Jump to content

Combined Queues - Final Discoveries


849 replies to this topic

#601 Andrzej Lechrenski

    Member

  • PipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 96 posts

Posted 28 May 2020 - 02:12 PM

View PostDaedra117, on 28 May 2020 - 11:21 AM, said:

My opinion is worth zero but id like to point out 2 conditions for the tier reset green light anyway:

1. This:


Not the other (easier) way around. ^This, exactly.



2. Once the new calculations are set, make them public



Spoiler



It's not zero-sum if you do it that way. Instead you will end up with a trend downward and another issue of clustering players in the wrong tier (tier 5 in this case). The only way to balance this is with a zero sum mechanism. The easiest (but not only) way to do that is by straight W/L. Honestly, it baffles me that people waste so much mental energy worrying about a couple of scrubs getting a free carry once in a while when the scale of what we are talking about is so much larger. I feel like there is a political point that could be made by analogy there, but I'm going to bite my tongue.

#602 David Sumner

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 1
  • 470 posts
  • LocationAuckland, New Zealand

Posted 28 May 2020 - 02:18 PM

View PostGladiolix, on 28 May 2020 - 04:57 AM, said:

Do the reset and give everyone "Tier 1 btw" decal as a compensation for losing their exp bar Posted Image


And / Or a Title to display "Ex-Tier 1 player"

#603 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 28 May 2020 - 02:20 PM

View PostSenator Blutarsky, on 28 May 2020 - 12:27 PM, said:

If you only play 12 matches a season, sure anything can happen, but if you play a couple hundred matches and your avg matchscore is say 250+, but your W/L is below 1.... I'm sorry but you don't belong in the top tiers same as the people who just grinded their way there after 5 years.



Again, if the queue were just solo, you'd have more standing to make that statement. But the impact of groups is huge. Take a look at this guy:

https://mwomercs.com...pe=0&user=Brios

Typically about a 1.3 WLR ratio and a solid low to mid 300's match score average. Definitely belongs in the top 10% and you can even watch his games on youtube to assess his play. He's played 500 matches this season, and the merge queue has hurt his match score, but really hurt his WLR. He can barely stay above 1. And this is a guy who does a lot of the intangibles - calls targets, coordinates with teammates, and just a generally good guy to have on your team (low salt too).

He's definitely a Tier 1 type of player (not elite, but good enough) but his WLR this season says maybe not.

I think we probably agree on more than we disagree, just pointing out how WLR is not as useful as it _should_ be.

#604 Andrzej Lechrenski

    Member

  • PipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 96 posts

Posted 28 May 2020 - 02:36 PM

View PostAnomalocaris, on 28 May 2020 - 02:20 PM, said:


Again, if the queue were just solo, you'd have more standing to make that statement. But the impact of groups is huge. Take a look at this guy:

https://mwomercs.com...pe=0&user=Brios

Typically about a 1.3 WLR ratio and a solid low to mid 300's match score average. Definitely belongs in the top 10% and you can even watch his games on youtube to assess his play. He's played 500 matches this season, and the merge queue has hurt his match score, but really hurt his WLR. He can barely stay above 1. And this is a guy who does a lot of the intangibles - calls targets, coordinates with teammates, and just a generally good guy to have on your team (low salt too).

He's definitely a Tier 1 type of player (not elite, but good enough) but his WLR this season says maybe not.

I think we probably agree on more than we disagree, just pointing out how WLR is not as useful as it _should_ be.


My experience with that particular player would seem to be entirely different from yours. I'm not at all surprised that his scores are suffering because I have found his level of team play to be lacking and his instances of needlessly antagonizing people in groups to be frequent. My suspicion is that he enjoyed being congratulated for holding the hands of pugs as they NASCAR'd in a big circle, and when other callers started to show up, he couldn't handle it. Bummer. For him.

#605 David Sumner

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 1
  • 470 posts
  • LocationAuckland, New Zealand

Posted 28 May 2020 - 02:52 PM

View PostJoeCold, on 28 May 2020 - 01:51 PM, said:

Though I'm definitely in favor of the PSR reset even though I don't think it will really fix things unless they make more tweaks to it than were discussed in the OP, I feel a little suspect of the entire OP due to one statement:


This is supposedly over the two weeks prior to the post, a period over which I was wondering why it "felt" like stomps had gone up suddenly. It bothered me since the post so today I did some quick and dirty tracking.

