Jump to content

Elo Is For Chess, Not Mwo


198 replies to this topic

#41 L Y N X

    Member

  • PipPipPipPipPipPipPip
  • The Nimble
  • The Nimble
  • 629 posts
  • LocationStrana Mechty

Posted 09 October 2014 - 02:38 AM

Karl, do yourself a favor, and begin making yourself aware of the number of individual match scores above 30 points on both the winning and losing side. In my experience, 99% of the time, the winning side has more match scores over 30 points. In stomps often there are half a dozen players on the losing side with match score in the single digits... SINGLE DIGITS?!!? yes

Those Players in the single digits have no competitive right to be in that match, they need to be paired up with other players of their callibre, this is the way to reduce the number of one sided stomps.

View PostLily from animove, on 09 October 2014 - 02:35 AM, said:


Wait a moment, what you and russ discsussed has nothign to do directly with ELO, it ahs to do with the factors influencing the elo rating.

Yes we could change elo rating not only based on wins. We could give oints for this like:

assist 10 points, kill 20 points, narc 1 point, spot 1 point, win 50 points. And then we compare every match participants gathered points and current elo to decide if his elo drops or raises.

Thsi would smooth the elo ratings, good palyers losing would not drop so much. bad players beign carried would not raise so much. And by this elo is changed softer giving people more of a realistic rating.


It would not be ELO if you made matchmaking based off of Average Match Scores. Please use the link provided and read up on ELO what it is. But Yes we need a better rating system.

Edited by 7ynx, 09 October 2014 - 02:40 AM.


#42 Torgun

    Member

  • PipPipPipPipPipPipPipPip
  • 1,598 posts

Posted 09 October 2014 - 03:00 AM

View Post7ynx, on 09 October 2014 - 02:38 AM, said:



It would not be ELO if you made matchmaking based off of Average Match Scores. Please use the link provided and read up on ELO what it is. But Yes we need a better rating system.


At this point I don't think anyone cares what it would be called as long as it would work better. An example: I had a game where I got stuck before even having contact with the opponents and became a turret. At the end of the match I did a whopping 57 damage with 1 assist, we won 12-2 and I went up in Elo as much as the guys that did tons more during that match. Good measure of my performance? Not at all. And this is happening each and every match right now to determine everyone's performance and subsequent Elo ranking. It's really no wonder it's failing.

#43 H I A S

    Member

  • PipPipPipPipPipPipPipPipPip
  • 4,971 posts

Posted 09 October 2014 - 03:20 AM

Elo works fine, dont touch it. I wouldnt match up with Noobs, they have high Matchscore, because they spread Damage all over the target.

Sry for my Bad english =)

Edited by HiasRGB, 09 October 2014 - 03:22 AM.


#44 Duke Nedo

    Member

  • PipPipPipPipPipPipPipPipPip
  • CS 2023 Top 12 Qualifier
  • CS 2023 Top 12 Qualifier
  • 2,184 posts

Posted 09 October 2014 - 03:48 AM

The main problems as I see them with the current ELO system is:

1) Unbalanced spawn points. On most maps actually, I have a very good feeling whether my team will win or lose depending on which side we get. This sort of incapacitates ELO input from these maps. Imo the only balanced maps are the near symmetic ones like HPG, canyon and factory. All other maps probably induces unlikely wins AND unlikely losses to such a degree that it taints the ELO and hides that small 1/12 player contribution and eventually cancels out to zero.

2) You can build ELO in the group queue and then drop in the solo queue with an ELO rating that has nothing to do with your expected performace in a PUG game. I assume this can work in both directions, either highly inflated or deflated depending on which kind of groups you regularly drop with. In the solo queue, at least from impressions during tournaments, everyone is rather competent but also very much mortal.

3) ELO completely disregards mech balance. This has been brought up many times. A Summoner is not a Timberwolf, an Awesome is not a DDC and a Magic Jesus Kitfox is not an Adder when it comes to capability to drive a win. Hell, even a Spider is not a Spider... (ecm vs no ecm)

From the story posted about the Nova, isn't the win-streak disturbing to you? Sure, one pilot is heavily underestimated by the ELO system, but since he is still only 1/12 of the team and he is in a NOVA, it sure looks like he was carried by the MM, and that should just never happen. At least to me that is not obvious why that would happen, but it could be the same thing that cause major streaks and MM-griefing, especially during tournaments when the solo queue is injected with all kinds of probably extreme ELO's and heavily scewed Mech class distributions. During these times I am getting win streaks in lights/mediums, and loss streaks in heavies/assault, but that may be entirely circumstantial, I know.

