Jump to content

Combining Group And Solo Queues - 4 Week Test


1579 replies to this topic

#561 Paul Meyers DEST

    Member

  • PipPipPipPipPipPipPip
  • CS 2024 Top 25
  • CS 2024 Top 25
  • 545 posts

Posted 30 April 2020 - 08:11 AM

View PostBrauer, on 30 April 2020 - 06:58 AM, said:

Win loss is clearly not stuck at 1.00 for everyone. If you look up some player stats you'll see that a number of players have sub 1.00 WLR and many have a WLR of over 2.00. These ranges are just going to expand further under this new system as good players who drop with other good players will able to achieve win rates upward of 90%.


This is making a Win/Loss based Matchmaker even more important,

#562 justcallme A S H

    Member

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

Posted 30 April 2020 - 08:16 AM

View PostRickySpanish, on 30 April 2020 - 05:38 AM, said:

So here is a thought - the Good Players don't enjoy stomps in the new SQ/GQ merger, so why play it? I call BS on that. Maybe you guys think you don't enjoy it, but something's making you play, eh? And of course, regular solo queue was absolutely the pinnacle of competitive play beforehand, right? So the only thing I can think of that you are missing is the old segregated group queue which correct me if I'm wrong, was rather dead already.



Nowhere did I say they would not play it. Also most people I know have barely played GroupQ in 6-8 months. GroupQ was mostly filled with low-mid skill users for the last 6 months. They became more and more upset when High Skill groups "invaded" "their" nights. I have even had them PM me to "stop us dropping"...

Anyway high skill users are playing it because they can group with like-minded individuals and essentially, prove the point.

Merging the queues does not make for more competitive matches for many. If the only way to make PGI see the argument is to give them data - well - give them data.

SoloQ was at the least, somewhat reasonable. 1 player could impact it, sometimes. All the matches people I know are playing in now - are complete stomp-fests for the most part. No one wins in this scenario.

#563 Horseman

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 4,746 posts
  • LocationPoland

Posted 30 April 2020 - 08:26 AM

View PostAlreech, on 30 April 2020 - 12:53 AM, said:

Solo & Group match making ist a different problem.

In Solo the match maker has to consider:
- Players Tier
- Players Mech Tonnage

In Group the match maker has to consider:
- Group Size
He also should consider "group tier" based on the players tier, but can't be done because not enough groups are aviable to set up matches due problematic group sizes (10+2, 9+3, 6+2, 5+3).

There are SOME modes where mech speed should be taken into account when building both teams, but anywhere where combat effectiveness is concerned you don't actually have to account for the mech class... just how the player performs with that mech. A poorly ran assault may not actually be worth more than a good medium player.

The problem is that the matchmaker as it is doesn't use any meaningful performance statistics (WLR, average MS, KDR) and instead matchmakes by cucumber.

Edited by Horseman, 30 April 2020 - 09:07 AM.


#564 BenMillard

    Member

  • PipPip
  • Moderate Giver
  • Moderate Giver
  • 38 posts
  • Twitch: Link
  • LocationLondon, UK

Posted 30 April 2020 - 08:26 AM

Occassional random changes have kept this game alive, year after year. The one thing it must not do is remain the same, as that will make it feel repetitive and stale.

I, for one, welcome our new 4-player overlords.

If this forces/encourages more solo players to enter LFG and make new friends, that can only be a good thing. With a shrinking population, it is in all our interests to make more friends amongst the players who remain.

It has certainly brought a bit more structure to QP games. It is now viable to use other tactics and expect a more organised response from the other side.

The limit of 4 players reduces the ability of meme pre-mades trolling en masse as happened in the 8vs8 test. It's a good balance. It's never going to be perfect and, indeed, perfect matchmaking would be a weirdly artificial experience.

Re-labelling "LFG" to simply say "Group" and using text labels for the 3 social icons at the bottom of the screen would make these valuable features more discoverable and intuitive. Solo players might never have used "LFG" since that is meaningless until explored. And they won't have text chatted during the ertswhile solo queue era since they were unlikely to solo drop with the same people on their team.

Experienced players will understand this simple change of label; they will continue to group up in their current ways.

