: http://wiki.teamliquid.net/starcraft2/Battle.net_Leagues
10/19/2017 10:26 AMPosted by xsun
2) No conspiracy of forced win/lose game intended towards players but from the bad MM system design ideas.


I think the conspiracy is that the system is working.

While I wouldn't put it past any game company to manipulate their players, I think a simpler explanation is that the system is poorly thought out, designed, and implemented.

All of the stats sites have huge amounts of trouble accurately gauging wins and losses if you flex heroes at any point it seems. I always wondered if that was unique to the way they pull the stats, or a problem in blizzards system.
Forgive me, I read on one of those sites that said that it was.. I knew LoL was a while ago, and people have mentioned that there is way less complaining and I just checked the forums on LoL and I can't find anyone complaining about bad comps. And in the SC2 I found already a couple. Not like Overwatch, but Overwatch has MANY more variables. Making a bell curve less effective. So, yeah makes sense blizzard believes in the math and philosophy of the 'bell curve'. It's a shame.
I enjoyed the read and the detailed analysis, but there's honestly just too many variables to come to any conclusion. Your calculations are all based on averages and 'ideal circumstances'

I think if they were fudging the system, you wouldn't have toxic players at Masters or above, and you wouldn't have people dwindling in low bronze. They would want the whole community to be happy and enjoying the game, more positive word of mouth=more sales

At any rate, it is their game. You agree to any terms of service before you start playing, and that includes modifying the terms with or without your consent. Everytime they buff/nerf a hero or patch something is proof in that. So could they fudge the system if they wanted to? You bet they could. Do they? I don't think so, it would be too big of a hit financially if they were doing any serious meddling. But they can do anything they want with their intellectual property, as long as they inform you what your license buys you.

The chips are in their hands. The house always wins.
You know I have over 100 games in my Data Repo, but my data is "too hard to read" for you :P

https://cloud.sky-net.me/index.php/s/oSkABXRtcNm4OMr

Started at: 1164 (Game 13)
Ended at: 1275 (Game 119)
Peak at: 1319 (Game 92)
I love when people only reply to the first post. We covered... everything you said except you agree to it by playing the game. Which we already knew. I press agree even if I don't agree cause I want to play the game. And as paying customers where the company claims they want us to be happy. We aren't I see hundreds of posts a day about the comp system. While the people who are in diamond + who don't have to deal with it as much say it's fine and 'getgud'.

We know the system works. That's the problem, and we don't agree with it anymore. When people realize it's broke as it is.. it will change hopefully.

Forum Mod Edit: This post has been edited by a moderator due to language. https://us.battle.net/forums/en/code-of-conduct/
10/19/2017 11:32 AMPosted by Shockey
I love when people only reply to the first post. We covered... everything you said except you agree to it by playing the game. Which we already knew. I press agree even if I don't agree cause I want to play the game. And as paying customers where the company claims they want us to be happy. We aren't I see hundreds of posts a day about the comp system. While the people who are in diamond + who don't have to deal with it as much say it's fine and 'getgud'.

We know the system works. That's the problem, and we don't agree with it anymore. When people realize it's broke as it is.. it will change hopefully.

Forum Mod Edit: This post has been edited by a moderator due to language. https://us.battle.net/forums/en/code-of-conduct/


I only saw the first post as I wasn't constantly refreshing the page as I typed
[quote="207576214032"]I have always thought 500 SR tier size was a purposeful number and its interesting to see the simulation at +-25 basically showing a 500 range (+250/-250) around the "true" skill number.


The range is a tunable parameter of the model. I directly comes from sigma (above, set to 500). Don't take it too seriously. I just chose a reasonable value based on typical variation. For higher skill players (plat+), SR varies less and a lower sigma is appropriate.
10/19/2017 11:04 AMPosted by Shockey
Except you left out a few things.. Variables.

Cheaters, throwers, and boosters. For starters.
(*Edit* p.s. I have a small theory that a lot of throwers are boosters moving back down to boost someone else.... anyway, moving on)


A significant point of the model is to model things like throwers and boosters that cause random factors outside of your control. If there were no random effects like this, sigma would be very small and people's SR would change by very little once they were at their proper rank.