It would be nice if it would be possible to somehow disconnect group elo from solo elo... and more balanced maps would probably help too.

The ELO system does its job somehow I guess, but I don't think it would get worse if you instead just used experience brackets to sort the beginners from the seasoned, and thereby roughly sort the skilled from the unskilled. Should be very effective for filtering out the trial mechs, but less effective to separate the truly awesome from the average pilots.

Edited by Duke Nedo, 09 October 2014 - 03:51 AM.


#45 Karl Streiger

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 20,369 posts
  • LocationBlack Dot in a Sea of Blue

Posted 09 October 2014 - 03:51 AM

View Post7ynx, on 09 October 2014 - 02:38 AM, said:

Those Players in the single digits have no competitive right to be in that match, they need to be paired up with other players of their callibre, this is the way to reduce the number of one sided stomps.


Good God is it really that hard to understand.
OK lets suggest it works that way.
You take all that crappy player that deal single digit Matchscore - like me - and toss them into the same barrel?
How do I climb out of that barrel?
By getting Matchscores of > 100?
Hey here I'm again in your match - because i dealt top 3 match score.

And what happen to you - bad luck, a nasty spotter - or you did the mistake to place yourself in a DDC.
You get primed, cored and what ever before you are even able to fire a single shot.

And BULLSHIT yes its obvious that in a roflstomp the looser team has a lot of players that didn't deal any damage - thats not bad play - thats typical occurrence when you become a prime - I lost an Atlas in 3seconds. See target - react at target - dead....


View PostDuke Nedo, on 09 October 2014 - 03:48 AM, said:

The ELO system does its job somehow I guess, but I don't think it would get worse if you instead just used experience brackets to sort the beginners from the seasoned, and thereby roughly sort the skilled from the unskilled. Should be very effective for filtering out the trial mechs, but less effective to separate the truly awesome from the average pilots.

and how do you want to do this?
I run my AS7-D (F) as a AS7-D - no frankenmonster - just the good old LRM20, AC20, 4MLAS and a SRM6.

Edited by Karl Streiger, 09 October 2014 - 03:55 AM.


#46 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 09 October 2014 - 03:57 AM

I don't like Elo but I HATE Battle Value. I saw the way it can be abused/used. Replacing one flawed system for another is just bad.

Bishop in a Centurion-AH v Me in a Centurion-AH. Both stock. Bishop is a better Medium pilot then I am so unless there is a variable for experience, BV won't be any more fare here than on TT.

#47 Glythe

    Member

  • PipPipPipPipPipPipPipPip
  • 1,566 posts

Posted 09 October 2014 - 04:02 AM

Elo was a nice try for online games but if fails hardcore.

In the end what it does is screw you over for being good. If you are good you are (or at least were when they first put it in) more likely to get half wit team mates when the game can't find a good match for you.


On the other hand if you're terrible then you love Elo because you are a baddie. You couldn't buy a win if you tried so the game has to hold your hand and make sure that you're having fun.

What ever happened to winners win and losers get their arse kicked to the curb?

You could always have a 'cadet' league for those who want unranked games or for new players in trial mechs.


Elo causes match failing to take longer meaning you play less games. Over a lifetime of playing MWO (how ever much time that is) you will get better/fair matches if you play more games.

I played a long time before Elo and let me tell you it was a lot more fun that way. Click battle and it starts in about 5 seconds. I've waited in the queue for a game longer than the game actually lasted. Thanks for that MM I'm so glad you are "working".

Match tonnage, clan mechs and ECM but just throw away the Elo please.

Edited by Glythe, 09 October 2014 - 04:08 AM.


#48 Chrithu

    Member

  • PipPipPipPipPipPipPipPip
  • Bad Company
  • 1,601 posts
  • LocationGermany

Posted 09 October 2014 - 04:04 AM

View PostRhent, on 08 October 2014 - 09:20 PM, said:

ELO + Weapon Load Outs = Workable System
ELO Only = Random Variability that makes it completely, utterly totally useless

Lets say the following:
ELO matchmaker sets up two teams who have a score of 100 for each sides ELO, but by the luck of the draw its Alpine and one team has 4 LRM mechs + 2 ECM light spotters versus a team of brawlers w/o ECM. Equal ELO, but the map and the mechs are set that the brawlers are going to lose all the time, assuming the lights w/ ECM play as spotters and its just game over, very, very quickly.

