Jump to content

Get Rid Of Elo-Based Matchmaking


69 replies to this topic

#41 EgoSlayer

    Member

  • PipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 1,909 posts
  • Location[REDACTED]

Posted 16 August 2014 - 12:05 PM

View PostAlistair Winter, on 16 August 2014 - 10:48 AM, said:

MM is supposed to give you a 50% chance of winning by balancing ELO scores on both teams.

**** that.

Just let veterans play with veterans and new players play with new players. If a new player wants to play with the veteran players, he can join the group queue.

I've played thousands of matches. I don't want to play with people who don't know how minimap coordinates work.


You obviously don't realize that a match maker giving you an equal chance of winning means that you are matched against equal opponents....

The problem isn't entirely the match maker. It's the player population and weight class distribution and the player skills across those weight classes that are the key factors. If you want to wait 30 minutes for a match you could get perfect matches every time. Most, myself included, don't want to wait that long.

#42 EgoSlayer

    Member

  • PipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 1,909 posts
  • Location[REDACTED]

Posted 16 August 2014 - 12:12 PM

View PostMiken, on 16 August 2014 - 11:13 AM, said:

We need BattleValue based balance. Not Elo or silly 4x3...


BV isn't the answer because it doesn't account for skill [Player 1 with 10 matches and a BV of 2306 (Summoner prime) vs Player 2 with 8000 matches and BV of 1634 (Victor 9K) - who is going to win? - Hint it's not Player 1] and doesn't translate directly to the MWO environment so they would all need to be recalculated.

Edited by EgoSlayer, 16 August 2014 - 12:23 PM.


#43 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 16 August 2014 - 12:13 PM

View PostMiken, on 16 August 2014 - 11:13 AM, said:

We need BattleValue based balance. Not Elo or silly 4x3...


BV was incredibly unbalanced and never worked. That's why it went to 2.0 and was then replaced in pretty much any and every competition by house rules.

Elo works and 4x3 has done more to balance the game than any other MM change in the games history.

#44 Alistair Winter

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Storm
  • Storm
  • 10,823 posts
  • LocationBergen, Norway, FRR

Posted 16 August 2014 - 12:36 PM

View PostMischiefSC, on 16 August 2014 - 12:05 PM, said:

matches played don't matter. You can't prevent stomps - they are a byproduct of team vs team play. That dynamic creates stomps, not Elo. Are you saying stomps never happen in competitive team play?

That's my point. You can't really stop it. And it doesn't matter how good a player is when their whole team is annihilated. You may have an average damage of 1000 per match, but when your whole team is wiped out instantly, you're gonna have a bad time.

But at least let me play with people who are competent. Today, I had a game in Alpine and no one knew where to go. So I said "Let's go to E8, E9, since we don't have ECM." Nobody replied, everyone just stood still or walked around in random directions. When the enemy team started attacking, some people had barely moved 200 meters from their spawn points, and our whole team was distributed along a 1,5 kilometer long line.

I don't mind losing if we have a team of good players with good communication, and we just get outplayed. What I can't stand, is herding cats.

View PostEgoSlayer, on 16 August 2014 - 12:05 PM, said:

You obviously don't realize that a match maker giving you an equal chance of winning means that you are matched against equal opponents....

I'm not matched against equal opponents. I regularly meet opponents that are both infinitely worse than me and considerably better than me. I regularly have teammates that have either been playing since closed beta or just got the game a few weeks ago. I regularly see Founders toy with new players in champion trial mechs, who don't know about LRM minimum range.

Nobody knows how matchmaker truly works, but I can tell you for damn sure that it has nothing to do with equal opponents.

#45 EgoSlayer

    Member

  • PipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 1,909 posts
  • Location[REDACTED]

Posted 16 August 2014 - 12:49 PM

View PostAlistair Winter, on 16 August 2014 - 12:36 PM, said:


I'm not matched against equal opponents. I regularly meet opponents that are both infinitely worse than me and considerably better than me. I regularly have teammates that have either been playing since closed beta or just got the game a few weeks ago. I regularly see Founders toy with new players in champion trial mechs, who don't know about LRM minimum range.

Nobody knows how matchmaker truly works, but I can tell you for damn sure that it has nothing to do with equal opponents.


Posted Image


If the match maker was perfect (it's not) then having an equal chance of winning means you are matched against like skill opponents. Mathematically that is undeniable. We do know enough about how the match maker works. And it does try for like skilled opponents based on Elo.

In order for that to happen you have to have a deep enough pool of players with matching skills, or barring that wait until that (potentially) small group of matching players are all available at the same time which could take hours in edge cases. Nobody wants to wait for games that long when the average play time is 7-8 minutes.