Then there is the learning curve. How much someone learns from the game how fast. MATH can't calculate skill like you are trying.

The biggest problem with the "skill curve" is is based on the Bell Curve. Which is something a few competitive games have started to move away from. Cause it's a terrible way to calculate skill which ends up in people feeling like they are being screwed, cause the bell curve forces everyone to be in the middle. And through my 20 years of competitive gaming it actually works more like a pyramid. Which also handles variables better. In a perfect world a bell curve would maybe work. But it doesn't. This is why I don't play CS:GO anymore.

Here are some references

https://casualaggro.wordpress.com/2013/11/08/the-noob-elitist-bellcurve/

https://blog.skipfault.com/2017/08/02/my-pubg-stats-and-the-top-monopoly-of-the-dedicated/

The graph on the second is the natural skill ratings people fall into when you start everyone at zero. Because casuals on the left make up the vast majority of those players making way for dedicated players to move right on the scale. He mentions in the article that he thinks that if everyone played equal amounts of time that it would be a bell curve.. and holy HE'S RIGHT. But not everyone plays equal time now do they? Also, they don't play with the same vigour.

And I found this while continuing a thought. And this guy has got the nail on the head. https://www.reddit.com/r/Competitiveoverwatch/comments/515dk5/the_bell_curve_distribution_makes_competitive/

The pubg scalene scale is natural to skill levels as it starts at 0 and graduates people through the scale, and LoL and SC2 use this and people complain a LOT less about comp games etc.

Bell Curve just throws everyone into a well.. Bell curve and says.. Have fun with our computer trying to figure out how to put 80% of you in the middle evenly! Peace!

Forum Mod Edit: This post has been edited by a moderator due to language. https://us.battle.net/forums/en/code-of-conduct/


You are very confused about bell curves. Blizzard does not force people to distribute themselves in a bell curve. People do that naturally. As people win and lose and take points from each other, that is how SR ends up being distributed. This happens in many natural processes, and is a consequence of the Central Limit Theorem (https://en.wikipedia.org/wiki/Central_limit_theorem). That is, people are not forced into the mushy middle with most of humanity. People naturally belong in the mushy middle with most of humanity. If some other distribution was more correct, then that would just naturally appear.
10/19/2017 11:11 AMPosted by Cheese
The range that your system has players vary is about 500 SR, I always thought the the current system varied from 200-300 SR.


Based on reports I've seen, the variation tightens up as you go up the scale. Basically, less random stuff happens at the higher ranks. Also, as I mention above, the amount of variation is a tunable parameter of the model, so don't take it too seriously.

10/19/2017 11:11 AMPosted by Cheese
I mean this helps back the concept that match making attempts to place you in 50/50 matches and when you under perform you fall. If the system didn't account for skill at all, you fluctuate up to 1400 SR on pure 50/50 whereas assuming you don't improve at all and play at the same skill level you can fluctuate up to 500 SR.

The most interesting thing would be now to run this test multiple times with different seeds and see the results. (Seed as the random number generator that exists.)


It behaves about how you'd expect. The CDF plots are virtually identical. The coin flip SR over time shows a lot of variation in final SR, but otherwise looks the same. I can post the matlab code tonight if you'd like to play with it. If you don't have a license, you can get a free 30 day trial pretty easily. You'll need the statistics toolbox.
10/19/2017 12:00 PMPosted by Kaawumba
10/19/2017 11:04 AMPosted by Shockey
Except you left out a few things.. Variables.

Cheaters, throwers, and boosters. For starters.
(*Edit* p.s. I have a small theory that a lot of throwers are boosters moving back down to boost someone else.... anyway, moving on)


A significant point of the model is to model things like throwers and boosters that cause random factors outside of your control. If there were no random effects like this, sigma would be very small and people's SR would change by very little once they were at their proper rank.

10/19/2017 11:04 AMPosted by Shockey
Then there is the learning curve. How much someone learns from the game how fast. MATH can't calculate skill like you are trying.