Does this scale, though? If most players now learn to form groups but only 1 group is placed on each team, what then? QP matches may take as long to start as FP while the abundant groups wait for solo players to fill the other 8 slots.

The only issue I've seen is one or two extremely new players (2 games and 140 games, specifically) being on the same side, facing a a >90% 4-player group. They probably thought their game was broken - and it terms of matchmaking, it was.

While the tier system indicates experience rather than skill (and I'm a case in point for this) that's still a useful metric for keeping the least experienced players in a somewhat more forgiving play-pen. The old "your current level +-2 levels" logic for matchmaking seems to have been lost during the integration? Maybe it was just low population at that moment but I have seen some dramatic differences in awareness in QP since the integration.

Faction Play logic of ranking groups and placing them on opposite sides, then filling in the remaining slots with solo players to balance tonnage and ranks, actually works well. Especially when group sizes are fairly small, such as 2-4 players each. Applying that to QP, perhaps with the match countdown and visible queue length, would be a logical move given the proven success of those systems.

Other games tend to do it that way, don't they?

So long as you keep changing the game, people will keeping playing it. For the novelty value if nothing else.

Edited by BenMillard, 30 April 2020 - 08:34 AM.


#565 justcallme A S H

    Member

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

Posted 30 April 2020 - 08:32 AM

View PostSpare Knight, on 30 April 2020 - 07:08 AM, said:

I am a solo only player, so that doesn't help me. I would have to say that I am probably T4 or at best T3, but I am set as T2 and will soon hit T1. I sure don't have the skills to belong there. Even if I have a bad performance, if the team wins, I still get an =.


I mostly SoloQ as well.

I have a WLR well, well above 1.0.

So what you are saying is factually incorrect. The MM currently does not at all aim, or try to keep, players at 1.0 ratios.

#566 Gakuseinozen

    Member

  • PipPipPip
  • Giant Helper
  • 55 posts

Posted 30 April 2020 - 08:38 AM

View PostNearly Dead, on 30 April 2020 - 06:39 AM, said:

Took yesterday off and tried some games this morning;

12-3 Silent
6-12 Called
2-12 Silent
11-8 Silent
6-12 Silent
12-5 Silent
2-12 Silent
3-12 Silent
8-12 Called
3-12 Silent

Only two games had any voice communications in the game. If there was outside comms obviously I would not have been aware of it.





Guys, please don't stop using the in game comms. The quality of the game client VOIP system is one of the best things about MWO. Discord chat with your Lance is nice between matches, but set your Teamspeak or Discord to the same PTT key as in-game comms so the ENTIRE team is on the same page during matches, not just your 'mates.

Edited by Gakuseinozen, 30 April 2020 - 08:40 AM.


#567 Brain Cancer

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • 3,851 posts

Posted 30 April 2020 - 08:43 AM

View Postjustcallme A S H, on 30 April 2020 - 08:32 AM, said:


I mostly SoloQ as well.

I have a WLR well, well above 1.0.

So what you are saying is factually incorrect. The MM currently does not at all aim, or try to keep, players at 1.0 ratios.


It tries. If you can carry hard enough, though- it sometimes just can't put enough bad players on a team to compensate for it.

In other news, attempting to "fix" groups by reducing weight will likely result in this happening.

Posted Image

160 tons, 11/12 kills.

Weight isn't the issue- we're seeing people manage to end up with nearly all-assault matches with the currently borked MM, including matches where the better 4-man lance's team is considerably out-tonnaged. The issue is you've let organized, skilled players get together in sufficient numbers to easily skew every "solo" game they're in.

#568 Spare Knight

    Member

  • PipPipPipPipPip
  • 119 posts
  • LocationAlso StankDog the damp

Posted 30 April 2020 - 08:45 AM

View Postjustcallme A S H, on 30 April 2020 - 08:32 AM, said:


I mostly SoloQ as well.

I have a WLR well, well above 1.0.

So what you are saying is factually incorrect. The MM currently does not at all aim, or try to keep, players at 1.0 ratios.

Well, I guess I am wrong. Posted Image

Please help me to understand how I keep going up in Tier when my Average Match Score is 230?

At least now, I do see players of my skill level, with the tier levels thrown out the window.