I played 24 matches and tracked the outcomes. I didn't know if there is an official definition of "stomp" in the game. So, I defined "stomp" as the losing team having no more than 4 kills AND the winning team having at least 10 kills. In 24 matches, 13 were stomps by this definition. It's worthy to note that 5 of the 11 "non-stomps" were won 12-5. So, if one has a more loose definition, the volume goes up. I don't think of 12-5 as a stomp, but its certainly not competitive, That means at least 75% were not competitive.

I probably should have recorded more detailed results. But, like I said, it was "quick and dirty" to satisfy the itch that has been drilling at me since the OP. I don't think prior to recently I experienced this much lopsidedness.


I use 12:2 = stomp. 12:3 = meh, better = OK.

Anything where there wasn't a "all of a team dies" makes me cheer because it means someone used the mission objectives.
That happens maybe 1 in 40 matches?

5 of my first 6 matches yesterday were losses.
Half were stomps by my definition. One was meh.

Why would I continue to play?

Edited by David Sumner, 28 May 2020 - 02:56 PM.


#606 David Sumner

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 1
  • 470 posts
  • LocationAuckland, New Zealand

Posted 28 May 2020 - 02:58 PM

Does the PSR have to be wiped? (I'm fine if so)

What would the effect be if you just multiplied it by a percentage such that the average new PSR of active players matched your target average.
(Is there actually a max PSR you can have, and are there actually a lot of people hard at that limit)

Or calculate a new PSR based on the stats for your last 200-300 matches.

Edited by David Sumner, 28 May 2020 - 02:59 PM.


#607 nuttyrat

    Member

  • PipPipPip
  • The Cyber Warrior
  • The Cyber Warrior
  • 94 posts
  • Facebook: Link
  • Twitter: Link
  • Twitch: Link
  • LocationVancouver BC

Posted 28 May 2020 - 03:06 PM

View PostDavid Sumner, on 28 May 2020 - 02:58 PM, said:

Does the PSR have to be wiped? (I'm fine if so)



Paul said it has to be wiped, so the original post answered your question. Even if it didn't have to be wiped, it SHOULD be wiped because there are a lot of people in positions where they shouldn't be.

#608 Mindwyrm

    Member

  • PipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 101 posts
  • LocationIndiana USA

Posted 28 May 2020 - 03:08 PM

31 pages in... so i don't know if i'm repeating anyone.. but it's worth repeating anyway.

I think the thing REALLY upsetting current psr isn't that you move up when you lose and perform well... it's when you move up because you win and perform badly.

#609 David Sumner

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 1
  • 470 posts
  • LocationAuckland, New Zealand

Posted 28 May 2020 - 03:28 PM

View Postnuttyrat, on 28 May 2020 - 03:06 PM, said:


Paul said it has to be wiped, so the original post answered your question.


Said and is are not necessarily the same.

If the PSR is stored as a number (cheap lookup, expensive update), that's different from it being calculated dynamically (expensive lookup. cheap update), and that's different from, it's dynamically calculated, and the new version needs data we don't currently collect (but that could potentially be handled using date boundary checks for some data)

#1 is easy to modify.
#2 happens automatically by changing the calculation
#3 happens automatically when you change the calculation, but you have to be much more careful with the calculation.

View Postnuttyrat, on 28 May 2020 - 03:06 PM, said:


Even if it didn't have to be wiped, it SHOULD be wiped because there are a lot of people in positions where they shouldn't be.


But the whole point of this is to move people from where they shouldn't be to where they should be.
So "change the calculation" has that effect.

The difference is where the starting point is.
Everyone in T3? Good argument being "most people will end up there anyway"
Everyone where they are now? Well, fewer people are unhappy about their Tier being shot down, but they'll (mostly) lose it anyway as they play if they didn't deserve it. Also it might lead to more stomps for a while.
Everyone shifted down a bit? That's a compromise solution, and how "nice" it feels depends on where you have to move stuff to get a median to median shift.

Edited by David Sumner, 28 May 2020 - 04:45 PM.


