Jump to content

Solo Queue - Mm Update (30/05/2018)


159 replies to this topic

#21 A Headless Chicken

    Member

  • PipPipPipPipPipPip
  • The Hungry
  • The Hungry
  • 273 posts
  • LocationImmersed in Stupid.

Posted 30 May 2018 - 07:10 PM

Can we get a PSR reset AND instead make PSR less of an EXP bar, more stringent to climb while we're at this?

#22 justcallme A S H

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2020 Referee
  • CS 2020 Referee
  • 8,987 posts
  • LocationMelbourne, AU

Posted 30 May 2018 - 07:19 PM

View PostPaul Inouye, on 30 May 2018 - 02:10 PM, said:

[color=orange]So what did I do to address this? Here's the scoop.[/color]

1) Stop allowing Tier 1 players to play against Tier 4 and 5 players completely. Hard line.. no ifs ands or buts no matter how low the player count is in off peak times. To do this, the value for Tier separation has been set to 2. The biggest skill gap allowed now is Tier 1 to Tier 3. Tier 2 to Tier 4. Tier 3 to Tier 5.

* Wait times have been reduced between 40-50%.
* MAXIMUM average wait times have dropped from 150 seconds to 80 seconds.
* Matches being played are a lot more challenging.
* Average PSR ranking differences between two teams has gone from 950 to 200.


I am very happy to see the matchmaker getting attention, ecstatic actually given I've dead set had enough of seeing T4 players in my games as I am sure they are sick of seeing me. However what you are doing Paul, while great, is grossly overlooking the major issue with PSR... Let me outline it and why PSR and why it doesn't really work right now.

Through the scraped data we know the 'Average Match Score' (AMS) across the active population of 30k-35k per month is between 200-220 AMS. It's been that way for a while so it's consistent enough to rely on

A player with a 220 AMS should be in, roughly, Tier 3. Maybe Tier 2 in terms of the Median but absolutely not Tier 1...

So what is the reality???

Right now there are players with a 170 AMS ending up in Tier 1... Yep, that's right - Players can make it into Tier 1 that are below the median score of the population.

Being that far below the median should, realistically, see you around Tier 4. Not Tier 3/Tier 2 and absolutely not Tier 1. To truely put attention into the the MM you fix the thresholds and be much tighter. For instance a "Match Win" should not count for anywhere near what it does now.

Once you go and do your internal calculations and tighten PSR/Tiers up so it ranks players properly.Tier 1 needs to be reserved for people 300 AMS or more, not people with 220 AMS (just as an example). Hell Tier1 might be reserved for 330 AMS or more. You'll have to get into the numbers to work out but only the top 5% should be ion Tier 1 at any given time, maybe even less ~3%. I haven't looked that far into the numbers to know exactly but the bulk should be Tier 3 and they certainly aren't if 170 AMS puts you in Tier 1.

I reckon you have two options:
  • Rustle all the Jimmies - Global reset the PSR bar once you fix the rating... And then let the system settle itself over time, which should happen within 30-60 days for most active players. While people won't be happy about being in a lower tier (Those in T1 and should be T4), at least match quality will be improved... Yes some matches will be rubbish while Tier 1 players get out of Tier 3-5, but it won't take long... Live with the short term pain.
  • Slow and painful - Avoid global reset, change the boundaries for Tier / Experience progression and over time those in T1 with 170 AMS will filter back into Tier 4 where they belong. This could take up-to 6 months and players won't like seeing the bar go down slowly over time. It is demoralising.

NOTE: The adjustments to 3/3/3/3, I totally support. Great job there, honestly great job. But it's only a half measure unless PSR is actually fixed.

Edited by justcallme A S H, 30 May 2018 - 11:34 PM.


#23 Kamikaze Viking

    Member

  • PipPipPipPipPipPip
  • 383 posts
  • LocationStay on Topic... STAY ON TOPIC!!!

Posted 30 May 2018 - 07:33 PM

View Postjustcallme A S H, on 30 May 2018 - 07:19 PM, said:


1. Rustle all the Jimmies - Global reset the PSR bar once you fix the rating... And then let the system settle itself over time, which should happen within 30-60 days for most active players.


I vote for this.

My reason being that the first few weeks after PSR was introduced were AMAZING FUN. and it was neat watching the people self sorting into the top tiers in the first few weeks. What followed was a period of probably 2-3 months after that where it seemed that the Matchmaker gods had smiled on us.

Do this once, then do smaller resets every 1 or 3 months based on ensuring that average match scores correleate with PSR and hence Tier levels. (standard distribution, on a curve, whatever you call it)

This will mean Good PSR data in -> Good MM data out! And when combined with Paul's OP the Matchmaker should start working properly again!