Until there are millions of concurrent users in MWO, there will never be perfect matches that get created in 2 minutes or less, unless you are in the heart of the skill curve. Removing the skill ranking from the match maker makes things *worse* not better because there is no basis of player abilities. The one thing that improves is the time to create a match, but it's far more likely to be completely unbalanced in player skills.

Edited by EgoSlayer, 16 August 2014 - 12:50 PM.


#46 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 16 August 2014 - 12:55 PM

View PostAlistair Winter, on 16 August 2014 - 12:36 PM, said:

That's my point. You can't really stop it. And it doesn't matter how good a player is when their whole team is annihilated. You may have an average damage of 1000 per match, but when your whole team is wiped out instantly, you're gonna have a bad time.

But at least let me play with people who are competent. Today, I had a game in Alpine and no one knew where to go. So I said "Let's go to E8, E9, since we don't have ECM." Nobody replied, everyone just stood still or walked around in random directions. When the enemy team started attacking, some people had barely moved 200 meters from their spawn points, and our whole team was distributed along a 1,5 kilometer long line.

I don't mind losing if we have a team of good players with good communication, and we just get outplayed. What I can't stand, is herding cats.


I'm not matched against equal opponents. I regularly meet opponents that are both infinitely worse than me and considerably better than me. I regularly have teammates that have either been playing since closed beta or just got the game a few weeks ago. I regularly see Founders toy with new players in champion trial mechs, who don't know about LRM minimum range.

Nobody knows how matchmaker truly works, but I can tell you for damn sure that it has nothing to do with equal opponents.


So a couple of important things.

I have games where I make a bad choice, end up way out of position and die early cuz I had THE DUMB. Happens to everyone.

Sometimes I'm derping in a troll build or a new mech. Same with everyone else.

Skill is not static; yesterday I killed TwinkyOverlord in a match. He's stupidly better than me but I got lucky, caught him at the right time when he was harvesting tears from my teammate and borked him from behind with my explodey-face TW build. He died early, his team followed.

Skill is not a true constant. It's situational and contextual and will vary match to match.

Finally, games can't be balanced. For Elo or personal improvement to work you need to win matches that are against you. This is a common misunderstanding about Matchmaking and Elo in general -

It does NOT give you even matches. You don't really gain Elo from balanced matches. You gain Elo by winning matches stacked against you; you lose Elo losing matches you should have won, your Elo doesn't change if you win matches you were expected to win or losing matches you were expected to do.

Without that there is no true way to measure improvement and success in the face of change. Newer players don't see how veterans play, veterans develop inbred habits because they only play with each other. The Matchmaker absolutely stacks matches for and against everyone, all the time. That is, in truth, its purpose. Sometimes you're the challenge to be overcome, sometimes you're the challenger.

The difference between real winners and everyone else is the ability to take that challenge as a challenge and differentiate themselves by passing more often than they fail. Blaming everyone else or the system or whatever is what keeps people in mediocrity.

#47 Alistair Winter

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Storm
  • Storm
  • 10,823 posts
  • LocationBergen, Norway, FRR

Posted 16 August 2014 - 01:02 PM

View PostEgoSlayer, on 16 August 2014 - 12:49 PM, said:


Posted Image



The point didn't go over my head. I deliberately left out certain parts when quoting you, because I immediately dismissed it as a combination of nonsense and stuff that wasn't relevant to the point I was making.

#48 EgoSlayer

    Member

  • PipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 1,909 posts
  • Location[REDACTED]

Posted 16 August 2014 - 01:06 PM

View PostAlistair Winter, on 16 August 2014 - 01:02 PM, said:

The point didn't go over my head. I deliberately left out certain parts when quoting you, because I immediately dismissed it as a combination of nonsense and stuff that wasn't relevant to the point I was making.


So you like ignoring math based facts? Ok then....

#49 Alistair Winter

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Storm
  • Storm
  • 10,823 posts
  • LocationBergen, Norway, FRR

Posted 16 August 2014 - 01:16 PM

View PostMischiefSC, on 16 August 2014 - 12:55 PM, said:

Without that there is no true way to measure improvement and success in the face of change. Newer players don't see how veterans play, veterans develop inbred habits because they only play with each other. The Matchmaker absolutely stacks matches for and against everyone, all the time. That is, in truth, its purpose. Sometimes you're the challenge to be overcome, sometimes you're the challenger.

First of all, I'm not here to suggest how MM should be improved for better ELO matching. I'm saying that I don't think PGI can do it well enough, so I want to get rid of the system.

