Jump to content

Combined Queues - Final Discoveries


849 replies to this topic

#1 Paul Inouye

    Lead Designer

  • Developer
  • Developer
  • 2,815 posts
  • LocationVancouver, BC

Posted 25 May 2020 - 03:41 PM

So, the 4 weeks are up. I want to start by sharing some findings.

The last time I communicated I set the Match Maker configs (release valves) in this manner and left it for a while...

Posted Image


What I didn't tell you was that two Fridays ago after reading all the comments and watching some vids the community made, I made this adjustment to see what would happen over a longer period of time:

Posted Image

Basically I decreased the weight class valve to no longer allow it to open up completely. I also reduced the allowable Tier spread and increased how long the MM would wait until it added another Tier to the viable team composition.

As you can see, the Average Team PSR values narrowed drastically between any two opposing teams. The Weight Class balance would open up to maximum first, then eventually Tier separation would open up. This is why at times in the above graphic will still have tight Tier separation but the Weight Class balance would be a little lop sided.

These settings seem to be working well in terms of finding a relatively happy medium.

* Matches are still kicking off much faster than before.
* Tier/PSR separation is drastically tighter than before.
* Weight class balancing is also much tighter than before.

On another note, when we just combined the queues, stomp rates went up by 5%. With the current settings in the match maker, that went back down to 1%.

It is because of these results, we have decided to keep the current combined queue.

HOWEVER:

There's still an elephant in the room. PSR calculations. Yes it's true that it biases upward movement. That bias comes from the formula currently saying that if you LOSE, you can still move up if you perform well. THIS is the aspect that breaks zero sum distribution.

That being said, we have a plan to make PSR zero sum... but, there's a big draw back. If we make this change, the PSR rankings for all players will need to be reset. That means if you're Tier 1, you'll be going back to Tier 4 for a bit. If you're Tier 5, you're also going back to Tier 4.

There is a multiplier for the first 20 games all players play that push your PSR changes higher or lower to speed the seeding process of the PSR player distribution. After those 20 matches, regular PSR zero sum values will be used.

We need your feedback on whether this is acceptable or not (losing current Tier standings).

#2 Brightsoul

    Rookie

  • Ace Of Spades
  • 3 posts

Posted 25 May 2020 - 04:18 PM

This is great data, and as someone who has a mid Tier 2 account and has played a lot of Tier 1 & Tier 2 matches in the past month, it does seem like things have been a bit more balanced since you tweaked things 2 weeks ago. Though I will say a well coordinated Tier 1 pre-made is much much better than a corresponding number of Tier 1 pugs, and I don't think the matchmaker takes that into account.

Like many others I think that the current PSR system is broken and would appreciate any change that would move it in the right direction, so I would support resetting all PSRs to Tier 4, and to change the calculation for increases/decreases to PSR to make it more zero sum.

It would be better still if there were a way to distinguish between the top 25% of players (currently mostly tier 1?) and the top 1% of players (which are lumped into Tier 1 right now but probably should have their own tier). But, I suppose that's probably too much to wish for. It would be best for all of us if the match maker could try to balance the number of top 1% people on each team...

#3 Paul Inouye

    Lead Designer

  • Developer
  • Developer
  • 2,815 posts
  • LocationVancouver, BC

Posted 25 May 2020 - 04:22 PM

View PostBrightsoul, on 25 May 2020 - 04:18 PM, said:

It would be better still if there were a way to distinguish between the top 25% of players (currently mostly tier 1?) and the top 1% of players (which are lumped into Tier 1 right now but probably should have their own tier). But, I suppose that's probably too much to wish for. It would be best for all of us if the match maker could try to balance the number of top 1% people on each team...


Just want to touch on this. It's known that MWO has a small playerbase now. Even if we do start to separate players into more finite tiers, the release valves would still have to open up fairly wide to get matches to kick off. Even strict Elo would result in people at both ends of the curve waiting forever for a match to kick off.. and by forever, I'm talking in the 50 minute range.

#4 Kageru Ikazuchi

    Member

  • PipPipPipPipPipPipPipPip
  • The Determined
  • The Determined
  • 1,190 posts

Posted 25 May 2020 - 04:23 PM

View PostPaul Inouye, on 25 May 2020 - 03:41 PM, said:

That being said, we have a plan to make PSR zero sum... but, there's a big draw back. If we make this change, the PSR rankings for all players will need to be reset. That means if you're Tier 1, you'll be going back to Tier 4 for a bit. If you're Tier 5, you're also going back to Tier 4.