#24 Hydrocarbon

    Member

  • PipPipPipPipPipPipPip
  • WC 2017 Qualifier
  • WC 2017 Qualifier
  • 659 posts

Posted 30 May 2018 - 07:38 PM

View PostNavid A1, on 30 May 2018 - 04:59 PM, said:

Have the matchmaker pick out 24 players based on the current rules.


Making a single large team is the answer IMO - you then split it into 2 teams based on a mix of tonnage & skill.

Making 2 "dream teams" separately runs into the problem of matching a junior varsity "dream team" against a seasoned pro team that just won the world championship. You wouldn't see a corgi win a greyhound race for the same reason.

#25 justcallme A S H

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2020 Referee
  • CS 2020 Referee
  • 8,987 posts
  • LocationMelbourne, AU

Posted 30 May 2018 - 07:38 PM

View PostKamikaze Viking, on 30 May 2018 - 07:33 PM, said:


This will mean Good PSR data in -> Good MM data out! And when combined with Paul's OP the Matchmaker should start working properly again!



Yes, exactly. I realise what I am suggesting won't be entirely popular and will cause some pain and upset people. Particularly those not being in "t1" anymore. However for the betterment of the game and the potential of better balanced matches, it needs to happen.

If you fix the way PSR works (I'm happy to help on a taskforce Paul, I also know other players too)... And then tie it in with these MM changes which are excellent - you've then basically done all you can to fix the SoloQ matches and that is what all players want - better matches.

#26 Xetelian

    Member

  • PipPipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 4,371 posts

Posted 30 May 2018 - 07:51 PM

Fix PSR along with these changes.

#27 MisterSomaru

    Member

  • PipPipPipPipPipPip
  • The Death Wish
  • The Death Wish
  • 252 posts

Posted 30 May 2018 - 08:28 PM

Putting myself, my team mates, and my opponents in a match where we belong will improve the solo queue experience vastly. As it stands right now, I will NOT touch solo queue due to the disparity in skill levels. it's simply not fun when you get stacked with new players on your team, and skilled opponents on the other in a tier 1 match. Nor is it fair to the new players.

#28 YUyahoo

    Member

  • PipPipPipPipPipPip
  • 339 posts

Posted 30 May 2018 - 09:25 PM

Thanks for updating us Paul, and giving hope that the quality of match making might actually improve after 2 years of us asking for quality of matchmaking to be addressed. As many people have pointed out, PSR should be addressed along with quality of match making because a 500+ damage game loss (with kills, assists, ect) should never be worth less than a 80 damage win (without any special stats beyond an assist or two)...a pilots average performance should count far more in determining an increase in PSR or decrease in PSR per match than if their team won or lost.

maybe its just me but I also see a problem with how the tiers will be allocated in matches...if a T1 player can play with a T3 player and a T3 player can play with a T5 doesn't that mean that any match can still have a T1 and a T5 (as long as thee is at least 1 T3 player on the team)? As well balancing the number of assaults/heavies/meds/lights on both teams is one thing but total mech tonnage on each side should be a factor as well...because as we all know a 100T assault can carry alot more firepower than an 80t assault (and the same can be said about each weight class both in terms of firepower and armor).

I think you are off to a good start in terms of improving match making quality, but there is much more to consider than just tier restrictions and classes of mechs per side. I know it may be a matter of pride, but I still think if you truly want to balance matches as effectively as possible it might be to PGI's benefit to consult a member of a match/balance team from one of the other competitive fps companies out there to gain additional insight on how better balance can be achieved.

#29 JinSR

    Member

  • PipPip
  • Ace Of Spades
  • Ace Of Spades
  • 32 posts

Posted 30 May 2018 - 11:41 PM

Are things being done similarly to group queue?

I always try to bring new friends into the game and its always the issue where we would group and sit in queue for 20 minutes sometimes (not exaggerating) and then My friends would give up and go play a different game.

I know group queue is abit iffy and is normally discouraged by vet players because the stigma is 'only group if you are coordinated and good' but for the game to grow and get more fresh blood its even harder for Me to introduce friends to MWO if the hook is "Get the game and play by yourself until you get a good grasp of the game and THEN we can play together"

#30 justcallme A S H

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2020 Referee
  • CS 2020 Referee
  • 8,987 posts
  • LocationMelbourne, AU

Posted 30 May 2018 - 11:56 PM

GroupQ is a different beast and it's issues revolved far more around population and trying to 'group match' a 12v12 scenario.

Would suggest leaving that out of discussions like this TBH. It's issues are very much different.

#31 Daidachi

    Member

  • PipPipPipPipPipPip
  • The Clamps
  • The Clamps
  • 461 posts
  • LocationThe Andromeda Initiative

Posted 31 May 2018 - 12:23 AM

I agree with Ash - in saying that, I will also applaud the way you've communicated these changes and the work put in to do them Paul.

