Jump to content

Stats Study: Matchmaker Is Unfair

Balance

344 replies to this topic

#81 Too Much Love

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 787 posts

Posted 20 April 2017 - 02:01 AM

View PostTarogato, on 20 April 2017 - 01:52 AM, said:

So then why do you often get teams stacked with all the best players on one side, and nobodies on the other?

That's exactly the question I'm asking.

Really - why?

If it was only the PSR problem, then situations like you described could happen sometimes.

But what we see - and what I tried to show in my original post - it happens all the time constantly. If there were "real" Tier 1 and "fake" Tier 1 they would be distributed evenly across the teams. Some "real" and some "fake" in team A, and some "real" and some "fake" in team B.

But how can it be that the team A gets all "real" Tier 1, and team B gets all "fake"?

It means some mechanics envolved.

#82 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 20 April 2017 - 02:29 AM

View PostTarogato, on 20 April 2017 - 01:52 AM, said:

So then why do you often get teams stacked with all the best players on one side, and nobodies on the other?


Simple, because these best players lost enough for the day and since it is much easier to accomodate a win for them when they are all on one team MM does that ...

View PostTarogato, on 20 April 2017 - 01:52 AM, said:

Because if you really think that is how the matchmaker works (tries to force people toward 1.0), than it is failing even worse than the one we actually have (which is trying to match players of equal skill.)


I don't think anything. I consider things I'm seeing, is all. Nine out of ten matches I'm in I can predict the outcome by just looking at names. Eight out of ten times the prediction is correct and there is a stomp. Since it is obvious to me, it should be even more obvious to a MM.

#83 James Argent

    Member

  • PipPipPipPipPipPipPip
  • 721 posts

Posted 20 April 2017 - 08:57 AM

I think it's a mistake to focus at all on the win/loss record. (I readily acknowledge that this may be what prompted the investigation, however.) We're not looking for equality of outcomes, but equality of opportunity. If the matchmaker is intentionally skewing the team composition, that's enough of a problem even if all matches resulted in a tie.

If the matches are made properly, the teams should be as even as is statistically possible. That's it. The resulting win/loss is just an outcome. Sure, it's the outcome about which we most directly care, but let the players decide that through playing the match. There will still be 12-0 (and 0-12) stomps because momentum will always be a thing no matter who's on either side. If the teams are equal, over a statistically significant number of matches the win/loss ratio will tend to take care of itself.

#84 Widowmaker1981

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Widow Maker
  • The Widow Maker
  • 5,031 posts
  • LocationAt the other end of the pretty lights.

Posted 20 April 2017 - 09:08 AM

The matchmaker (in solo queue, anyway) literally only looks at your tier, that is all. It doesnt know or care about your W/L, your K/D, or any of your other stats. A T1 is a T1, and a T3 is a T3. It tries to match tier level on each side. That is how it is described to work, and there is absolutely no way PGI went to the trouble of designing some complicated algorithm to try and keep everyones win / loss at 1, because why would they do that?

It produces games of highly variable quality because its basing all of its information on an upward biased PSR system that cant tell how good people are, and also because at times it doesnt have many players to draw from, i assume.

#85 Vanguard319

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,436 posts
  • LocationTerra

Posted 20 April 2017 - 12:33 PM

Rejoice, all who feel the MM is crap, vindication is ours!

#86 Too Much Love

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 787 posts

Posted 20 April 2017 - 01:15 PM

View PostJames Argent, on 20 April 2017 - 08:57 AM, said:

I think it's a mistake to focus at all on the win/loss record. (I readily acknowledge that this may be what prompted the investigation, however.) We're not looking for equality of outcomes, but equality of opportunity. If the matchmaker is intentionally skewing the team composition, that's enough of a problem even if all matches resulted in a tie.

If the matches are made properly, the teams should be as even as is statistically possible. That's it. The resulting win/loss is just an outcome. Sure, it's the outcome about which we most directly care, but let the players decide that through playing the match. There will still be 12-0 (and 0-12) stomps because momentum will always be a thing no matter who's on either side. If the teams are equal, over a statistically significant number of matches the win/loss ratio will tend to take care of itself.
But the very problem is that the teams are NOT equal. They are created unequal.

Edited by drunkblackstar, 20 April 2017 - 01:27 PM.


#87 Too Much Love

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 787 posts

Posted 20 April 2017 - 01:23 PM

View PostWidowmaker1981, on 20 April 2017 - 09:08 AM, said:

The matchmaker (in solo queue, anyway) literally only looks at your tier, that is all. It doesnt know or care about your W/L, your K/D, or any of your other stats. A T1 is a T1, and a T3 is a T3. It tries to match tier level on each side. That is how it is described to work, and there is absolutely no way PGI went to the trouble of designing some complicated algorithm to try and keep everyones win / loss at 1, because why would they do that?

It produces games of highly variable quality because its basing all of its information on an upward biased PSR system that cant tell how good people are, and also because at times it doesnt have many players to draw from, i assume.

May be. But

1) Who really knows how it works? How can we discuss it, if we simply don't know. It's all rumors + some little pieces of information PGI made public.
But know we can analyze the comparision of teams in detail, that is the result of MM work.

