Jump to content

Should We Switch From Personal Elo To Per-Mech Elo?

Balance

22 replies to this topic

#1 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,307 posts

Posted 14 June 2014 - 10:14 PM

I constantly see this situation in game - full team of "green" (i mean camo) players vs premade roflstomp. Some of them are noobs, who are still using trial mechs, some of them are players, that want to try new trial mechs before buying it and some of them are players, who have just bought their new mechs. The problem with that players - that they are underperforming or have underperforming mechs. And I constantly ask myself a question, why can't matchmaker distribute this players on both teams evenly? I'm just trying to get an answer and, I guess, the problem is in fact, that ELO is personal. I.e. game doesn't feel difference in player between fully tuned 2xPPC+AC killer poptart and player in trial mech with stock build. Matchmaker still puts me into noobbish teams against premades and thinks, that I should stomp the whole enemy team solely, even if I'm driving mech with stock build, I've just bought. So I'm asking this question: should PGI implement per-mech ELO instead of personal one?

Edited by MrMadguy, 14 June 2014 - 10:20 PM.


#2 SirLANsalot

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,540 posts
  • LocationWashington State

Posted 14 June 2014 - 10:15 PM

ELO is split into 4 groups, aka the 4 Tonnage Classes.


But I do agree that it should be based on a per mech basis rather then a per Class basis.

#3 DaisuSaikoro Nagasawa

    Member

  • PipPipPipPipPipPipPip
  • Mercenary
  • Mercenary
  • 974 posts
  • Google+: Link
  • LocationTaipei, Taiwan

Posted 14 June 2014 - 10:54 PM

Was the whole point of this so you could impress people by the fact that you take a Highlander poptart. And that that is somehow correlated with you being a great player?

#4 Clit Beastwood

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,262 posts
  • LocationSouthern California

Posted 14 June 2014 - 11:15 PM

View PostMrMadguy, on 14 June 2014 - 10:14 PM, said:

I constantly see this situation in game - full team of "green" (i mean camo) players vs premade roflstomp. Some of them are noobs, who are still using trial mechs, some of them are players, that want to try new trial mechs before buying it and some of them are players, who have just bought their new mechs. The problem with that players - that they are underperforming or have underperforming mechs. And I constantly ask myself a question, why can't matchmaker distribute this players on both teams evenly? I'm just trying to get an answer and, I guess, the problem is in fact, that ELO is personal. I.e. game doesn't feel difference in player between fully tuned 2xPPC+AC killer poptart and player in trial mech with stock build. Matchmaker still puts me into noobbish teams against premades and thinks, that I should stomp the whole enemy team solely, even if I'm driving mech with stock build, I've just bought. So I'm asking this question: should PGI implement per-mech ELO instead of personal one?



It puts you where you belong.

#5 Appogee

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 10,967 posts
  • LocationOn planet Tukayyid, celebrating victory

Posted 14 June 2014 - 11:31 PM

It's an interesting question. Would the current high Elo players be just as effective if they didn't take meta builds?

I play occasionally with one of the recent tournament winners. He has good skills in any chassis, however, when he takes a non-meta build he's certainly not as effective.

Personally, I have a very bad time whenever I am trying to level a new chassis, because I get handed PUG after PUG team which I cannot personally carry against high Elo teams, unless I'm in one of my highly-optimised Heavies or Assaults.

So I think your idea has merit, complicated though it would be to implement because loadouts can vary significantly even on one particular chassis.

Edited by Appogee, 14 June 2014 - 11:34 PM.


#6 Arnold J Rimmer

    Member

  • PipPipPipPipPipPipPip
  • Civil Servant
  • Civil Servant
  • 892 posts

Posted 15 June 2014 - 12:33 AM

View PostAppogee, on 14 June 2014 - 11:31 PM, said:

It's an interesting question. Would the current high Elo players be just as effective if they didn't take meta builds?

You'd have to pry them out of their Victors first.