Thanks to everyone in this thread for keeping conversation civil as well. No posts have made me roll my eyes and go 'ergh, really?'. :)

#32 Horseman

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 4,694 posts
  • LocationPoland

Posted 31 May 2018 - 12:58 AM

View PostPaul Inouye, on 30 May 2018 - 02:35 PM, said:

This will work as well. Will also put in a request for Match ID in pub matches.
In which case, is there any chance to give players the ability to pull up a list of Match IDs of the pub matches they were in?

#33 mikelovskij

    Member

  • PipPipPip
  • Philanthropist
  • 60 posts

Posted 31 May 2018 - 12:58 AM

View Postjustcallme A S H, on 30 May 2018 - 07:19 PM, said:



Being that far below the median should, realistically, see you around Tier 4. Not Tier 3/Tier 2 and absolutely not Tier 1. To truely put attention into the the MM you fix the thresholds and be much tighter. For instance a "Match Win" should not count for anywhere near what it does now.



I think that "match win/loss" should remain the main factor in determining a pilot tier, while other factors should merely be a way for pilots that consistently perform exceptionally/awfully in the games they are matched in to quickly reach their "real" tier. I know that in a game with 12 players per team, the victory or loss is often determined by the rest of the team, but by averaging over a large amount of matches, the capability of a player to contribute to a win or loss can be extrapolated.

#34 ForceUser

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 894 posts

Posted 31 May 2018 - 01:04 AM

View Post1 21 Giggawatts, on 30 May 2018 - 05:43 PM, said:

Great to see something finally being done about this... Correct me if I am wrong - essentailly rather than reducing restrictions on tier difference you are reducing adhereance to 3,3,3,3 team composition.

Now i dont mean to seem churlish here, but people have been screaming about match quality in solo for what - approximately 2 years now? And all that needed to be done was reduce inflexibility in team comp? Wow....

Ok - if this was such a trivial thing to fix - maybe paul can also look at FW - i reckon there would be some quick wins he can get there. Might bring some people back to this game - myself included.

Thats vastly oversimplifying the problem and omitting key events/facts.

There has always been comlaints about match making. Every game has such complaibts (WoT is infamous for theirs since they dont have any skill based mm). There is constant changes made to the MM based off of what the players complain about. The 3/3/3/3 rule was implented based off of player feedback. Same with loosening the tier restrictions, removal of groups from solo q, implementation of PSR in the first place, etc etc.

It just depends on what the most current issue is the playerbase has.

#35 Unit 86

    Member

  • PipPip
  • Ace Of Spades
  • 45 posts

Posted 31 May 2018 - 02:14 AM

Posted Image

Those changes are totally working. The game outcome is totally not decided in the first 20 seconds where 3 people always die before any actual combat begins.

#36 East Indy

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,207 posts
  • LocationPacifica Training School, waiting for BakPhar shares to rise

Posted 31 May 2018 - 03:40 AM

PGI -- you guys would be better able to maintain Tier compatibility and control weight classes in solo queue by offering players a "First Choice/Second Choice" option.

UI a bit like a drop deck but only one 'Mech is picked by the matchmaker per individual queue. Players can choose to use "First Choice/Second Choice" or not. If they do, they select two 'Mechs, each from different weight classes. "First Choice" is what they'd prefer to play; "Second Choice" is intended as their pick to help speed 23 other players way to the lobby in something close to 3/3/3/3, at most times a light or medium. Similar to map/mode voting, each time a player completes a solo QP match in their "Second Choice" 'Mech, the matchmaker increases that player's priority to drop in their "First Choice," and of course resets the modifier if the "First Choice" drops or if the 'Mech selections change.

Again, players can choose to use this or not.

But this will help the other match quality factor of not having 90% heavies and assaults every time -- a factor that your change might

Edited by East Indy, 31 May 2018 - 05:46 AM.


#37 East Indy

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,207 posts
  • LocationPacifica Training School, waiting for BakPhar shares to rise

Posted 31 May 2018 - 03:43 AM

I'll add that you guys should probably contract Karl Berg for a weekend, but I'm sure you already knew that!

#38 KodiakGW

    Member

  • PipPipPipPipPipPipPipPip
  • The Jaws
  • The Jaws
  • 1,775 posts
  • LocationNE USA

Posted 31 May 2018 - 04:36 AM

How about showing Tiers in the end match screen? Would go a long way in helping you determine if it is working. Better than asking everyone in chat and getting some ***** saying “I am!!!” Causing you to get a lot of screen shots that waste your time.

#39 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 31 May 2018 - 06:44 AM

If PGI can shuffle the players in a weight class between teams to minimize skill gap (not tiers, actual skill), it would minimize the appearance of stomps.

#40 Nothing Whatsoever

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 3,655 posts
  • LocationNowhere

Posted 31 May 2018 - 08:10 AM

Interested in learning what data gets collected out of these changes.


I wonder if we'll see further changes to Heat Capacity and additions like Target Interlocking Circuits in how weapons are grouped together and fired?





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users