2) Your words don't explain the things like streak-defeats or streak-victories. If the teams were randomly picked up by tiers only, they would be more or less even chance of defeat/victories. The situation of 10 defeats-stomps in a row wouldn't be possible.

View PostAlabaster Croft, on 20 April 2017 - 09:17 AM, said:

I don't like this post.
I'm too busy enjoying the game

Why are you wasting your time writing this post then? You better enjoy gaming.

Edited by drunkblackstar, 20 April 2017 - 11:02 PM.


#88 Leeroy Lazer Vomit Kerensky

    Member

  • PipPip
  • Ace Of Spades
  • Ace Of Spades
  • 23 posts

Posted 20 April 2017 - 01:34 PM

These posts always make me laugh, like every match should be a nail biter or else its "gg matchmaker".

#89 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 20 April 2017 - 01:44 PM

View PostLeeroy Lazer Vomit Kerensky, on 20 April 2017 - 01:34 PM, said:

These posts always make me laugh, like every match should be a nail biter or else its "gg matchmaker".


I think it's awesome that NKVA is so dedicated to the lore - all Kerenskys had silly names and were pessimistic and pretentious as ****.

It's like you guys are living in a MW LARP.

#90 Dryderian

    Member

  • PipPipPipPipPip
  • Philanthropist
  • 112 posts

Posted 20 April 2017 - 02:08 PM

The time I raged because of the matchmaker is over. This is a team game and if the voice chat is silent I try to take the lead and in most cases people chime in and something like coordinated match takes place - sometimes I take the lead, sometimes it is a mix of several people, sometimes another one takes lead.

If you lose you can say you have at least tried instead of dying with everyone in silence on voice com. Sometimes the team does not chime in on voice com and that is in most cases a bad sign. After some lost matches with people not following instructions or staying passive I might get a bit angry dying alone and in most cases it is a loss again.

What I take from that, try to lead, try to call targets in many cases the team will follow or some other guys chime in. Most teams lose because there is no coordination and utter silence on voice com.

The matchmaker does not decide the win, coordination does in many cases and if the lead or the live voice com can fetch people.and take them out of their usual bounds.

Edited by Dryderian, 20 April 2017 - 02:11 PM.


#91 Escef

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 8,530 posts
  • Twitter: Link
  • Twitch: Link
  • LocationNew England

Posted 20 April 2017 - 02:16 PM

12 matches is far, far too small a sample size to be of any worth.

#92 Humpday

    Member

  • PipPipPipPipPipPipPipPip
  • The Pharaoh
  • The Pharaoh
  • 1,463 posts

Posted 20 April 2017 - 02:22 PM

I dunno, I've finally dug myself out of my n00b hole to attain 1.3 w/l. Not sure I've noticed any sudden swing.
I've more noticed the time of day I sign on or day itself determines play quality. Like yesterday, I signed on and the teams got stomped 3x in a row...like stomp stomp as in I'm the only one who killed something, stomp. I proceeded to shelf the game for the rest of the day after that.

The day before however it was opposite, I ran 6-8 games straight without losing once, closer games, but I never died once and racked 1-5 kills in each match, more toward the 2-3 kill area.

Your sample size is too low to be of any factual value. You'd need to scale that out to hundreds or thousands of matches while recording the time and zone in which you play. Those are variables that will swing the result violently.

Maybe do like another run of like 50 games? See if you can repeat your results?

#93 Exilyth

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bridesmaid
  • 2,100 posts
  • LocationTerra

Posted 20 April 2017 - 02:47 PM

View Postdrunkblackstar, on 19 April 2017 - 09:06 AM, said:

Among 12 analyzed matches ...

... over 100 matches data


Needs more samples: https://en.wikipedia...e_determination

#94 The Jerol

    Member

  • PipPipPip
  • Survivor
  • 59 posts
  • LocationNorthern California (Inner Sphere)

Posted 20 April 2017 - 02:56 PM

This is all rather silly.

MANY other games have figured this out. It's not hard. PGI is tracking all the relevant stats they need to make a decent matchmaker. You balance each game based on chassis weight and average match score in that weight class. That would be a start, anyway. Make sure the teams have roughly equal average match scores and I think you'd see a lot more competitive games.


TJ

#95 Alistair Winter

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Storm
  • Storm
  • 10,823 posts
  • LocationBergen, Norway, FRR

Posted 20 April 2017 - 03:02 PM

View PostBud Crue, on 19 April 2017 - 09:17 AM, said:

I guess the results are not exactly a shock. I mean it would have been a shock to me if the comparative stats on each team were remotely similar. But what you found? It's depressing but not surprising.

The main thing to take away from the OP, if we assume that we can generalize from 12 matches to most matches, is that it contradicts the common "truism" in the MWO forums where deathballing is 1) inevitable and 2) random. In other words, either team is likely to get 1 or 2 kills first, and then it just starts snowballing from there, thus making the ultimate result of the match pretty random.