The biggest problem with the "skill curve" is is based on the Bell Curve. Which is something a few competitive games have started to move away from. Cause it's a terrible way to calculate skill which ends up in people feeling like they are being screwed, cause the bell curve forces everyone to be in the middle. And through my 20 years of competitive gaming it actually works more like a pyramid. Which also handles variables better. In a perfect world a bell curve would maybe work. But it doesn't. This is why I don't play CS:GO anymore.

Here are some references

https://casualaggro.wordpress.com/2013/11/08/the-noob-elitist-bellcurve/

https://blog.skipfault.com/2017/08/02/my-pubg-stats-and-the-top-monopoly-of-the-dedicated/

The graph on the second is the natural skill ratings people fall into when you start everyone at zero. Because casuals on the left make up the vast majority of those players making way for dedicated players to move right on the scale. He mentions in the article that he thinks that if everyone played equal amounts of time that it would be a bell curve.. and holy HE'S RIGHT. But not everyone plays equal time now do they? Also, they don't play with the same vigour.

And I found this while continuing a thought. And this guy has got the nail on the head. https://www.reddit.com/r/Competitiveoverwatch/comments/515dk5/the_bell_curve_distribution_makes_competitive/

The pubg scalene scale is natural to skill levels as it starts at 0 and graduates people through the scale, and LoL and SC2 use this and people complain a LOT less about comp games etc.

Bell Curve just throws everyone into a well.. Bell curve and says.. Have fun with our computer trying to figure out how to put 80% of you in the middle evenly! Peace!

Forum Mod Edit: This post has been edited by a moderator due to language. https://us.battle.net/forums/en/code-of-conduct/


You are very confused about bell curves. Blizzard does not force people to distribute themselves in a bell curve. People do that naturally. As people win and lose and take points from each other, that is how SR ends up being distributed. This happens in many natural processes, and is a consequence of the Central Limit Theorem (https://en.wikipedia.org/wiki/Central_limit_theorem). That is, people are not forced into the mushy middle with most of humanity. People naturally belong in the mushy middle with most of humanity. If some other distribution was more correct, then that would just naturally appear.


I am not confused about bell curves. It's your reliance on the math that clouds your judgement on what it takes to accept skill can't be measured on a bell curve in a competitive format.

When you start at zero for the starting point. Most people never make it half way. The pubg and LoL skill rating proves this. Also, Blizzard IS forcing people into a bell curve as they are the ones enforcing the curve. You didn't read one of those articles I posted by the way you responded and have very little respect from me at this time.

For further information.

http://read.navi-gaming.com/en/team_news/the_new_curve_csgo_matchmaking

This graph shows how cs go was having problems with their bell curve, and they have aligned it further towards a scalene scale moving people down. The system has been better. But it's still not working as well as it could if it was true scalene.

How about you do some math on LoL and the scalene system and tell me how that works out.
10/19/2017 10:26 AMPosted by xsun
Blizz should hire this man for their MM system... At least the simulation agrees what I guessed before:

1) SR is not truly skill rating but somehow Lucky Rating.
2) No conspiracy of forced win/lose game intended towards players but from the bad MM system design ideas.
3) long lose/win streak is abnormally high with the algorithm, it hurts players game experiences.

conspiracy chain: Blizz designed MM system like this and doesn't want to reset SR for each season. -> Blizz wants you know little bit of the MM system. -> Blizz wants you finally give up and buy a new copy of game to hard reset your position. -> Blizz can say to others that they had 35 million copies sold out...


Thanks for the compliment. However, I'm not sure you really understood the post.

As far as improvements go, the main thing is to try to reduce natural variation. That is, reduce sigma. Better banning of throwers, boosters, and trolls would help. A role select queue would reduce variation, but I think this is bad for the game in other ways. However, there is only so much that blizzard can do. 12 random players thrown together are going to behave in random ways, not always under thecontrol of individual players.
10/19/2017 11:02 AMPosted by mrjman

Wut?

This data is simply showing that streaks are a normal phenomena and the SR range that someone can expect is distributed fairly evenly around their "true" rank which in this case was 2500.

