Jump to content

Mm Filters & Psr


37 replies to this topic

#21 DerHuhnTeufel

    Member

  • PipPipPipPipPip
  • 195 posts

Posted 19 March 2018 - 01:01 AM

View PostKrivvan, on 18 March 2018 - 11:18 PM, said:

My opinion may also be impacted by the fact that, usually playing a Light, I rely quite a bit on how solid the core of the team is. I can't be the one to initiate a push for example, or take a position that leads to the rest of the team repositioning.


As a light, you should be enabling a push because you should be distracting and/or providing targets. If all you're doing is waiting to take things out when they're distracted with others you aren't making good use of your maneuverability and speed.

#22 arcana75

    Member

  • PipPipPipPipPipPipPipPip
  • Shredder
  • Shredder
  • 1,161 posts

Posted 19 March 2018 - 01:07 AM

View PostEscef, on 18 March 2018 - 05:13 PM, said:

To be honest, I think the worst potatoes in solo-QP are people that rode into tiers 1 and 2 on the backs of coordinated groups. A little coordination goes a long way into covering up people's shortcomings in gunnery or mechbuilding. And it's easier to be aggressive when you know you have 3 or 4 friends at your side. But these guys in solo-QP? Oh, gawds... Atrocious builds, horrible awareness, no idea how to use their builds, and about as aggressive as a quadriplegic sloth.

I think I agree with you, there will be some players who failed to learn things the hard way, too much mollycoddling. But it won't speak for every player in T1.

View PostNightbird, on 18 March 2018 - 05:45 PM, said:


Stomps happen because of unbalanced teams, repeated stomps reduce population.

Stomps in MWO happen because of the nature of MWO's battles of unlimited weaponry vs limited armour in a coordinated fashion. If stomps didn't happen, it's because that match resulted in several disconnected battles across the map, rather than a single large engagement between 2 teams. If it's the latter once there's a slight numerical advantage, if the team seizes that advantage and pushes, that small numerical advantage snowballs quickly into a stomp. It nearly always never matters then who's on the opposing team.

#23 The Basilisk

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 3,270 posts
  • LocationFrankfurt a.M.

Posted 19 March 2018 - 01:29 AM

View PostBlueStrat, on 18 March 2018 - 04:06 PM, said:

The PSR system in MWO has many problems, one of which is the low population's effect on the matchmaker algorithms for solo-PUG and how that affects PSR calculations based on the baked-in assumptions the PSR algorithm makes about players.

When the que has fewer players it starts to drop filters (Tier, weight class, etc). When MM starts putting almost everyone into one 'bucket' that invalidates all of the player's PSR scores that match as PSR is calculated with the assumption that players won't be matched against others who have significantly different levels of skill & experience.


Every player skill based matchmaker in a game like MWO is a lie, especially in an environment like MWOs Quickplay.

MM can not:

-- take account for a player prefering a playstyle that may or may not fit to his assigned group

-- take account for the build the player is bringing (trollbuild much?)

-- take into account that the player may be well versed in one weight class but averange at best in an other

-- take into account the combination of fielded builds and the environment the own team and enemy team is put and if this environment favors a certain combo of builds or not

To make this short and simple:
As long as a match in MWO is 2/3 just dumb luck and complete randomness in maps and teams the whole MM thingy is just a poor excuse for not doing a more sophisticated combat value driven system that takes into acount the competitveness of a chassis in relation to map and fielded loadout, combined with the player modificator.

Sure this will make some loadouts on both ends of the scala having LOOOOOONG waiting times.
And this will lead us to a starvation of variety since the more extreme builds on both ends will get more seldom to be seen in games.

And there you go the endless pabulum is ready for you to get eaten over and over again.

#24 Krivvan

    Member

  • PipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 4,318 posts
  • LocationUSA/Canada

Posted 19 March 2018 - 01:56 AM

View PostDerHuhnTeufel, on 19 March 2018 - 01:01 AM, said:

As a light, you should be enabling a push because you should be distracting and/or providing targets. If all you're doing is waiting to take things out when they're distracted with others you aren't making good use of your maneuverability and speed.

You distract things by hitting them when they're paying attention to your team. That's Light 101. You don't poke your head out multiple times as the only mech visible to the enemy unless you want to instantly die. You can't just run into an enemy team alone and hope they miss all their shots on you. That only works when up against the worst of players.