There is a multiplier for the first 20 games all players play that push your PSR changes higher or lower to speed the seeding process of the PSR player distribution. After those 20 matches, regular PSR zero sum values will be used.

We need your feedback on whether this is acceptable or not (losing current Tier standings).

Do it. Please.

Edited by Kageru Ikazuchi, 25 May 2020 - 04:23 PM.


#5 Brightsoul

    Rookie

  • Ace Of Spades
  • 3 posts

Posted 25 May 2020 - 04:23 PM

Yep I get that. What I'm saying is that if you have a pre-made of top 1%s you need to put a full team of Tier 1 people on the other team to counter balance that, and then give the top 1%s a bunch of tier 2 or tier 3 players. That might make the match less stompy...

#6 Zookeeper Dan

    Member

  • PipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 487 posts
  • LocationBeer City USA

Posted 25 May 2020 - 04:23 PM

I'm for it. Reset the tiers! And this is from a too agressive player that will most likely move down. Worth it if it means better matches.

#7 My Lord and Saviour Jesus Christ

    Member

  • PipPipPipPipPipPip
  • Major General
  • Major General
  • 475 posts

Posted 25 May 2020 - 04:24 PM

Lets go with a hypothetical. Lets say there is only T5s in queue and a 4 man T5 group, and then a T1 group joins queue.

Will they get matched with the T5s as they're the only group? Or will the MM rightfully keep the T1 group out?

#8 Gagis

    Member

  • PipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 1,731 posts

Posted 25 May 2020 - 04:29 PM

View PostPaul Inouye, on 25 May 2020 - 03:41 PM, said:

If we make this change, the PSR rankings for all players will need to be reset.

I am perfectly fine with this.

#9 AnUndeadMonkey

    Rookie

  • Legendary Founder
  • Legendary Founder
  • 9 posts
  • LocationUSA

Posted 25 May 2020 - 04:33 PM

Don’t rest any tier rankings. I’m not going to grind from T4 to T1 again. Last time I sat in T3&T2 for close to a year before moving out of them. I’ll quit, and I played since beta. The matches are far from balanced. I have had matches were one team has had 1 assault class Mech and the other had 4 assault class mechs, this past Saturday. The group queue and solo mix screwed everything up. You have 4 mans running off doing their own thing ignoring the rest of the “team” and everyone loses by a 12-0 stomp. The solo QP games are trash, bring back knockdown and escort since you want the game play to suck. Must really be trying to push players away.

#10 Wesxander

    Member

  • PipPipPipPipPipPip
  • Field Marshal
  • Field Marshal
  • 319 posts

Posted 25 May 2020 - 04:35 PM

Against it. The obvious cow in the room is you have not played in enough matches to understand that so called stomp matches happen over and over again. Nothing like drawing an elite 4 man on the opposing team that you know is fielding the best elites or mediums or lights as team that can be had then you look over at your side and it's composed of tier 3 players and 2 tier 1's. Example lets say we draw the elite 4 man team concept a lot are playing 3 MC II and light scout for them. Then you factor in the programs has also given them 2 tier 2 players in elite assaults. Then on the other side you have casual 4 man team that are in 2 heavy and 2 mediums and not nearly the same skill level. Additionally match maker just for fun throws in 2 guys in assaults that are armed with specialty fail mechs. Things like say a assault that's all LRM's on close combat map. The personal favorite of some guy being jerk with 6 flamers on 85 plus ton mech. Yep real competitive there. You have on your side 2 assaults that are not worth their weight on or off paper while the other team has 5 assaults' 3 of which are top end elites. What could go wrong there. It's all good because of our 2 tier 1's are in lights or medium and can make up for that error.

Edited by Wesxander, 25 May 2020 - 04:37 PM.


#11 Warning incoming Humble Dexterer

    Member

  • PipPipPipPipPipPipPipPip
  • Bridesmaid
  • 1,077 posts

Posted 25 May 2020 - 04:38 PM

View PostPaul Inouye, on 25 May 2020 - 03:41 PM, said:

What I didn't tell you was that two Fridays ago after reading all the comments and watching some vids the community made, I made this adjustment to see what would happen over a longer period of time:

We noticed : There's even a thread in the General forum from a player ragequitting over the weight disparity he wasn't told about.

View PostPaul Inouye, on 25 May 2020 - 03:41 PM, said:

There's still an elephant in the room. PSR calculations. Yes it's true that it biases upward movement. That bias comes from the formula currently saying that if you LOSE, you can still move UP if you perform well. THIS is the aspect that breaks zero sum distribution.