#610 Uakari

    Member

  • PipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 116 posts

Posted 28 May 2020 - 04:53 PM

View PostMindwyrm, on 28 May 2020 - 03:08 PM, said:



I think the thing REALLY upsetting current psr isn't that you move up when you lose and perform well... it's when you move up because you win and perform badly.


THIS

There are games where I completely **** up and still go up because my team wins. It should be way harder then it is now to go up, and way easier to loose a tier.
If it stays like it is bad players can be in tier 1 again in a couple of weeks.

#611 Senator Blutarsky

    Rookie

  • 8 posts

Posted 28 May 2020 - 05:10 PM

View PostAnomalocaris, on 28 May 2020 - 02:20 PM, said:

Again, if the queue were just solo, you'd have more standing to make that statement. But the impact of groups is huge. Take a look at this guy: https://mwomercs.com...pe=0&user=Brios


This Brios fella is mid-90s percentile. Like you said avg 300+ matchscore. He's gonna get a big move up on wins and a small move down on losses. Even if his w/l is 1 he's still going to wind up at the tippety top when the dust settles just maybe not as high as someone with a similar avg matchscore but a w/l of 2 though Tier 1 no doubt. Maybe its all the groups? I dunno, teamwork is OP, kinda makes sense.

View PostHorseman, on 28 May 2020 - 01:52 PM, said:

What about penalizing being a deadweight? Because that's what it comes down to: players on the winning team who didn't contribute to the win should not profit from it.


Penalizing deadweight is unnecessary as the same works in reverse. If you're carrying some potato to victory he's going to get a small boost on his wins if he's averaging sub 200 matchscore but he'll get a big drop on losses and find himself in tier 4 or whatnot in due course. Plus to be honest I don't trust your or anyone else' definition of deadweight. Winning matters, or else it ain't hockey its figure skating. That deadweight might have kept half the red team occupied while you racked up damage and kills. Maybe kamikaze pilots are a valid winning strategy.

Matchmaking hopefully focuses more on PSR than tiers bc if tiers are going to be evenly distributed its worth pointing out that with the tails of the bell curve t1 will encompass everyone from roughly 250 avg matchscore up to the stone killers running mid-400s. Quite the range.

Also as a consolation for losing precious T1 status I think veterans get something of a boost in that they probably spend most of their time in skilled up mechs versus newer players who are pretty much always in the middle of skilling up something.

#612 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 28 May 2020 - 05:45 PM

View PostAndrzej Lechrenski, on 28 May 2020 - 02:36 PM, said:


My experience with that particular player would seem to be entirely different from yours. I'm not at all surprised that his scores are suffering because I have found his level of team play to be lacking and his instances of needlessly antagonizing people in groups to be frequent. My suspicion is that he enjoyed being congratulated for holding the hands of pugs as they NASCAR'd in a big circle, and when other callers started to show up, he couldn't handle it. Bummer. For him.


We'll have to agree to disagree on that one. I've never group dropped with him, but the guy is more communicative than about 95% of the pugs I've played with and actually responds to calls as well. Never seen him antagonize anyone myself. Really he's about as much as I could hope for in a solo queue when it comes to the intangibles. There are better players out there, but he ticks a lot of boxes in terms of what I look for in solo queue teammates. He's also been surprisingly supportive of the merge queue even though it's hurt his game - can't explain that one except maybe he thinks its better for the game overall (I disagree).

View PostSenator Blutarsky, on 28 May 2020 - 05:10 PM, said:


Matchmaking hopefully focuses more on PSR than tiers bc if tiers are going to be evenly distributed its worth pointing out that with the tails of the bell curve t1 will encompass everyone from roughly 250 avg matchscore up to the stone killers running mid-400s. Quite the range.


This we can agree on. It can't be just tiers because it will repeat the problems of the current matchmaker, just with a different flavor. You've got to match like to like when it comes to player ranking. Again, I'm not optimistic PGI is up for it. We will see.

#613 FRAGTAST1C

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Fighter
  • The Fighter
  • 2,946 posts
  • LocationIndia

Posted 28 May 2020 - 06:50 PM

View PostAnomalocaris, on 28 May 2020 - 05:45 PM, said:

This we can agree on. It can't be just tiers because it will repeat the problems of the current matchmaker, just with a different flavor. You've got to match like to like when it comes to player ranking. Again, I'm not optimistic PGI is up for it. We will see.