Nor does distracting an enemy or "providing targets" actually mean your team is actually going to push. What it seems you're suggesting is the stereotypical "why didn't the rest of the team push, we were flanking/distracting them" mistake you often see.

Edited by Krivvan, 19 March 2018 - 01:59 AM.


#25 El Bandito

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • Big Daddy
  • Big Daddy
  • 26,736 posts
  • LocationStill doing ungodly amount of damage, but with more accuracy.

Posted 19 March 2018 - 02:30 AM

View PostKrivvan, on 19 March 2018 - 01:56 AM, said:

Nor does distracting an enemy or "providing targets" actually mean your team is actually going to push. What it seems you're suggesting is the stereotypical "why didn't the rest of the team push, we were flanking/distracting them" mistake you often see.


Yep, pugs are not trustworthy in general.

#26 Asym

    Member

  • PipPipPipPipPipPipPipPipPip
  • Nova Captain
  • 2,186 posts

Posted 19 March 2018 - 02:52 AM

Oh good grief.... A little reality: stomps happen in real life using real mechs everyday of every year.... It's not a week that goes by at the any nation's armored training centers where the "good guys" get eaten alive by the OPFOR... Everyday. Day in, Day out. Been there, done that and it's normal.......... why? Because the OPFOR cheats.......they know every square inch of the terrain. They know the enemy. They have vastly superior time-on-the-ground teams who have been together for months if not years on the same terrain.... It's a rare bird that defeats the OPFOR and when we find those guys, we pay attention to who is leading them... In real life, teams are life and good leaders, the real eaters-of-flesh, are rare..... 5 out of 100. Same here.

I do not agree with a lot El Bandito of what says of late, teams in MWO are just about gone and Solaris, if it works, will be the end of them.

OP, stomps are normal and if you find a 10-12 match, cherish it, because you've been involved is a place and time where luck, skill and fate has conspired to make the game FUN>...............

#27 Krivvan

    Member

  • PipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 4,318 posts
  • LocationUSA/Canada

Posted 19 March 2018 - 04:17 AM

View PostEl Bandito, on 19 March 2018 - 02:30 AM, said:


Yep, pugs are not trustworthy in general.

It's even in the case with trustworthy teammates. You don't want Lights to be the first and only target the enemy team shoots at. You want them to be the second target when they're already busy with the bulk of the team and cannot afford to spend the relatively little amount of firepower it takes to kill you.

#28 Nightbird

    Member

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

Posted 19 March 2018 - 05:49 AM

Bottom line is still that PGI's attempt to balance teams with the Tier system is flawed as Tiers are based on matchscore and how match score is calculated is entirely disconnected from a player's crontribution to a team's win.

#29 Electron Junkie

    Member

  • PipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 192 posts

Posted 19 March 2018 - 05:57 AM

Considering stomps are 9 out of 10 matches, at this point I would just be happy with match maker taking less than 5 minutes to put an actual match together. I would rather play than sit at the searching screen.

Why not just turn match makers brains off and throw everyone minus the new in the last 30days account in one big bucket.

#30 ZippySpeedMonkey

    Member

  • PipPipPipPipPipPip
  • Survivor
  • 356 posts
  • LocationSomewhere on Dropship Earth

Posted 19 March 2018 - 08:51 AM

Stomps / Losses happen because you are all predictable and have little to no situational awareness.

In most games I play, you all, predictably NASCAR to the right. Even as you know they are coming from the left you’re still all dead set on going right and not setting up a line to stop them.

To make matters worse you then all try to keep up with the Lights is a pathetic game of follow the leader. This then allows their Lights ( coming from the left ) to kill the Assaults.

To compound this even as team mates are dying in the rear none of you will turn around to try and stop them even though the map in the middle of you screen has the red triangles on it showing you where they are.

All in all MM is not the problem. It’s your own habits and instincts which make you predictable that lead to the loss.

P.S. As most of you know the only predictable thing about me is that I will run away from the group at the start of the match and generally work alone.

#31 BlueStrat

    Member

  • PipPipPipPipPipPip
  • 241 posts

Posted 19 March 2018 - 09:53 AM

View PostElectron Junkie, on 19 March 2018 - 05:57 AM, said:

Considering stomps are 9 out of 10 matches, at this point I would just be happy with match maker taking less than 5 minutes to put an actual match together. I would rather play than sit at the searching screen.