Second of all, it's fairly easy to imagine a system that splits the whole mass of player into two groups (or more), forcing lower skilled players to climb to the top of their hill before possibly getting a high enough ranking to move on to the higher group. In chess, a player who just started playing will never be matched up against the highest ranked player in the world, or even a GM, for the sake of " a challenge to overcome".

View PostMischiefSC, on 16 August 2014 - 12:55 PM, said:

The difference between real winners and everyone else is the ability to take that challenge as a challenge and differentiate themselves by passing more often than they fail. Blaming everyone else or the system or whatever is what keeps people in mediocrity.

Telling the players to accept poor matchmaking as some exercise in stoicism is what keeps games in mediocrity.

View PostEgoSlayer, on 16 August 2014 - 01:06 PM, said:

So you like ignoring math based facts? Ok then....

Oh snap. You used the word 'math' and 'facts' followed by an ellipsis. I guess you won the argument. GG WP.

#50 Kjudoon

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Rage
  • Rage
  • 7,636 posts
  • LocationWisconsin

Posted 16 August 2014 - 01:19 PM

View PostMystere, on 16 August 2014 - 11:12 AM, said:


Well, there is a difference between "taking a defensive position" and "camping". The former, when done properly, actually works extremely well. The latter just gets you killed by attrition. Most PUG players, though, cannot tell the difference.




Well, Banzai charges seem to work for me, more often than not anyway. ;)

Banzai charges were always tactics of last resort and often resulted in the utter destruction of the Japanese force or a Pyrrhic victory that was quickly overturned in the next battle.

Most people don't know the difference between "Camping" and "Defensive Positions" unless they're either tryhards on the path to competitive level play, or competitive players themselves. Then they often know the difference. They also seem to know what cover is, firing lines, some map awareness and don't hyperfocus like a kid off his ritalin.

By the same token, I've watched time and time again, bad players charge the pug zapper or Radar Dish or Cauldera to be slaughtered by an entrenched position that they seem to think they automatically will win 'because they are Rambo and therefore have a forcefield of awesome'. This is the typical MWO player, not the exception. It is a known fact in military science circles that it takes 3-5 times the force to defeat a fortified position. When the position also has support (like turrets) it's even more difficult.

That is why defensive play is the best way to win in the majority of cases. But it's a game, and we're here to have a good time so that excuses all the bad play and stupid choices in the world.

#51 PappySmurf

    Member

  • PipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • 842 posts

Posted 16 August 2014 - 01:20 PM

HAHAHA ELO was a myth never happened never worked a year ago they scraped it now all the MM does is pick 24 random players and throw them together for a big Free For All. You are all just Clueless.

#52 EgoSlayer

    Member

  • PipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 1,909 posts
  • Location[REDACTED]

Posted 16 August 2014 - 01:33 PM

View PostAlistair Winter, on 16 August 2014 - 01:16 PM, said:

Oh snap. You used the word 'math' and 'facts' followed by an ellipsis. I guess you won the argument. GG WP.


Considering you ignored the reasons for your point about the match maker creating poor matches, then propose some other unknown skill system that would have the same population based issues as the current system, I'm not entirely sure what you think you are proving.

#53 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 16 August 2014 - 01:41 PM

View PostAlistair Winter, on 16 August 2014 - 01:16 PM, said:

First of all, I'm not here to suggest how MM should be improved for better ELO matching. I'm saying that I don't think PGI can do it well enough, so I want to get rid of the system.

Second of all, it's fairly easy to imagine a system that splits the whole mass of player into two groups (or more), forcing lower skilled players to climb to the top of their hill before possibly getting a high enough ranking to move on to the higher group. In chess, a player who just started playing will never be matched up against the highest ranked player in the world, or even a GM, for the sake of " a challenge to overcome".


Telling the players to accept poor matchmaking as some exercise in stoicism is what keeps games in mediocrity.


Oh snap. You used the word 'math' and 'facts' followed by an ellipsis. I guess you won the argument. GG WP.


You're not playing nubs 1 on 1 vs top tier players. You're playing pretty closely balanced (40-90 pt spread) matches, which is about perfect. It's balanced between teams. Not everyone always carries their weight every match and sometimes people do better than expected.

That is the while *point* of a ranking system. Nobody is ever static in their skill. If I drop you in the same match with the same teams on the same map 10 times in a row you'll do better every match - but so will everyone else. By match 10 you'll all do that map together very well and who wins/loses will differ though because some people will adapt and improve more quickly than others. If you won 12-0 in match 1 you might lose 0-12 in match 10 because the other guys adapted bettered, learned faster and improved more in those 10 matches.