In the clan I used to run with, one guy became bored at the lack of challenge presented to him in MWO, it all became 'too easy' to win/kill. I heard him say this quite a bit, then dropped with him before he gave in. Nothing but meta builds. I asked him why he didn't challenge himself with a different chassis once in a while, maybe even (:ph34r:) with a balanced loadout, and he just responded with confusion. His actual words were 'Why would I do that?'

Stuff like that is why I have no time for people stroking their own egos while humping the meta.

But I live in hope that the next patch brings change :huh: The incoming Buckton fix should breathe life back into the brawlers, for instance...

#7 Appogee

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 10,967 posts
  • LocationOn planet Tukayyid, celebrating victory

Posted 15 June 2014 - 12:46 AM

I play with a guy who constantly tries ''weird builds that shouldn't work'' - and is consistently successful with them.

(Eg. at the moment he has declared that "small lasers are the new meta" and he's busy killing everyone with those.)

It keeps him entertained. And he's a great pilot from all the diverse experience he gets. I'd hate to come up against him in a genuine meta build, he'd be unstoppable.

#8 Torgun

    Member

  • PipPipPipPipPipPipPipPip
  • 1,598 posts

Posted 15 June 2014 - 01:21 AM

The problem is matchmaking has to be remade from scratch. It never made any sense to base it directly off win/losses without a single reflection on how you do individually. It doesn't matter if you get a single digit damage before dying or try to carry hard by doing more than your entire lance. Both these performances are judged equally purely based on the outcome of the match. In no shape or form can that make sense unless you only do 12-mans with the exact same team every time. You have to judge the performances individually or else the matchmaking will never have the right data to work with.

Edited by Torgun, 15 June 2014 - 01:24 AM.


#9 Zerberus

    Member

  • PipPipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 3,488 posts
  • LocationUnder the floorboards looking for the Owner`s Manual

Posted 15 June 2014 - 02:44 AM

View PostMrMadguy, on 14 June 2014 - 10:14 PM, said:

I constantly see this situation in game - full team of "green" (i mean camo) players vs premade roflstomp. Some of them are noobs, who are still using trial mechs, some of them are players, that want to try new trial mechs before buying it and some of them are players, who have just bought their new mechs.


And some of them are players that want you to think exactly that. I used to switch my mechs back to basic green whenever that chassis happened to be a trial mech.Posted Image

Not to mention that the 5 basic F2Pcolors are pretty useless as "camo", wheras green actually has some merit.
:huh:

Quote

The problem with that players - that they are underperforming or have underperforming mechs. And I constantly ask myself a question, why can't matchmaker distribute this players on both teams evenly? I'm just trying to get an answer and, I guess, the problem is in fact, that ELO is personal. I.e. game doesn't feel difference in player between fully tuned 2xPPC+AC killer poptart and player in trial mech with stock build. Matchmaker still puts me into noobbish teams against premades and thinks, that I should stomp the whole enemy team solely, even if I'm driving mech with stock build, I've just bought. So I'm asking this question: should PGI implement per-mech ELO instead of personal one?

Per mech Elo is both a database and New player experience nightmare just waighting to happen.... Every time a new mech is released, people who just wreck face in that weight class would be pitted against the steering wheel underhive for weeks until the scores even out. If someone rarely plays that chassis, his score will never even out properly.

We have weight class ELO, and IMO that strikes an ideal balance between having a global ELO with a huge amount of uncomparable information and having a Per chassis Elo with a minute amount of very specific information, and thereby actually has a effective accuracy greater than either extreme solution because the sample size is both reasonably large and reasonably comparable.

And since the matchmaker is just days away from being dropped in in a completely recoded state, how about testing it first when it does instead of wasting breath complaining about the old, soon to be irrrevant one? :ph34r:

Edited by Zerberus, 15 June 2014 - 02:50 AM.


#10 Thorqemada

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 6,396 posts

Posted 15 June 2014 - 02:46 AM

Mech Bound ELO would break every time you sell a Mech?

#11 Torgun

    Member

  • PipPipPipPipPipPipPipPip
  • 1,598 posts

Posted 15 June 2014 - 04:06 AM

View PostThorqemada, on 15 June 2014 - 02:46 AM, said:

Mech Bound ELO would break every time you sell a Mech?