Just make that formula say that if you WIN, you can still move DOWN if you perform bad. Because PSR is meant to rate the player's performance, not his luck.

View PostPaul Inouye, on 25 May 2020 - 03:41 PM, said:

That being said, we have a plan to make PSR zero sum... but, there's a big draw back. If we make this change, the PSR rankings for all players will need to be reset.
We need your feedback on whether this is acceptable or not (losing current Tier standings).

First fixing the PSR formula and then resetting the PSR rankings is one of the oldest standing player requests.

#12 CptJudas

    Member

  • PipPip
  • The Tip of the Spear
  • The Tip of the Spear
  • 22 posts

Posted 25 May 2020 - 04:39 PM

PSR reset should have happened ages ago. Please move forward with it. Everyone knows how broken it is and nobody enjoys it.

#13 John Bronco

    Member

  • PipPipPipPipPipPipPip
  • The Fighter
  • The Fighter
  • 966 posts

Posted 25 May 2020 - 04:41 PM

Please reset PSR.

#14 Kin3ticX

    Member

  • PipPipPipPipPipPipPipPipPip
  • The People's Hero
  • The People
  • 2,926 posts
  • LocationSalt Mines of Puglandia

Posted 25 May 2020 - 04:42 PM

reset it

#15 D U N E

    Member

  • PipPipPipPipPip
  • Bad Company
  • Bad Company
  • 131 posts

Posted 25 May 2020 - 04:44 PM

Better PSR - This is... I can't believe it.

#16 Kurlon

    Member

  • PipPipPip
  • 59 posts

Posted 25 May 2020 - 04:44 PM

At the moment, PSR doesn't mean anything anyways, (watching cadets match against T1s right now) so resetting it can't negatively impact games any more than leaving things as is. So yeah, reset my PSR, no worries.

On the stomp side, please don't report vague percentage increases, as we have no basis to evaluate that on. A 3% increase if stomps are already at 50% of matches is different than if they were 2% of matches. What is the percentage of matches that show as a stomp at the moment?

#17 Ballistic Panicmode

    Member

  • PipPipPip
  • The Scattershot
  • 53 posts
  • LocationMaryland, USA

Posted 25 May 2020 - 04:46 PM

I'm in support of trying this at least, as I've felt out of place since I hit tier 1 a few years ago. Going by skill, I should probably be somewhere in tier in a proper PSR system. Due to the other issues mentioned, however, I'm afraid tho only true solution would be to track average match scores on a per variant basis, and using those to seed the teams. It would account for both "meta" and "troll" builds, as well as players working outside of their comfortable styles. I have no clue how difficult such a system would be to implement, but a fair MM system might go a long way toward bringing people back to the game. (not to mention retaining more of the current player base)

#18 BARRY SHlTPEAS

    Member

  • PipPipPip
  • Overlord
  • Overlord
  • 80 posts
  • LocationBrown Sea Buccaneer

Posted 25 May 2020 - 04:48 PM

View PostKurlon, on 25 May 2020 - 04:44 PM, said:

At the moment, PSR doesn't mean anything anyways, (watching cadets match against T1s right now) so resetting it can't negatively impact games any more than leaving things as is. So yeah, reset my PSR, no worries.

On the stomp side, please don't report vague percentage increases, as we have no basis to evaluate that on. A 3% increase if stomps are already at 50% of matches is different than if they were 2% of matches. What is the percentage of matches that show as a stomp at the moment?


Yeah what this dude said.

EDIT: Oh yeah, on the 19th Steam applied a 7.4MB update to MWO, but no patch notes. What did you do there?

Edited by -Verti-, 25 May 2020 - 04:53 PM.


#19 D U N E

    Member

  • PipPipPipPipPip
  • Bad Company
  • Bad Company
  • 131 posts

Posted 25 May 2020 - 04:49 PM

To anyone not wanting this - Not being Tier 1 isn't a bad thing. It's a system to ensure you are on the right side of the field. It shouldn't be a bar to grind. It should be a system to ensure fair balance in games to avoid stomps and make closer matches.

#20 Krasnopesky

    Member

  • PipPipPipPipPipPip
  • CS 2022 Gold Champ
  • CS 2022 Gold Champ
  • 217 posts

Posted 25 May 2020 - 04:50 PM

Reset the tiers and make them zero sum. A lot of us have wanted this for years.





5 user(s) are reading this topic

0 members, 5 guests, 0 anonymous users