It's not a bad matchmaker - it's fine for what it does. About optimal IMO for investment vs reward. If the matchmaker was absolutely *perfect* and got your skill right to the exact point... you'd still play differently and throw that calculation off. Without precognition there is no way to truly predict, just give good guesses. That's why it's done in averages over many games, those miscalculations wash out.

If what you're wanting is a game where you only play with people of your exact same skill level you'll need to strip all the customization out of mechs and situations and have a 'you must have X sleep and have drank Y coffee to play on this team' sort of measurements because your skill level will change over the day.

There is no stoic acceptance of anything. You're in a team vs team environment with 12 v 12. The quality of teammates and opponents will shift. If you don't want that... well, you're going to need to re-evaluate your wants. It's not logical or feasible.

It sounds like what you're wanting is everyone on both teams to be good at what they do every match. You'll need to play 12v12 private matches for that. If you're pugging.... well, you're going to play in a pick-up group. Want better teammates and opponents on average? Win more.

#54 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 16 August 2014 - 01:47 PM

View PostKjudoon, on 16 August 2014 - 01:19 PM, said:

Banzai charges were always tactics of last resort and often resulted in the utter destruction of the Japanese force or a Pyrrhic victory that was quickly overturned in the next battle.

Most people don't know the difference between "Camping" and "Defensive Positions" unless they're either tryhards on the path to competitive level play, or competitive players themselves. Then they often know the difference. They also seem to know what cover is, firing lines, some map awareness and don't hyperfocus like a kid off his ritalin.

By the same token, I've watched time and time again, bad players charge the pug zapper or Radar Dish or Cauldera to be slaughtered by an entrenched position that they seem to think they automatically will win 'because they are Rambo and therefore have a forcefield of awesome'. This is the typical MWO player, not the exception. It is a known fact in military science circles that it takes 3-5 times the force to defeat a fortified position. When the position also has support (like turrets) it's even more difficult.

That is why defensive play is the best way to win in the majority of cases. But it's a game, and we're here to have a good time so that excuses all the bad play and stupid choices in the world.


You get to say that when you're winning all the time. I base my opinions off of what wins and what doesn't by how the people who win all the time play. If you've dropped with any of the names in the top 5 of the various classes in the tournament for example you'll see them tell everyone 'push center, don't stop'. Generally that's all they'll say. As an example I watched Sun Cobra do that earlier today. He was the first one in and was kicking a$$ when the rest of us showed up.

Without voip and familiarity with each other fortification is no advantage in MW:O. It just makes you an easy target. Moving you are not an easy target. The person who pushes dictates the point and cadence of the fight, the person who pushes hardest wins. There are no 'fortified positions' in MW:O. Every point on every map has several ways in and out and is easy to flank. Also... arty + airstrike make clustering in one place a bad choice.

I'm not saying I'm a better player or smarter or any e-peen waving. I am however saying that you just got done complaining your win/loss was less than 1.0 for a lot of your mechs. There are people with 2.0 and up after many, many matches. They don't play how you play (or me for that matter, I'm bad at pulling back when badly damaged - I tend to let myself get ground down in a 'take you with me' approach), so wouldn't it make sense to play more like they play?

#55 Alistair Winter

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Storm
  • Storm
  • 10,823 posts
  • LocationBergen, Norway, FRR

Posted 16 August 2014 - 02:07 PM

View PostMischiefSC, on 16 August 2014 - 01:41 PM, said:

If what you're wanting is a game where you only play with people of your exact same skill level you'll need to strip all the customization out of mechs and situations and have a 'you must have X sleep and have drank Y coffee to play on this team' sort of measurements because your skill level will change over the day.

I'll try to explain my point more concisely.

I don't demand to play with othe rpeople with the exact same skill level. I demand to be placed in a separate queue from people who have only just started playing the game. Now, PGI has already done something like that. I think cadets have an artificially lowered ELO untill they finish their cadet matches, or something. But I want more of that. I don't think a Founder with 5000 matches should be in the same queue as someone with 50 matches.

That's all.

#56 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 16 August 2014 - 02:20 PM

View PostAlistair Winter, on 16 August 2014 - 02:07 PM, said:

I'll try to explain my point more concisely.

I don't demand to play with othe rpeople with the exact same skill level. I demand to be placed in a separate queue from people who have only just started playing the game. Now, PGI has already done something like that. I think cadets have an artificially lowered ELO untill they finish their cadet matches, or something. But I want more of that. I don't think a Founder with 5000 matches should be in the same queue as someone with 50 matches.

