344 replies to this topic

[Alan Hicks]

Unbalanced teams is the most common fact in QP.

The incredibly flawed mm is the main reason this game's community refused to grow. It is also the death sentence to MWO.

Disastrous experiences no matter your effort in a match is what drives people far, far away from considering spending their time piloting a mech.

[Variant1]

Bots would help population, but pgi dont do ai

[BumbaCLot]

Variant1, on 17 October 2018 - 08:18 AM, said:

thank you OP weve been tellin others that 12-0 means mm don goofed not both sides were equal thats why it lead to a stomp

mm needs to balance by loadouts and stats. Have both teams have meta build and even stats.
Infact id prefer if mm just paired meta cookie cutter builds against each other, that way us filthy casuals with non meta bracked builds could enjoy slugging it out while not being torn to shreds by high alpha strikers. just my 2 centurions

Go ahead and report me for harassment since that's what you said you intend to do anyway, but you are the one checking out my user profile. You don't have enough recorded games for me to take anything you say seriously, but when you are taking LRM jenners in to battle that's not bracket, that's just bad.
And I just played a T5 alt this weekend (for two matches, one I had a broken keyboard and ended up spinning in a circle and getting 37 damage and another I got 950 damage in a trial dire wolf).
There are plenty of non-meta "alpha strikers" in the low divisions. If anyone is playing meta builds they will end up in T2 within a 1-2000 matches with horrible stats.
So again, if you are running awful 'fun' builds and your WLR is under 1, you are to blame, not the matchmaker, given enough matches played.

[BumbaCLot]

Tiewolf, on 16 October 2018 - 11:25 PM, said:

I think we can all agree that in 70% of the time we can predict the outcome of a game by looking at player/unit/cmd signatures. That w/l ratio always pushes towards the mean (1) is not pgi or the coding of the MM but the mathematical phenomenon of regression to the mean. That is a strong indicator that MM is purely random.

Personally I wonder more if MM considers secondary variables like premium time? Would be smart of pgi to give paying costumers advantages without the unpleasant p2w feeling.

Statistically speaking the biggest decider in outcome is AFK/disconnect/easy first kills as the snowball effect is more powerful than anything. Less guns shooting/less targets to shoot at.
Same goes for out of position teams where more than one are nowhere near the battle.
So don't chase lights and don't follow your own lights either!

[MrXanthios]

Tiewolf, on 16 October 2018 - 11:25 PM, said:

I think we can all agree that in 70% of the time we can predict the outcome of a game by looking at player/unit/cmd signatures. That w/l ratio always pushes towards the mean (1) is not pgi or the coding of the MM but the mathematical phenomenon of regression to the mean. That is a strong indicator that MM is purely random.

Personally I wonder more if MM considers secondary variables like premium time? Would be smart of pgi to give paying costumers advantages without the unpleasant p2w feeling.

I wouldn't be surprised if premium players' mm is actually rigged. And I wouldn't be surprised if mm was also rigged for some streamers, especially one.

[HARDKOR]

Appogee, on 19 April 2017 - 09:27 AM, said:

I know it's weird, but it feels to me like win:loss ratio is still being used in the background to allocate people to teams. I say that because my win:loss ratio cycles up to 1.1 then down to 1.01. Up to the top of that range, then down to the bottom again. And it's been doing that for two years. It's as if when I get to 1.1 some kind of switch kicks in and I get a run of awful teams that not even Proton could carry. I have no proof for this, other than that my win:loss stat seems like it hasn't behaved in a random fashion over the course of thousands of matches.

Can confirm... I hover around 1 no matter if im playing terribly or playing well.I would assume that if someone has a high ratio, it's because they only drop with their comp group, because the matchmaker will always even you out.

[Tarl Cabot]

HARDKOR, on 17 October 2018 - 02:03 PM, said:

Can confirm... I hover around 1 no matter if im playing terribly or playing well.I would assume that if someone has a high ratio, it's because they only drop with their comp group, because the matchmaker will always even you out.

That is to be expected, up to a point. The times where in SQ of having a higher W/L rate if the player enhances the team's gameplay like a Bard and said member provides tactics and the team follows. This is then enhanced even more when the team is made up of like minded individuals who are more used to playing as a team and taking directions, ie faction play and group queue, than the other tryhards/rambos who may treat everyone else as if they were NPC with no control.

That can NOT be measured by the MM. And the PSR that is used, ie is more of a Player Experience Rating than Player Skill Rating. Everyone may reach different plateaus with how they perform, individually, but that performance can be enhanced in a group controlled setting but with many SQ drops there is no real control.

So any research done will show that majority of the time the side with the better stats, be it a T1 or T2 or T3 drop, will generally win because the MM is only using Tier designation, matching weight class to weight class, then opening up when appropriate tier is not available.