I have argued that the first kill and snowballing phenomenon as similar to a study I saw a while ago about a world jiujitsu championship. The study showed that the athlete that gets the first point (by getting a dominant position) usually wins the match. Now, some people will say that this shows the importance of a dominant position. However, it seems more likely that the best athlete is most likely to score the first point anyway.

Same thing with first blood in MWO. Snowballing is a real thing, but the best team is likely to get the first kill or, certainly, the first 2-kill lead that really speeds up the snowballing.

#96 slide

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,768 posts
  • LocationKersbrook South Australia

Posted 20 April 2017 - 10:29 PM

The is assumption that the match maker tries to balance the team. That might be it's intent, but the people who have hit launch at the precise time it is trying to make a match would have a far bigger influence than most people would give credit for.

In an ideal scenario the MM has 24 players to choose from. Ideally some metric is used to rate those players from 1-24. to create a roughly equal match you would put all the odd players on one team and all the evens on the other (school yard pick style).

What I think is happening most of the time (particularly during low population times) is the match maker is grabbing 12 players and making a team (because they have been waiting). It then grabs the next 12 players that hit launch, and because the valves have opened already, it launches the match. It would be similar to taking the above 24 players and putting in a teams split 1-12 and 13-24. Which would give an entirely different out come to the match up.

The prevalence of matches where you are teamed with or against the same players match after match has got more to do how people finish a game and how long it takes them to press launch again than it has to do with MM trying to even up the teams.

MM is all but non functional except during the highest population times. Even if there were 2400 (a number based on roughly double the Steam count) people playing at a given time. That's only 100 possible matches it can launch. With the vagaries of match length, random logins, mech choices, PSR etc the MM will have only a small percentage of people it can actually put in a game. At any given time there are maybe 60-100 people for the MM to work with, that doesn't give it a whole lot of choice about who's on what team.

I doubt that even at peak times the MM is launching more than 1 in 10 matches in the way it should be. Bump the population up to 24k and it might be a different story.

#97 Too Much Love

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 787 posts

Posted 20 April 2017 - 11:01 PM

View PostLeeroy Lazer Vomit Kerensky, on 20 April 2017 - 01:34 PM, said:

These posts always make me laugh, like every match should be a nail biter or else its "gg matchmaker".

It's so cute that you liked your own post.
Posted Image

Edited by drunkblackstar, 20 April 2017 - 11:22 PM.


#98 Too Much Love

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 787 posts

Posted 20 April 2017 - 11:19 PM

View PostEscef, on 20 April 2017 - 02:16 PM, said:

12 matches is far, far too small a sample size to be of any worth.


View PostHumpday, on 20 April 2017 - 02:22 PM, said:

You'd need to scale that out to hundreds or thousands of matches while recording the time and zone in which you play.


View PostExilyth, on 20 April 2017 - 02:47 PM, said:


For all those guys who SUDDENLY became an experts in statistics:

1) I knew that there would be comments like that (though I didn't expect that there are so much statistics professors playing MWO), so I made the special remark concerning this problem in my original post. Apparantly, it didn't help. Nobody reads the post till the end, I know.

2) The Tarogato's study with much more samples showed the same results (but the conclusions were different).

3) You should understand one important thing about matchmaker: it's not a public opinion, it's not some natural phenomena, it's a mechanism. It is designed to produce the same results. It's like you are examining some machine that authomatically produces wooden blocks (sometimes the produced blocks are flawed, and it brings some distortion to the stats). It's not necessary to have 1000 of samples to prove that every time this machine makes a wooden block.

#99 Too Much Love

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 787 posts

Posted 20 April 2017 - 11:33 PM

View PostAlistair Winter, on 20 April 2017 - 03:02 PM, said:

The main thing to take away from the OP, if we assume that we can generalize from 12 matches to most matches, is that it contradicts the common "truism" in the MWO forums where deathballing is 1) inevitable and 2) random. In other words, either team is likely to get 1 or 2 kills first, and then it just starts snowballing from there, thus making the ultimate result of the match pretty random.

I have argued that the first kill and snowballing phenomenon as similar to a study I saw a while ago about a world jiujitsu championship. The study showed that the athlete that gets the first point (by getting a dominant position) usually wins the match. Now, some people will say that this shows the importance of a dominant position. However, it seems more likely that the best athlete is most likely to score the first point anyway.

Same thing with first blood in MWO. Snowballing is a real thing, but the best team is likely to get the first kill or, certainly, the first 2-kill lead that really speeds up the snowballing.

The model of "snowball" works well when the teams are equal (e.g. some championship where all the participants have the same skill level, so those who get the first advantage win).

I don't believe that it suits MWO case.

If the teams are created unequal, team A is strong and team B is weak, then there is a very high possibility that team A will get the first and second kills. It looks like "snowball effect", but in fact it only mirrors the force balance.

Usually it happens this way. The team B has a lot of "noobs", so one of them or two go straight to the enemy positions and die immediately. It's not a snowball effect, it shows the skill level of team B.

#100 Shifty McSwift

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,889 posts

Posted 20 April 2017 - 11:37 PM

They need to just make QPsolo/group non ranked casual mode, and force people to make units or groups for 'ranked' play keeping it at the minimum of a lance (4 players).





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users