Are you sure about "streaks are a normal phenomena"? As I said, any abnormal streak will hurt play experience. You can say the streak is "as designed" but I dont see why blizz need the streak to downgrade user experience. I'd rather believe that is the side effect caused by the MM system.

10/19/2017 11:02 AMPosted by mrjman

I have always thought 500 SR tier size was a purposeful number and its interesting to see the simulation at +-25 basically showing a 500 range (+250/-250) around the "true" skill number.

[/quote] 500 SR can go across two tiers... any well designed rating system wont have this big variations. As I said, the SR system according to the simulation is basically lucky pot.
10/19/2017 12:09 PMPosted by Shockey
<span class="truncated">...</span>

A significant point of the model is to model things like throwers and boosters that cause random factors outside of your control. If there were no random effects like this, sigma would be very small and people's SR would change by very little once they were at their proper rank.

<span class="truncated">...</span>

You are very confused about bell curves. Blizzard does not force people to distribute themselves in a bell curve. People do that naturally. As people win and lose and take points from each other, that is how SR ends up being distributed. This happens in many natural processes, and is a consequence of the Central Limit Theorem (https://en.wikipedia.org/wiki/Central_limit_theorem). That is, people are not forced into the mushy middle with most of humanity. People naturally belong in the mushy middle with most of humanity. If some other distribution was more correct, then that would just naturally appear.


I am not confused about bell curves. It's your reliance on the math that clouds your judgement on what it takes to accept skill can't be measured on a bell curve in a competitive format.

When you start at zero for the starting point. Most people never make it half way. The pubg and LoL skill rating proves this. Also, Blizzard IS forcing people into a bell curve as they are the ones enforcing the curve. You didn't read one of those articles I posted by the way you responded and have very little respect from me at this time.

For further information.

http://read.navi-gaming.com/en/team_news/the_new_curve_csgo_matchmaking

This graph shows how cs go was having problems with their bell curve, and they have aligned it further towards a scalene scale moving people down. The system has been better. But it's still not working as well as it could if it was true scalene.

How about you do some math on LoL and the scalene system and tell me how that works out.


You realize that this thread is mostly tackling 50/50 and random chance.

If you wish to talk about how that is effected by a players skill (performances based SR) then we can do that in saying that it does lead itself to being naturally toward high gold-plat.

In CS:GOs case it was because people used the older system to inflate their ranks which happened with winstreaks and, probably, matchmaking manipulation. A similar thing happened here where winstreaks were toned down to almost non-existence in favor of MMR inflation to quickly boost/drop people out of ranks.

10/19/2017 11:02 AMPosted by mrjman

Wut?

This data is simply showing that streaks are a normal phenomena and the SR range that someone can expect is distributed fairly evenly around their "true" rank which in this case was 2500.

Are you sure about "streaks are a normal phenomena"? As I said, any abnormal streak will hurt play experience. You can say the streak is "as designed" but I dont see why blizz need the streak to downgrade user experience. I'd rather believe that is the side effect caused by the MM system.


Yes, streaks are normal and over a long enough period of time the detriment of throwers, griefers, trolls, ect. gets balanced out by good games and good people. In a sense human nature is its own balancing factor.
10/19/2017 11:21 AMPosted by Macbeth
10/19/2017 10:26 AMPosted by xsun
2) No conspiracy of forced win/lose game intended towards players but from the bad MM system design ideas.


I think the conspiracy is that the system is working.

While I wouldn't put it past any game company to manipulate their players, I think a simpler explanation is that the system is poorly thought out, designed, and implemented.

All of the stats sites have huge amounts of trouble accurately gauging wins and losses if you flex heroes at any point it seems. I always wondered if that was unique to the way they pull the stats, or a problem in blizzards system.
yes the system real working. There may be some side effects like long win/lose streak. But other than that the system pretty much is working as designed.
the only key part of conspiracy is: in order to sell more copies, does the vendor intend to design the MM system to drag someone in low level and force them to hard reset?

Join the Conversation

Return to Forum