I doubt they will. You see, they still haven't figured out how to assess the threat value of a group. That is entirely down to PSR and not Tiers 'cause groups can have mixed tiers.

#614 Andrzej Lechrenski

    Member

  • PipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 96 posts

Posted 28 May 2020 - 06:58 PM

View PostAnomalocaris, on 28 May 2020 - 05:45 PM, said:


We'll have to agree to disagree on that one. I've never group dropped with him, but the guy is more communicative than about 95% of the pugs I've played with and actually responds to calls as well. Never seen him antagonize anyone myself. Really he's about as much as I could hope for in a solo queue when it comes to the intangibles. There are better players out there, but he ticks a lot of boxes in terms of what I look for in solo queue teammates. He's also been surprisingly supportive of the merge queue even though it's hurt his game - can't explain that one except maybe he thinks its better for the game overall (I disagree).



This we can agree on. It can't be just tiers because it will repeat the problems of the current matchmaker, just with a different flavor. You've got to match like to like when it comes to player ranking. Again, I'm not optimistic PGI is up for it. We will see.


Fair enough. I shouldn't have said anything ill, anyway. Its just that we had an argument a few days ago and my rear is still chafed over it.

#615 AnassRhammar

    Rookie

  • 2 posts

Posted 28 May 2020 - 09:11 PM

View PostAndrzej Lechrenski, on 28 May 2020 - 02:28 AM, said:


People are spreading rumors (I won't raise from the ASHes any names) that cadets are getting matched against T1 players. Because of course the den of ego and idiocy that is Outreach HPG would be spreading that.


I'll call you on that.
But they are. I finished my cadet 2 weeks ago, during which I was matched up against justcallmeASH in a QP match with him in a group. Needless to say, I got absolutely decimated.

Tier 4 at the time.


My vote is for yes, there needs to be something done about the psr and would happily support a reset

Edited by AnassRhammar, 28 May 2020 - 09:14 PM.


#616 Horseman

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 4,749 posts
  • LocationPoland

Posted 28 May 2020 - 10:59 PM

View PostDavid Sumner, on 28 May 2020 - 02:58 PM, said:

Does the PSR have to be wiped? (I'm fine if so)

What would the effect be if you just multiplied it by a percentage such that the average new PSR of active players matched your target average.
It wouldn't redistribute the players across tiers the way it should. We'd still end up with a lot of bads in T1 until they play enough matches to drop lower.

Quote

(Is there actually a max PSR you can have, and are there actually a lot of people hard at that limit)
As far as we can tell, there seems to be.

View PostDavid Sumner, on 28 May 2020 - 03:28 PM, said:

If the PSR is stored as a number (cheap lookup, expensive update), that's different from it being calculated dynamically (expensive lookup. cheap update), and that's different from, it's dynamically calculated, and the new version needs data we don't currently collect (but that could potentially be handled using date boundary checks for some data)
Arguably, PGI could seed the initial positions after the reset using past few months of match data even if it's only in the aggregated form on the leaderboards.

#617 Taron

    Member

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 2,180 posts
  • LocationGermany

Posted 28 May 2020 - 11:26 PM

Well, matchmaking is faster, that's true, but it is more idiotic.
I mean: Thake a new idea of making the matchmaking faster, that way, it runs now, is completely stupid.

Edited by Taron, 28 May 2020 - 11:42 PM.


#618 nuttyrat

    Member

  • PipPipPip
  • The Cyber Warrior
  • The Cyber Warrior
  • 94 posts
  • Facebook: Link
  • Twitter: Link
  • Twitch: Link
  • LocationVancouver BC

Posted 28 May 2020 - 11:50 PM

View PostDavid Sumner, on 28 May 2020 - 03:28 PM, said:


Said and is are not necessarily the same.

If the PSR is stored as a number (cheap lookup, expensive update), that's different from it being calculated dynamically (expensive lookup. cheap update), and that's different from, it's dynamically calculated, and the new version needs data we don't currently collect (but that could potentially be handled using date boundary checks for some data)

#1 is easy to modify.
#2 happens automatically by changing the calculation
#3 happens automatically when you change the calculation, but you have to be much more careful with the calculation.



Go back and read his post, the fact that he 'said' it is necessary to do a reset means that it IS necessary. Since he is the game designer I'm happy taking him at his word here.