ELO has to take into account certain builds and tools, things like:
ECM
Large LRM Loadouts
Large AC Loadouts
Large Gauss Loadouts

You could have a high ELO player running a very subpar mech versus another high ELO player in a carry mech and the carry mech is going to win always. MM has to identify the carry mechs and take that into account for ELO. You can't ignore them and have ELO work.


This one has it right: Good match making must be based on a measurement of loadout strength AND player skill. It must use both or it won't get good results.

BUT, as the OP said Elo does NOT qualify as a good method of measuring player skill in MWO. In fact I think there is too much random factors in a team based game to measure player skill in a relative form. I think the best way to go about measuring player skill is to do an absolute measurement based on the match score system averaging over a set amount of past games to smoothen out the random factors.

During every solo challenge people agree that the point system is a good start point for skill measurement (just needs to cover support like NARC/TAG/ECM and the weighting of damage done better) and in any case a better measurement than Elo based on team win/loss.

I say scrap Elo completely, base the skill measurement on the match score system with adjustments and combine it with a Battle Value measuring the strength of a loadout including weapons, equipment, modules, upgrades, ammo and armor.

I think that is the best route to go and should be the top priority project for AFTER CW is released. Until then do no more changes to the MM and just get CW finished.

Edited by Jason Parker, 09 October 2014 - 04:21 AM.


#49 Duke Nedo

    Member

  • PipPipPipPipPipPipPipPipPip
  • CS 2023 Top 12 Qualifier
  • CS 2023 Top 12 Qualifier
  • 2,184 posts

Posted 09 October 2014 - 04:06 AM

View PostKarl Streiger, on 09 October 2014 - 03:51 AM, said:

and how do you want to do this?
I run my AS7-D (F) as a AS7-D - no frankenmonster - just the good old LRM20, AC20, 4MLAS and a SRM6.


Not sure if I understand the question... if you bring a bad build you want to play against trial mechs so you can have some fun, is that it? Perhaps I just misunderstand... ?

#50 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 09 October 2014 - 04:09 AM

View PostDuke Nedo, on 09 October 2014 - 04:06 AM, said:


Not sure if I understand the question... if you bring a bad build you want to play against trial mechs so you can have some fun, is that it? Perhaps I just misunderstand... ?

A Stock Atlas-D is not a bad build. Only a bad Pilot would think it is bad.

#51 Stormwolf

    Member

  • PipPipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 3,951 posts
  • LocationCW Dire Wolf

Posted 09 October 2014 - 04:10 AM

BV is the way too go, players with min maxed builds will always have higher BV due to all the expensive components.

#52 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 09 October 2014 - 04:14 AM

View PostStormwolf, on 09 October 2014 - 04:10 AM, said:

BV is the way too go, players with min maxed builds will always have higher BV due to all the expensive components.
But a Min/Max build in a fool's hands is less lethal than a well balanced Mech in a good pilot's hands.

#53 Glythe

    Member

  • PipPipPipPipPipPipPipPip
  • 1,566 posts

Posted 09 October 2014 - 04:15 AM

Match score is what everyone seems to want. Can we try playing with JUST match score as the determining factor for player matching? Wait that won't work.... because ECM, clan mechs, etc. Let the computer sort out the weight, ecm and clan balance as one category and then match score as the other. There probably needs to be some sort of match value inflation for really high poke "alpha" builds.

View PostJoseph Mallan, on 09 October 2014 - 04:09 AM, said:

A Stock Atlas-D is not a bad build. Only a bad Pilot would think it is bad.


It's bad compared to the same build with say double heat sinks and Endosteel and more heatsinks/ammo.

#54 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 09 October 2014 - 04:19 AM

View PostGlythe, on 09 October 2014 - 04:15 AM, said:

Match score is what everyone seems to want. Can we try playing with JUST match score as the determining factor for player matching? Wait that won't work.... because ECM, clan mechs, etc. Let the computer sort out the weight, ecm and clan balance as one category and then match score as the other. There probably needs to be some sort of match value inflation for really high poke "alpha" builds.



It's bad compared to the same build with say double heat sinks and Endosteel and more heatsinks/ammo.
So move the goal line. Got it. It's a good build just not the best build.