#569 Shaggy2176

    Rookie

  • Littlest Helper
  • Littlest Helper
  • 2 posts

Posted 30 April 2020 - 08:45 AM

Yeah this is garbage. Had to matches where we had 1 assault and the opfor had 3 or more assaults. I'll go play something else.

#570 Horseman

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 4,746 posts
  • LocationPoland

Posted 30 April 2020 - 08:47 AM

View PostGrumpy Old Man, on 30 April 2020 - 02:54 AM, said:

So one suggestion that might be quick to be implemented is a check for Mech performance. A lot of concerns are about people derping about vs. people playing meta builds. Why don't we use the Solaris System to classify the strength of Mechs? Whichever Mech is picked for the group, the OPFOR get a mech within one deviation of the same class with the same Solaris classification. So a "Tier 2" Assault Mech on one side could be offset by a "Tier2" Heavy Mech, for example, or a "Tier 1" Heavy Mech but not a "Tier 2" Medium Mech, to not cause problems with weight differences.

The second part of the majority of concerns seems to be about player skills. Since the two Tier PSR differences seem be be out the window... Why doesn't PGI calculate the Match Score average of the past 12 months or so per weight class (Lights, Mediums, etc.) once per month, divides them into some some brackets and then crossreferences them with the Mech table. We have that information already. In theory they could do the calculation in Excel for all I care, they already have all this information on the webpage... Have the system pick whoever is closest in brackets within the weight performance and form up groups. This is not perfect, because if a Top 10 player is playing at the same time as the worst 23 players in the game, they will still get matched together. But in the spirit of a MVP, this might be something relatively easy to implement that at least eases the current problems...

What do you think?
I think that the player's recent performance average by mech variant is all you really need, although you could apply some correction based on that variant's global performance for the given map/mode combination and you could make some estimations based on the player's overall performance and the variant's global performance if you don't have sufficient player-speicific data points for the variant.
But in general principle, I agree.


View Postkapusta11, on 30 April 2020 - 05:56 AM, said:

Mechdad groups shouldn't count as groups at all, if anything, that's 4 1/12 chances to have someone who can pull their weight wasted. Alternatively if there's a lance of .5%ers the other 8 players should be picked from the buttom 20% unless there's a group of equal strength on the other team.
Or we could apply a similar time-gating system as Solaris, where matches start every X minutes rather than as soon as there's enough queued players, so that the matchmaker has people to work with.
Ideally the matchmaker should alternate between building both groups "from the top" (taking most skilled players / groups) and "from the bottom" (taking least skilled players / groups) so that if there's enough decent players queued they can have the competitively-minded matches they desire and if there's enough casuals / mechdads they can have the derping-around matches they desire at a reasonably equal frequency.

View PostSpare Knight, on 30 April 2020 - 06:53 AM, said:

We all know that win/loss is roughly 50/50 due to the match maker. Why on earth would "Luck of the Draw" be a factor in Player Skill Rating?

Global win/loss is always 50/50. Individual players' win/loss isn't, and the player themselves is usually the deciding factor. PSR isn't based on luck, it's been a glorified XP bar for years.

Quote

To make it more accurate, just use an Average Last 100 Match Score. Your skill rating would fluctuate a bit if you are doing Events, or skilling out a new mech, but would only be temporary and would put players into matches with others of their skill level.

But we do seem to be converging on the same conclusion: that matchmaking should in some way reflect player performance. Which metrics (WLR, PSR, KDR) should be taken into account and how granular that should be (I mean, PGI could, at least in theory, do that on a per-mech level and correct based on map, mode and build) may differ from suggestion to suggestion but the core point stays the same.

View PostSpare Knight, on 30 April 2020 - 08:45 AM, said:

Well, I guess I am wrong. Posted Image
Please help me to understand how I keep going up in Tier when my Average Match Score is 230?
At least now, I do see players of my skill level, with the tier levels thrown out the window.

Because the tier system is so biased towards increase that there are people with average match score around 170-180 who have kept going up in tiers.

Edited by Horseman, 30 April 2020 - 08:48 AM.


#571 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 30 April 2020 - 08:52 AM

View PostBrain Cancer, on 30 April 2020 - 08:43 AM, said:

It tries. If you can carry hard enough, though- it sometimes just can't put enough bad players on a team to compensate for it.