Why not just turn match makers brains off and throw everyone minus the new in the last 30days account in one big bucket.


The new players would have to be thrown in as well, as there simply is not a large enough player population to have separate ques.

I understand that no skill-filtering system can work without a sufficient pool of players to pull from.

Many of the problems in MWO boil down to essentially not having a large enough player population because PGI lacks the money for proper marketing because player population is low. "It's turtles all the way down" in essence.

Mechwarrior *is* a niche game, always has been. Previous iterations benefited from being backed by companies like Microsoft that had the resources to market the game well and so attract a larger player base.

I just wanted to point out another facet adding to the dumpster-fire we all know as the PSR/Tier system.

#32 Duncan Aravain

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 416 posts
  • LocationBehind you with a sharp tool...er,mech

Posted 19 March 2018 - 01:53 PM

View PostZippySpeedMonkey, on 19 March 2018 - 08:51 AM, said:

Stomps / Losses happen because you are all predictable and have little to no situational awareness.

In most games I play, you all, predictably NASCAR to the right. Even as you know they are coming from the left you’re still all dead set on going right and not setting up a line to stop them.

To make matters worse you then all try to keep up with the Lights is a pathetic game of follow the leader. This then allows their Lights ( coming from the left ) to kill the Assaults.

To compound this even as team mates are dying in the rear none of you will turn around to try and stop them even though the map in the middle of you screen has the red triangles on it showing you where they are.

All in all MM is not the problem. It’s your own habits and instincts which make you predictable that lead to the loss.

P.S. As most of you know the only predictable thing about me is that I will run away from the group at the start of the match and generally work alone.


A bit harsh to say "all of you" and " most of you", but would agree with saying "many of you" and I understand your sentiment.
There are times I feel like Kevin Bacon in the parade scene in "Animal House". In addition, it still doesn't explain the mm having the tendency to place elite players unequally, an observation again noticed from this last weekend of heavy MWO playing. When you can start determining the match results before it starts because one side has 2 or more "elite" players more than the other (sometimes one side has ZERO "elite" players), you can start to understand a certain number of DCs and groans in TS as in the know players recognize what is about to happen. The mm algorithm needs to be changed; perhaps in increasing the importance of k/d ratio.

Edited by Duncan Aravain, 19 March 2018 - 02:31 PM.


#33 ShaneoftheDead

    Member

  • PipPipPipPipPip
  • 174 posts
  • LocationPA

Posted 19 March 2018 - 02:17 PM

Balancing matches is very complicated. PSR alone, even if it were perfect, would still turn out poorly balanced teams. Your piloting skill is only part of the equation in any given match. 'Mech load-outs, Map, and Mode are large factors as well.

PSR is flawed, mainly because it puts too much emphasis on Wins. You and your 'Mech have somewhere around 4% influence on the outcome of any given match. The other 96%, the vast majority of the influence, are the other people in that match. The calculations for it needs to totally ignore Wins and Losses.

'Mech capability, from what I understand, is largely ignored by the PSR. They are trying to address this in Solaris 7 by giving 'Mech's a Tier Level too. While this is a good start, it is too simplistic because it ignores the load-out of that 'Mech. 'Mech tonnage disparities is also a part of this.

The Map has a large influence on how you perform. Like short range weapons on Polar Highlands or hot weapons on Terra Therma. From what I understand, MM ignores this.

Game Mode has some influence, not as much as the others, but still more than historical Wins. Ever lose a Domination on Polar Highlands because no one on your team could make it into the circle in time because the enemy had a 165 kph Locust and the best your team had was an 70 kph medium? Or get out-capped by fast lights in Conquest?

Groups can make balancing impossible. A group of good players, if large enough, can be better than any combination of the available players within the Tier Levels. That is to say, you take an 8-man group in the bucket and pair them with the 4 worst players in the bucket then have them play against the best 12 remaining players in the bucket and the side with the 8-man team rolls them.

I suggest that we throw out the current PSR and replace it with a system that judges your performance separately for each 'Mech and each load-out you use. The calculation could factor in KD Ratio, Firepower rating, Heat Dissipation, Ambient Temperature, Speed, Game Mode. The point is: You use it to assign 1 number to that player in that 'Mech on that Map and that Mode. Then you keep a tally as you build the teams and you try to balance the sides using those calculations.