Did anyone really expect anything different to happen? Elo worked in similar fashion. Seeded with a player then picked people up around his Elo number, above and below, but stricter on the weight classes. The difference between Elo and PSR? Elo did not have a static threshold to be reached. If the team designed to be better team won and the team designed to lose lost, Elo changed little, dependent on the score difference. But if the team designed the underdog actually won, the scores earned were greater, again dependent on the score difference.

That is my view on it. Like my main -T1 vs alt- T3 (w/120 SQ games over a few seasons) W/L is 1.1 vs 1.5, avg 245 MS vs 320 MS. And Alt used freebies/trial mechs at first - Wolverine 6k, Battlemaster 3ERLL/3ERML w/SHS-lots!!. , Enforcer and Centurion (freebies). Hmm, Wolfhound and Roughneck (free one) and Warhammer 6R. Beside those stats, on the Jarl's list my alt is ranked lots higher than my main. hai.. like a wolf half-breed in amongst the sheep/goats.. never knowing what ya going to get...

Again, each of us hits a plateau, individually, dependent on who we are facing but if that plateau is at or slightly exceeds the PSR static threshold, eventually most of us will be Tier 1, a WIDE range of players. Should MM use additional stats or different stats other than PSR? Just remember, that Match Score is not simply what you do on the battlefield, but it is also a reflection of what you do in the mechlab too. There are so many variables...

Frak it.. beer run!!!!

[MischiefSC]

HARDKOR, on 17 October 2018 - 02:03 PM, said:

Can confirm... I hover around 1 no matter if im playing terribly or playing well.I would assume that if someone has a high ratio, it's because they only drop with their comp group, because the matchmaker will always even you out.

Nope. I have a lower w/l in group queue than solo queue and I'm well over a 1.0 and I'm upper end of mediocre. There are thousands of players with hundreds or thousands of matches with a w/l over 1.25 and hundreds with that many matches and a w/l over 2.0. I spent one month only playing QP and actually putting in my best effort and got a 2.92 and nobody is going to call me a top tier player.

Math is actually a thing and it works. You're 8.333% of your teams performance every time. Everyone else is playing in the same pool of randoms - your teammates 'wash out' vs everyone else in the average over time. What's left is your performance. A sample size over 60 is going to get you reasonably close, to place within a few percentage points of accuracy you need a couple of hundred matches to a representation of how likely you are to win matches with w/l as an indicator.

The only people who win constantly in group queue are the same people who have the skills to win consistently in QP. While their w/l is likely inflated in group queue vs solo queue it's going to be scaled relative to the environment - while mixing solo/group queue absolutely skews results it skews them largely by scale, not relevance. Someone who wins a lot in one will win a lot in the other. Only a few dozen players are good enough to have significant inflation from group queue, vs thousands and thousands of players on the board.

[N0MAD]

MischiefSC, on 17 October 2018 - 06:35 PM, said:

The only people who win constantly in group queue are the same people who have the skills to win consistently in QP.

Not so, most people who win constantly in group q win because generally they will play only group q with a team usually vs a pug team, there are a couple teams that will win constantly vs other organised teams but generally the people that only play with an organised group vs pug groups are stat padding.

[Full Meta Jacket]

9 losses in a row last night before I logged off and went and did something actually fun.

In one of my alleged "Tier 1" skirmish matches, five players did less than 100 damage.

It wasn't the snowball effect, either. These players stood in a clump, barely even firing, refusing to move ... even though I kept indicating where they should move to, highlighting the Mechs moving around to flank them, etc.

Noone can carry players this poor. So my question is: why am I being dropped with them so frequently?

Edited by Full Meta Jacket, 17 October 2018 - 07:26 PM.

[John Bronco]

I agree the mm is completely random.

For some more fun anecdata I played a match the other day with SlippnGriff and Magic Pain Glove on the same team vs noone else I recognized (and I play at the same time most days, I know who's bringing a big stick). Both about as ruthless as you will find in solo....why are they on the same team?

[RickySpanish]

I dunno guys, I feel like it's working as intended. I played some pretty meta builds on maps that supported them the other week and my w/l went above 2, recently I have been playing my Awesomes and Fun-Spider VG and it's tanked. Seems like if you shoot yourself in both feet at the start of the race you aren't going to cross the finishing line. Even had one guy remark about how every drop with me that evening had been a total loss. Sorry bud

Edited by RickySpanish, 17 October 2018 - 08:34 PM.

[Peaklander]

It feels like I've lost the last 10 games in a row now. I've just hit tier 2 and I whilst I have been knackered from work and playing very poorly I shouldn't drag down the entire team to lose the game this many times in a row. Anyway I read this thread and I was immediately blaming matchmaking of course

I did notice that not a single person talks in any of these games. No mic or text - not a sausage. Normally I can at least get a hello out of people. Stupid bot replacement conspiracy initiated!