In other news, attempting to "fix" groups by reducing weight will likely result in this happening.

Posted Image

160 tons, 11/12 kills.

Weight isn't the issue- we're seeing people manage to end up with nearly all-assault matches with the currently borked MM, including matches where the better 4-man lance's team is considerably out-tonnaged. The issue is you've let organized, skilled players get together in sufficient numbers to easily skew every "solo" game they're in.


It does not try to get WLR down to 1.0 for everyone, that's just nonsense. In the old system it simply grabbed players in the appropriate tiers and balanced by weight class to make a match. In the new system it seems like it might just use the old group queue matchmaker which basically tossed a match together with whatever groups are in queue (possibly now treating solo droppers as groups of one). What you are seeing in that screenshot is a result of the skill gaps within the playerbase not some matchmaker voodoo.

#572 Larsh

    Member

  • PipPipPipPipPipPip
  • The Lanner
  • The Lanner
  • 272 posts
  • Facebook: Link
  • Twitter: Link
  • Twitch: Link
  • LocationYinz all going to EnP at PGH n'at?

Posted 30 April 2020 - 09:04 AM

View Post-Tenshi-, on 30 April 2020 - 05:41 AM, said:

If earlier the games in QP only annoyed me with endless nascar and randomly running allies, now it feels like a completely different game, people finally started trying to play as a team, which is very pleasing.


I am sincerely sorry for the nascar-potatoes whose invincible strategy has collapsed and now they are running to the forum to write that solo quick play will now die. A couple of tips for casuals and nascar-bots: review your builds, always take AMS if you can, always take consumables, do not run far ahead alone, do not leave your assaults much behind, coordinate your actions with the team, find a unit or create your own.


Dear devs, you still have to work on the balance of the selection of players for matchmaking and group tonnage restrictions. But in general, you guys did an excellent job, greatly transforming the game for the better way, I am very grateful for this.


As a casual player, my group and I HATED Nascar. I can't say for certain where the mindset of just go right, or just go left, came from, but man, did it ever ruin solo queue for most of last year and earlier. If anything, I noticed Nascar a lot less with the new system. New system still needs fixing with lopsided weight and such.

Casual doesn't always mean just go right. Having the idea of strategy, and using the terrain will always be a more enjoyable experience. Also, use the terrain doesn't mean hill hump the ridge in the middle of the map, or poke from the rock across the map.

#573 DAEDALOS513

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Flame
  • The Flame
  • 2,634 posts
  • LocationArea 52

Posted 30 April 2020 - 09:04 AM

View PostMeanMachinE, on 30 April 2020 - 02:27 AM, said:

Played last night for 4 hours and I am still loving the change. Now it is possible to play quick play again with a buddy, which I have missed. Soloing worked just fine also. Did not see any 12-0 stomps as there usually was before in QP. At least compared to the situation before the queue change, the stomping has in my experience diminished. The tactics vary now more than before, which is really refreshing. I mostly play QP to level up my mechs. FP works best for groups or some competition matches and practices.

The only thing I am a bit worried is the new player learning curve to get into the game. If it can be avoided, tier 4-5 players might need some love at the beginning of their career.

As i suggested before.. lower tiers should get an armour boost as a handicap..

Edited by DAEDALOS513, 30 April 2020 - 09:05 AM.


#574 Horseman

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 4,746 posts
  • LocationPoland

Posted 30 April 2020 - 09:11 AM

View PostKamikaze Viking, on 28 April 2020 - 05:49 AM, said:

1) Changes to the value of the thresholds of the PSR score given at the end of each match, and potentially the amount that a player goes up or down per match. This should require Zero programming (or a minor amount with Decency's suggestion below)
This is what my version would be:
Posted Image
I'd disagree simply because that still builds some bias into the system.
What I'd personally see as the best solution is to have the PSR change based not on brackets of match score but on the difference between your personal match score and the average MS of your team - and then possibly some adjustment based on the difference between your score and the average of both teams combined.
That keeps the net PSR change at a zero sum and more importantly - adapts to however the match score formula changes in the future.