As I have written elsewhere, there are various solutions PGI could try. However, it all hinges on being able to determine if the sides are balanced much more accurately than they do now in MM using PSR and Tiers.</p>

1) Unbalance the numbers of players. If you cannot make a balanced 12 vs. 12 match, then don't force a 12 vs. 12 match. Have MM build the teams piecemeal, but when you add a player or group that throws the balance off just add to the other side until things are balanced again or you hit your 12 player team limit. 8 vs. 12 could be a balanced match if those 8 are great pilots in capable 'Mechs and the 12 player team are average pilots.

2) Give the team with a lower rating some in game advantages. We'll never get perfect balance every match because you'll never know what is in the bucket until you look. So if you make a match that is poorly balanced, give the weaker team an advantage(s). For example, you could give the weaker team Combat ID or Radar Sweep. Or you could take away Air Strikes and/or Artilley from the stronger team. Or some combination of these things depending on the disparity.

EDITED to fix formatting.

Edited by ShaneoftheDead, 19 March 2018 - 02:19 PM.


#34 pjnt

    Member

  • PipPipPipPipPipPip
  • Philanthropist
  • 425 posts
  • LocationDetention Block 23, Death Star

Posted 20 March 2018 - 10:18 AM

View PostEscef, on 18 March 2018 - 06:11 PM, said:


On the contrary, balanced teams make stomps more likely. In teams were everyone is equally valuable, losing 1 player is bad, and can end up snowballing into a stomp. In teams where not every player is equal, sure, losing your worst player doesn't hurt much, but losing your best is even worse than losing a player on the everyone is equal team.

Stomps are the result of one team making a mistake and the other exploiting it. Close games, however, are the result of BOTH teams making a lot of mistakes.

As for repeated stomps reducing the population, it can disenfranchise some people, absolutely. But balancing teams does not eliminate stomps, it just makes it harder to recover from a mistake.

In short, stomps are normal.


Oh my.

Not certain I agree with this. The game is incredibly fluid and battle awareness is king. I don't believe balance has much to do about it in this context, but skill does. 2 teams of equal skill will have players that are able to both fix mistakes by their side and capitalize on the other teams lesser skilled players errors. Unbalance the teams and the side that has less skill gets stomped should they make mistakes or not. 12 crap players playing a great game will get stomped by 12 great players playing a great game - period.

Distribute the skill equally and let the new players make their mistakes and learn while the skilled ones do their jobs.

View PostShaneoftheDead, on 19 March 2018 - 02:17 PM, said:

Balancing matches is very complicated. PSR alone, even if it were perfect, would still turn out poorly balanced teams. Your piloting skill is only part of the equation in any given match. 'Mech load-outs, Map, and Mode are large factors as well.

PSR is flawed, mainly because it puts too much emphasis on Wins. You and your 'Mech have somewhere around 4% influence on the outcome of any given match. The other 96%, the vast majority of the influence, are the other people in that match. The calculations for it needs to totally ignore Wins and Losses.

'Mech capability, from what I understand, is largely ignored by the PSR. They are trying to address this in Solaris 7 by giving 'Mech's a Tier Level too. While this is a good start, it is too simplistic because it ignores the load-out of that 'Mech. 'Mech tonnage disparities is also a part of this.

The Map has a large influence on how you perform. Like short range weapons on Polar Highlands or hot weapons on Terra Therma. From what I understand, MM ignores this.

Game Mode has some influence, not as much as the others, but still more than historical Wins. Ever lose a Domination on Polar Highlands because no one on your team could make it into the circle in time because the enemy had a 165 kph Locust and the best your team had was an 70 kph medium? Or get out-capped by fast lights in Conquest?

Groups can make balancing impossible. A group of good players, if large enough, can be better than any combination of the available players within the Tier Levels. That is to say, you take an 8-man group in the bucket and pair them with the 4 worst players in the bucket then have them play against the best 12 remaining players in the bucket and the side with the 8-man team rolls them.

I suggest that we throw out the current PSR and replace it with a system that judges your performance separately for each 'Mech and each load-out you use. The calculation could factor in KD Ratio, Firepower rating, Heat Dissipation, Ambient Temperature, Speed, Game Mode. The point is: You use it to assign 1 number to that player in that 'Mech on that Map and that Mode. Then you keep a tally as you build the teams and you try to balance the sides using those calculations.