That's all.


That's good logic. I would generally agree.

One problem though is people starting new accounts and farming nubs. it happens plenty.

Another is that many matches /= higher skill. There are plenty of bads with a lot of matches.

Should nubs be completely isolated? The problem there is how do they get better? Who helps them? A newbie queue that skilled players can drop in would be great for those who want to help shepherd new players but otherwise how are they going to learn that LRMs don't work inside 180 unless you're clans, PPCs not under 90, or anything about Ghost Heat.

Consider them an 'environmental hazard'. Like lava. They're a threat to everyone. Sometimes you accidentally drop into the lava, sometimes the other team does.

#57 VonFrundsberg

    Member

  • Pip
  • 17 posts

Posted 16 August 2014 - 03:05 PM

View PostMischiefSC, on 16 August 2014 - 12:13 PM, said:


BV was incredibly unbalanced and never worked. That's why it went to 2.0 and was then replaced in pretty much any and every competition by house rules.

Elo works and 4x3 has done more to balance the game than any other MM change in the games history.


At this point, what would it hurt to try BV2.0? They've got a test server, right?

#58 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 16 August 2014 - 03:34 PM

View PostVonFrundsberg, on 16 August 2014 - 03:05 PM, said:


At this point, what would it hurt to try BV2.0? They've got a test server, right?


So eliminate player skill completely because all that matters is a rough and unreliable estimate of the value of the mech?

Good players do good in bad mechs. Bad players do bad in good mechs. Or do you think that anyone can be a good player with the right mech?

#59 Kjudoon

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Rage
  • Rage
  • 7,636 posts
  • LocationWisconsin

Posted 16 August 2014 - 03:35 PM

View PostMischiefSC, on 16 August 2014 - 01:47 PM, said:


You get to say that when you're winning all the time. I base my opinions off of what wins and what doesn't by how the people who win all the time play. If you've dropped with any of the names in the top 5 of the various classes in the tournament for example you'll see them tell everyone 'push center, don't stop'. Generally that's all they'll say. As an example I watched Sun Cobra do that earlier today. He was the first one in and was kicking a$$ when the rest of us showed up.

Without voip and familiarity with each other fortification is no advantage in MW:O. It just makes you an easy target. Moving you are not an easy target. The person who pushes dictates the point and cadence of the fight, the person who pushes hardest wins. There are no 'fortified positions' in MW:O. Every point on every map has several ways in and out and is easy to flank. Also... arty + airstrike make clustering in one place a bad choice.

I'm not saying I'm a better player or smarter or any e-peen waving. I am however saying that you just got done complaining your win/loss was less than 1.0 for a lot of your mechs. There are people with 2.0 and up after many, many matches. They don't play how you play (or me for that matter, I'm bad at pulling back when badly damaged - I tend to let myself get ground down in a 'take you with me' approach), so wouldn't it make sense to play more like they play?


Complaining? No just stating historical record. You're the one inferring that I'm winning all the time so it makes it easy to say. I'm losing because of of the Banzai charges and precisely what you claim is your flaw. Too many people practicing the 'take you with me' style instead of being smart and trying to survive the match. I mean, come on... listen to an expert on aggressive warfare, George Patton: "The object is not die for your country, but to make the other poor dumb B* die for his!"

That's what I do every time. Those who don't cost their teams games just about every single time unless inordinately lucky.

View PostMischiefSC, on 16 August 2014 - 03:34 PM, said:


So eliminate player skill completely because all that matters is a rough and unreliable estimate of the value of the mech?

Good players do good in bad mechs. Bad players do bad in good mechs. Or do you think that anyone can be a good player with the right mech?

Define "Skill" first." If skill is pavlovian twitch instinct, it's not skill. Bad mechs hurt all players. Good players can overcome the flaws of bad mechs. Good mechs make bad players only okay, unless they have REAL skill, not just good reflexes. Good mechs make good players great because they can use all the benefits.

Edited by Kjudoon, 16 August 2014 - 03:36 PM.


#60 ApolloKaras

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,974 posts
  • LocationSeattle, Washington

Posted 16 August 2014 - 03:44 PM

View PostMischiefSC, on 16 August 2014 - 03:34 PM, said:


So eliminate player skill completely because all that matters is a rough and unreliable estimate of the value of the mech?

Good players do good in bad mechs. Bad players do bad in good mechs. Or do you think that anyone can be a good player with the right mech?



Exactly you can roll some derp build, and roll an abhorrently low BV. BV does not account for skill. Well I guess it could, but then again you're looking at a more complicated match system.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users