We don't actually know what Paul's plan is so we can only speculate, however many other games that measure skill on some kind of ranking/elo system go through resets as part of normal operations; Dota2 ranked MMR gets reset regularly (annually I believe) and Solaris 7 ELO resets every season. The current PSR data is just bad, it is inaccurate, and causes a lot of frustration.

Yes, Paul needs to reset Tiers.
Yes, it needs to happen.
No, just resetting the calculation on it's own is not the answer as it doesn't address one of the primary reasons for resetting Tiers; better matches.

Most of the people responding want this to happen and have been asking for it for years.

Quote


But the whole point of this is to move people from where they shouldn't be to where they should be.
So "change the calculation" has that effect.

The difference is where the starting point is.
Everyone in T3? Good argument being "most people will end up there anyway"
Everyone where they are now? Well, fewer people are unhappy about their Tier being shot down, but they'll (mostly) lose it anyway as they play if they didn't deserve it. Also it might lead to more stomps for a while.
Everyone shifted down a bit? That's a compromise solution, and how "nice" it feels depends on where you have to move stuff to get a median to median shift.


So in the OP, Paul already stated that everyone would drop to Tier 4 (which is where you start as a new player in the game) and your first 20 matches will be accelerated, which basically means they are used for placement. Perform very well and you will accelerate quickly out of Tier 4, and if you don't perform as well then you will find your own spot to land.

Just changing the calculation and leaving everyone where they are will not address match quality issues that are present TODAY! Yes, when PSR resets everyone will be considered the same, but that is an expected outcome and will be temporary as players will accelerate at their own pace to where they should be. Resetting will level set everyone and be more fair and accurate calculation of their skill.

All this being said, we don't yet know what the actual PSR/Tier plans are. Heck, we don't even know if Tiers are even going to be a thing! Don't forget that there is already an ELO system for Solaris that they could copy/paste (more or less) which would absolutely require a reset if they went that direction.

#619 DeadWeight18

    Member

  • PipPipPipPipPipPip
  • Survivor
  • Survivor
  • 266 posts
  • LocationLuxembourg

Posted 29 May 2020 - 01:21 AM

View PostZ Paradox, on 28 May 2020 - 08:49 AM, said:

I expect to have less T4 and T5 and more real T1 and T2 in future games.


You won't get that with the low MWO population at hand. Best way for you would be to invite some equal level guy to a private lobby and leave us weekend soldiers alone.

View PostHorseman, on 28 May 2020 - 01:52 PM, said:

...
What about penalizing being a deadweight? Because that's what it comes down to: players on the winning team who didn't contribute to the win should not profit from it.
...


Luck is a fundamental part of life.

View PostHorseman, on 28 May 2020 - 04:38 AM, said:

Simply put, no. Tier is already supposed to be a handicap in that the higher you go the better opponents you're supposed to play against. The reason that doesn't work properly right now is that PSR progression is garbage and allowed too many [political incorrect term] to rotate their way into T1 by sheer number of matches played.


Yeah "supposed" but it does not happen.

Tiering is not currently working with our MW population and your stats show me that you are not pitted often enough against your peers Don't you agree?

Hence if you guys want challenging matches you have to accept to take EXTRA nerfs. Of course I perceive a certain toxic attitude against a balanced = equal kill/death and win/loss ratio. Logic demands if you above 1 the game is TOO easy for you.

BUT that is only my definition of balance.

I wish you guys would show more respect for the feelings of your fellow players. After all the majority of us is supporting your fun here at MWO.

Perhaps we need a visible ranking system based on the time spend in MWO (like in military) and a invisible tiering system to create better matches and "balancers" between different Tiers if tiering does not work.

#620 -P U R E-

    Member

  • PipPipPip
  • The Predator
  • The Predator
  • 71 posts
  • Twitch: Link
  • LocationWarsaw

Posted 29 May 2020 - 02:20 AM

View PostAndrzej Lechrenski, on 28 May 2020 - 02:28 AM, said:


People are spreading rumors (I won't raise from the ASHes any names) that cadets are getting matched against T1 players. Because of course the den of ego and idiocy that is Outreach HPG would be spreading that.


Cadet in trial mech in T1 match.
''Muh rumors ''

Posted Image





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users