As I have written elsewhere, there are various solutions PGI could try. However, it all hinges on being able to determine if the sides are balanced much more accurately than they do now in MM using PSR and Tiers.</p>

1) Unbalance the numbers of players. If you cannot make a balanced 12 vs. 12 match, then don't force a 12 vs. 12 match. Have MM build the teams piecemeal, but when you add a player or group that throws the balance off just add to the other side until things are balanced again or you hit your 12 player team limit. 8 vs. 12 could be a balanced match if those 8 are great pilots in capable 'Mechs and the 12 player team are average pilots.

2) Give the team with a lower rating some in game advantages. We'll never get perfect balance every match because you'll never know what is in the bucket until you look. So if you make a match that is poorly balanced, give the weaker team an advantage(s). For example, you could give the weaker team Combat ID or Radar Sweep. Or you could take away Air Strikes and/or Artilley from the stronger team. Or some combination of these things depending on the disparity.

EDITED to fix formatting.


Lots of good stuff here. ID the KPI for each pilot and use that as the players ranking.... I mean they do an OK job of it determine how many credits you get... don't they?

I do not agree with adding things into the game for a lesser team or creating uneven numbers of players on each side. There are 2 things at work here, the visible balance and the invisible one. The visible one is the number of players and the tonnage they have - this needs to be as close as possible. The invisible side is where organising those players into a equal matchup. It is here you tweak. Otherwise the can of worms opened up would be epic. The torches would be lit, pitchforks brought out and children would hide under their beds. Keep the visible completely balanced, only half the headache on the forums.

Edited by xlentx, 20 March 2018 - 10:24 AM.


#35 Dragonporn

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 657 posts

Posted 07 April 2018 - 10:19 PM

It gets really annoying sometimes... Tier system in this game isn't really bad, but there's only one significant downside: system built in mind, so after terrible performance, for player not to feel too bad, it doesn't give heavy enough penalties, which makes every single scrub end up in tier 1 sooner or later. Fixing this little thing would make whole system simply perfect. Sadly, system built like this in many other different competitive games, with exactly same issues.

Not all belong in T1, and even T2 or T3. After losing match or having poor score, you shouldn't get ^ in rating unless you made some significant impact. Winning match even if you have crap score guarantees ^. That's wrong. PSR should be accounted only per score, and winning or losing match should have much less impact. Of course Lights would be rather screwed with this system, so it needs additional tweaks to fix this crap, but overall, it feels weird to meet players in higher tiers who don't know basics. Yeah, it doesn't happen on regular basis, but because of that, it is even more obvious and sad.

*Venting mode off*

#36 Dogstar

    Member

  • PipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 1,725 posts
  • Twitch: Link
  • LocationLondon

Posted 08 April 2018 - 10:39 AM

View PostNightbird, on 18 March 2018 - 05:45 PM, said:


Stomps happen because of unbalanced teams


Utterly, utterly wrong. Stomps are a normal product of the game design, unbalanced teams just exacerbate it. If you don't understand this you really shouldn't be talking about it.

#37 Nightbird

    Member

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

Posted 08 April 2018 - 10:49 AM

View PostDogstar, on 08 April 2018 - 10:39 AM, said:


Utterly, utterly wrong. Stomps are a normal product of the game design, unbalanced teams just exacerbate it. If you don't understand this you really shouldn't be talking about it.


LOL

That's like saying, Death is a part of life, getting shot in the head only exacerbates it!!

LMAO

Edited by Nightbird, 08 April 2018 - 10:59 AM.


#38 Bulletsponge0

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Vicious
  • The Vicious
  • 2,950 posts

Posted 08 April 2018 - 02:04 PM

View PostEscef, on 18 March 2018 - 06:11 PM, said:


On the contrary, balanced teams make stomps more likely. In teams were everyone is equally valuable, losing 1 player is bad, and can end up snowballing into a stomp. In teams where not every player is equal, sure, losing your worst player doesn't hurt much, but losing your best is even worse than losing a player on the everyone is equal team.

Stomps are the result of one team making a mistake and the other exploiting it. Close games, however, are the result of BOTH teams making a lot of mistakes.

As for repeated stomps reducing the population, it can disenfranchise some people, absolutely. But balancing teams does not eliminate stomps, it just makes it harder to recover from a mistake.

In short, stomps are normal.

that only applies if both teams are evenly balanced, and both teams are of relatively high skill where focus fire is more common and natural.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users