View PostLarsh, on 30 April 2020 - 09:04 AM, said:

As a casual player, my group and I HATED Nascar. I can't say for certain where the mindset of just go right, or just go left, came from, but man, did it ever ruin solo queue for most of last year and earlier. If anything, I noticed Nascar a lot less with the new system. New system still needs fixing with lopsided weight and such.
Easy: Most people are right-handed and will attempt to flank targets on their dominant side. NASCAR is a result of people following flankers when they shouldn't have, but once it reaches critical mass other mechs have to choose between following the merry-go-round or dying.

View PostBenMillard, on 30 April 2020 - 08:26 AM, said:

Faction Play logic of ranking groups and placing them on opposite sides, then filling in the remaining slots with solo players to balance tonnage and ranks, actually works well. Especially when group sizes are fairly small, such as 2-4 players each. Applying that to QP, perhaps with the match countdown and visible queue length, would be a logical move given the proven success of those systems.
TBH... agreed, that is another potentially viable direction to take it.

Edited by Horseman, 30 April 2020 - 09:11 AM.


#575 justcallme A S H

    Member

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

Posted 30 April 2020 - 09:19 AM

View PostDAEDALOS513, on 30 April 2020 - 09:04 AM, said:

As i suggested before.. lower tiers should get an armour boost as a handicap..


rofl

#576 Paladin IIC

    Member

  • PipPipPipPipPip
  • The Carnivore
  • The Carnivore
  • 119 posts

Posted 30 April 2020 - 09:26 AM

Drop the tier system, bring back 8 mans and leave it. Simple solution. Drop the resources you're wasting on the empty husk that is Solaris and put them toward quick play. You could also [redacted] upgrade your server hardware.

Edited by GM Patience, 30 April 2020 - 10:40 AM.


#577 Thorqemada

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 6,395 posts

Posted 30 April 2020 - 09:34 AM

View PostZirconium Kaze, on 29 April 2020 - 08:00 PM, said:

This is the truth.


Afair FW had respwan an mech pre selection so it is not true.

#578 o0cipher0o

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 353 posts
  • LocationItaly

Posted 30 April 2020 - 09:40 AM

So, PGI, where the heck did weight balance go?
Seeing way too many matches with teams having wildly different total weights.
Make the damn matchmaker balance the weight of the solo players at the very least.

#579 R Valentine

    Member

  • PipPipPipPipPipPipPipPip
  • Heavy Lifter
  • Heavy Lifter
  • 1,744 posts

Posted 30 April 2020 - 09:43 AM

View PostRickySpanish, on 30 April 2020 - 05:38 AM, said:

So here is a thought - the Good Players don't enjoy stomps in the new SQ/GQ merger, so why play it? I call BS on that. Maybe you guys think you don't enjoy it, but something's making you play, eh? And of course, regular solo queue was absolutely the pinnacle of competitive play beforehand, right? So the only thing I can think of that you are missing is the old segregated group queue which correct me if I'm wrong, was rather dead already.


I love how the "GROUP QUEUE DEAD!" crowd conveniently forgets that even the very brief 8v8 group queue was completely buzzing, even among players who weren't of the "cult of competitive". Players of all skill ranges largely reported satisfaction with the 8v8 group queue. But somehow, "group queue dead!" is still your main argument even when that argument has been shot to pieces. The merging of solo and group queue has only removed features from the game. It hasn't added anything. In the week before the queues were merged, I could already drop as a two man, three man, or four man group and reliably get matches. Even 5 and 6 man teams saw reliable queue pops. I could also drop solo and be premade free, should I feel the urge. Now I can't. Premade free queues are no longer a thing, but somehow you think the loss of a feature enjoyed by the majority of players is a good thing.

#580 DAEDALOS513

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Flame
  • The Flame
  • 2,634 posts
  • LocationArea 52

Posted 30 April 2020 - 09:45 AM

Still having a blast in new grolo queue.. if you are a particularly skilled piloy try leaving your comfort zone and challenge yourself.. take out 'non-meta' for a spin..

I know this will be difficult for some bec God forbid you under-perform in the eyes of the masses.. but remember, we all know who you are, you have absolutely nothing to prove to us.. ego makes one angry and happiness elusive. Don't let it dominate your life..

"When ego is lost, limit is lost. You become infinite, kind, beautiful." Harbhajan Sing Yogi





6 user(s) are reading this topic

0 members, 6 guests, 0 anonymous users