I don't see how, it's like you can sell a mech and when you rebuy it you still have the same amount of mech exp? It's just a question of keeping an Elo number on every mech you own or have owned.

#12 Levi Porphyrogenitus

    Member

  • PipPipPipPipPipPipPipPipPip
  • Mercenary
  • Mercenary
  • 4,763 posts
  • LocationAurora, Indiana, USA, North America, Earth, Sol, Milky Way

Posted 15 June 2014 - 04:53 AM

The best system I've seen uses 50 game thresholds.

1 - Start with an account-wide Elo rating with initial 50 games at New Player level and seeding after the 50th match.

2 - Apply the account-wide Elo to every weight class with fewer than 50 games played.

3 - Once a weight class has 50+ games played, use weight class Elo for every chassis in that class.

4 - Once a chassis within a weight class has 50+ games, use chassis Elo for every variant within that chassis.

5 - Once a variant has 50+ games, use variant Elo for every drop using that specific variant.

This allows for scaling levels of precision with Elo rating, and should eliminate most of the problems with experienced players dropping in new (to them) variants/chassis (since they'd use their Account or Weight Class Elo until they hit 50 games with the new chassis).

I'm also a fan of basing Matchmaker Rating on a combination of win/loss, kills/death (assists count as half a kill), and average XP per match.

#13 NextGame

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,072 posts
  • LocationHaggis Country

Posted 15 June 2014 - 04:58 AM

Sure, it would have to be at chassis level rather than variant level though due to the interchangeability of clan mech parts.

#14 Wispsy

    Member

  • PipPipPipPipPipPipPipPipPip
  • Talon
  • Talon
  • 2,007 posts

Posted 15 June 2014 - 05:32 AM

You have a few things wrong there as far as I know. Even if you have individual Elo scores for every different chassis it will not change how matchmaker makes groups. What most people get confused with is that the games are not balanced smartly. When you get put in a game matchmaker thinks you will lose, you simply lose less Elo rating. It does not prevent you from being put in games in which you will lose. From what I can tell it will pick a 4man and start making a game around them, it will then pick another 4man within like 1000 Elo and place them on the other team. Then it will try and fill in the rest of the spaces, keeping groups even etc. What it does not do, is pick 24 people with Elo ratings as close to each other as possible, and divide them up to give each team as fair a chance as possible. So no matter what your Elo, you will be in a game in which it is completely random how much your team will be favoured to win/lose and the Elo will simply be adjusted accordingly. So you can still quite happily have a game full of trial mechs getting dominated, just like you can have a game in which 3 top Elo premades get put on the same team as 3 lore/casual premades and lose in less then 2mins.

#15 Reitrix

    Member

  • PipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 1,130 posts

Posted 15 June 2014 - 05:34 AM

View PostAppogee, on 15 June 2014 - 12:46 AM, said:

I play with a guy who constantly tries ''weird builds that shouldn't work'' - and is consistently successful with them.

(Eg. at the moment he has declared that "small lasers are the new meta" and he's busy killing everyone with those.)

It keeps him entertained. And he's a great pilot from all the diverse experience he gets. I'd hate to come up against him in a genuine meta build, he'd be unstoppable.


Heh, i slapped an AC2 and 3 MGs into my Huginn. It's surprisingly effective. And quite fun, i haven't enjoyed Ravens as much as this for a while. Low damage, but i'm getting kills since I'm fairly accurate.

Edited by Reitrix, 15 June 2014 - 05:35 AM.


#16 ScarecrowES

    Member

  • PipPipPipPipPipPipPipPipPip
  • Knight Errant
  • Knight Errant
  • 2,812 posts
  • LocationDefending the Cordon, Arc-Royal

Posted 15 June 2014 - 06:25 AM

Elo is a joke anyway and should be dropped for a true modern skill system. Of course, for that to work, PGI would have to implement a proper scoring system as well. How are you going to have a skill system based on factors like wins/losses, which are entirely dependent on the skill of the 11 other players on your team relative to the skill of the 12 players on the other team? You want the skill system to work? Start taking into account those factors... like every modern multiplayer skill system out there today.