Edited by Dilbert Prime, 18 October 2018 - 01:52 AM.

[Snowhawk]

Full Meta Jacket, on 17 October 2018 - 07:16 PM, said:

9 losses in a row last night …. five players did less than 100 damage.
Noone can carry players this poor. So my question is: why am I being dropped with them so frequently?

Same here…. I often see assaults with less than 50 damage and I'm bored about it. Seems that Pgi has opended the "Flood gates" again and Tier 4 and 5 players are mixed into Tier 1 matches.

[Tiewolf]

BumbaCLot, on 17 October 2018 - 11:20 AM, said:

Statistically speaking the biggest decider in outcome is AFK/disconnect/easy first kills as the snowball effect is more powerful than anything. Less guns shooting/less targets to shoot at.
Same goes for out of position teams where more than one are nowhere near the battle.
So don't chase lights and don't follow your own lights either!

Yes thats true for one specific game but it doesn`t matter if all do that or all don`t because over 1k games randomization evens it out as long as MM doesn`t intentionaly create weaker teams.

I am more interested in the variance of the data. Do you get a normal destribution around the artifical w/l mean 1 or is the destribution skewed?

Just for fun I have to disappoint all "a sample of 12 is not enough" preachers. If the power of the effect is very strong you don`t need hugh samples to prove it. But MWO is very complex and has hundreds of covariables so the sample sizes must be very large to prove anything.

[Tiewolf]

HARDKOR, on 17 October 2018 - 02:03 PM, said:

Can confirm... I hover around 1 no matter if im playing terribly or playing well.I would assume that if someone has a high ratio, it's because they only drop with their comp group, because the matchmaker will always even you out.

Yes and No if the MM is realy random.

Yes you have to make sure you cheat the MM in group queue by stacking more skilled players/better loadouts/teamwork on your side to achieve a hugh W/L or K/D ratio. Some argue that W/L ratio goes up and K/D ratio goes down when you team up with better players. But thats not the case because you might get less kills but your chances of surviving till the end of a game is much higher.

No you can drop exclusively in solo QP and have a constant W/L ratio above or lower then 1. First you have to play a lot of games. If not then your W/L ratio bounces around 1 randomly. After a lot of games the difference between the precentage above or lower then 1 is your personal influence in % on the outcome off all the games you participated in. Good dropcallers, solely meta users or extreme skilled players should have a constant higher % W/L ratio in solo QP then others. If not...well...

Edited by Tiewolf, 18 October 2018 - 03:22 AM.

[RickySpanish]

Dilbert Prime, on 18 October 2018 - 01:50 AM, said:

It feels like I've lost the last 10 games in a row now. I've just hit tier 2 and I whilst I have been knackered from work and playing very poorly I shouldn't drag down the entire team to lose the game this many times in a row. Anyway I read this thread and I was immediately blaming matchmaking of course

I did notice that not a single person talks in any of these games. No mic or text - not a sausage. Normally I can at least get a hello out of people. Stupid bot replacement conspiracy initiated!

Not talking is actually something I have noticed too. T3 is the sweet spot for that, T2 is extremely quiet until you lose... I imagine by T1 everyone twists their knickers as a matter of course before the map selection is complete...

[Obiwayne]

I'd go so far to say that every competitive multiplayer team game tries to keep players at a ~50% W/L ratio. It's a pretty common thing which is called elo hell.

It would be much more accurate and satisfactory if all of the matches were balanced in terms of player skill.

If you want to get above 50% you dont only have to play good, you have to play godlike.

Edited by EdSt0nE, 18 October 2018 - 06:40 AM.

[RickySpanish]

EdSt0nE, on 18 October 2018 - 06:38 AM, said:

I'd go so far to say that every competitive multiplayer team game tries to keep players at a ~50% W/L ratio. It's a pretty common thing which is called elo hell.

It would be much more accurate and satisfactory if all of the matches were balanced in terms of player skill.

If you want to get above 50% you dont only have to play good, you have to play godlike.

Elo hell does not exist, or if it does, then it exists at all levels of play. Put another way, there is no level of play at which one does not accuse one's team mates of throwing a game.

[BumbaCLot]

Dilbert Prime, on 18 October 2018 - 01:50 AM, said:

It feels like I've lost the last 10 games in a row now. I've just hit tier 2 and I whilst I have been knackered from work and playing very poorly I shouldn't drag down the entire team to lose the game this many times in a row. Anyway I read this thread and I was immediately blaming matchmaking of course

I did notice that not a single person talks in any of these games. No mic or text - not a sausage. Normally I can at least get a hello out of people. Stupid bot replacement conspiracy initiated!

It happens. Even Juju Shinobi lost for 2.5 hours on his stream a week ago, but his WLR only dropped to 2.6.