#17 draiocht

    Member

  • PipPipPipPipPipPipPip
  • 791 posts

Posted 15 June 2014 - 02:26 PM

View PostMrMadguy, on 14 June 2014 - 10:14 PM, said:

even if I'm driving mech with stock build, I've just bought. So I'm asking this question: should PGI implement per-mech ELO instead of personal one?

Your proposal doesn't go as far as your premise alludes. That could be rated per loadout... or as could be reasoned, per loadout per map. However, that'd be a fairly substantial database, never mind doing that per player (or anything more accurate). Might make sense to use those stats to adjust c-bill/xp rewards, though. (I don't know; I'm assuming a partyback has significantly worse performance on alpine than a gaussjager. Maybe that's not the case at all.)

Just throwing that out into the void forums....

#18 LowSubmarino

    Member

  • PipPipPipPipPipPipPipPip
  • Bad Company
  • 1,091 posts

Posted 15 June 2014 - 04:11 PM

I sometimes don't quite understand elo. Lately I really dont understand the matchmaker at all. I have been playing a pretty long time and repeatedly end up in groups that just seem utterly confused. They wander about aimlessly, dont take cover, have no idea where enemy could be (even on mwo's huge number of ridiculously tiny maps), dont watch their radar so dont react to teammates being slaughtered right behind them, and die like flies.

On the other side I find myself opposing organized teams that walk in formation and just terminate those aimlessly wandering totatlly noobish players.

Damn, I don't care that there are newbs. Everybody starts fresh. But how in hell is it even possible to have highest elo players in organzied teams fight against an all out lone wolf, completly skillless team of pugs. And why Jesus Christ do I get those teams. Like somebody said lately I get demolished 10 times with maybe one victory in between. Sometimes the streaks last longer and I stopped playing before getting a win.

A few months ago (maybe half a year) it would be pretty balanced. I'd win a few and lose a few. Now its like 95 % losses. Like every game. And not just losses but complete annihilations. I wanne track this with some gameplay footage over a few hours so nobody can claim that i am exagrerating cause I am not. I literally lose all matches.

This is so weird. At first I thought it was an unlucky streak. But it went on and on. For weeks. Then I stopped for a few weeks cause I thought that its a bug or something and the matchmaker is messed up. I come back and get this for several days again. Then I stop and come back to see if its still there. It is like some low or medium elo ppl get mixed up against some of the best teams. Sometimes i get several 12 : 0s or 12 : 1s in a row.

If somebody thinks the matchmaker works even remotely. It doesn't. It mixes experienced premades against the worst players the universe has seen. And it does so in 9 out of 10 times. Not sometimes not pretty often but almost always. And Diablo himself punishes me every single time I drop by condemning me to die beside those confused random pugs. Well, lately just by the names I can predict if we lose and im always right. I should start a paranormal bureau of prophecy which players can insta check when a match starts and I predict their death.

Maybe I am a wizard or something. Or the matchmaker is so warped that nobody on earth or the underworth could ever fix it again. Just deinstall the matchmaker and replace it by some very early beta version. Anything is better than how it is right now.

#19 Hillslam

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,198 posts
  • LocationWestern Hemisphere

Posted 15 June 2014 - 04:40 PM

View PostFierostetz, on 14 June 2014 - 11:15 PM, said:

It puts you where you belong.

Now THIS HERE is an intriguing concept!

What if the PGI MM was not just, in fact, a matchmaking engine, but was, in fact, a matchmaking engine AND Karma Dispenser?!?!

The mind reels.
(...and I'm f**ked)

#20 Davers

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 9,886 posts
  • Facebook: Link
  • LocationCanada

Posted 15 June 2014 - 04:41 PM

My win/loss is almost exactly 50/50. While I experience lopsided matches, both ways, I generally have very good and fun matches. I find Elo works fine for me.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users