#55 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 09 October 2014 - 04:20 AM

View PostJoseph Mallan, on 09 October 2014 - 03:57 AM, said:

I don't like Elo but I HATE Battle Value. I saw the way it can be abused/used. Replacing one flawed system for another is just bad.

Bishop in a Centurion-AH v Me in a Centurion-AH. Both stock. Bishop is a better Medium pilot then I am so unless there is a variable for experience, BV won't be any more fare here than on TT.


And therein lies the rub.

Adding a person's "score" (whether it be Elo or some form of XP based level system) as a value to the BV would make a difference. Granted, I've seen BV/CV abused as much as you have, so we both know it can be metagamed.

One of the BIGGEST problems with the whole MM/Elo thing is that it seems that PGI is getting all their data from the group queue. Gotta tune the MM to make team Elo's as close to equal as possible....works pretty good in the group queue. In the Group queue, you don't get a lot of new players that have figured out how to group up.

In the solo queue, however, it's a different story. We get stuck with ALL skill levels because the MM was set up to make team Elo scores equal by "average." I realize that "average" isn't that huge in the group queue....but in the solo queue, you can expect to have to carry at least 4 new players every single match.

All the tweaking and nerfing of weapon systems, electronics, etc work differently in the group queue than the solo queue. You can tell right after they tweak something....the group players come to the solo queue to try out the new shineys on the unsuspecting baby seals. Then the whining begins.

#56 KuunLan

    Rookie

  • 2 posts

Posted 09 October 2014 - 04:22 AM

@karl

You don't want to climb out there , you keep your atlas at your elo land where you don't have monsters with 4 uac10's or 6 uac5's stalking you .You play with likwminded players and have fun .

I simply dislike the elo imbalance in matches , i know that is because the low server population but i would prefer a 4x4 game at same elo than a 12vs 12 with 3 guru's +7 noobs vs 4 vets and 7 med and a noob like today .

BV without some elo is a bad bad bad bad ideea . My kitfox almost ALWAYS outdamage one of the direwolfs of my team

Edited by KuunLan, 09 October 2014 - 04:25 AM.


#57 H I A S

    Member

  • PipPipPipPipPipPipPipPipPip
  • 4,971 posts

Posted 09 October 2014 - 04:24 AM

View PostStormwolf, on 09 October 2014 - 04:10 AM, said:

BV is the way too go, players with min maxed builds will always have higher BV due to all the expensive components.


BV is no answer for Playerskill :P

#58 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 09 October 2014 - 04:25 AM

View PostWillard Phule, on 09 October 2014 - 04:20 AM, said:


And therein lies the rub.

Adding a person's "score" (whether it be Elo or some form of XP based level system) as a value to the BV would make a difference. Granted, I've seen BV/CV abused as much as you have, so we both know it can be metagamed.

One of the BIGGEST problems with the whole MM/Elo thing is that it seems that PGI is getting all their data from the group queue. Gotta tune the MM to make team Elo's as close to equal as possible....works pretty good in the group queue. In the Group queue, you don't get a lot of new players that have figured out how to group up.

In the solo queue, however, it's a different story. We get stuck with ALL skill levels because the MM was set up to make team Elo scores equal by "average." I realize that "average" isn't that huge in the group queue....but in the solo queue, you can expect to have to carry at least 4 new players every single match.

All the tweaking and nerfing of weapon systems, electronics, etc work differently in the group queue than the solo queue. You can tell right after they tweak something....the group players come to the solo queue to try out the new shineys on the unsuspecting baby seals. Then the whining begins.
And why wouldn't you take that data? It is a game meant to be played by teams against teams? I play mostly PUG now since Most of Murphy's moved on when Teams were nerfed to 4 or less. I don't blame them either, a Team game that caters to Soloist is breaking the idea of the game. make Solaris7 for the PUGs who want personal glory, and keep the rest of the game as is for those who want a real fight.

You would not see me on Solaris very often With or with out the Law.

Edited by Joseph Mallan, 09 October 2014 - 04:46 AM.


#59 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 09 October 2014 - 04:27 AM

View PostJoseph Mallan, on 09 October 2014 - 04:25 AM, said:

And why wouldn't you take that data? It is a game meant to be played by teams against teams? I play mostly PUG now since Most of Murphy's moved on when Teams were nerfed to 4 or less. I don't blame them either, a Team game that caters to Soloist is breaking the idea of the game. make Solaris7 for the PUGs who want personal glory, and keep the rest of the game as is for those who want a real fight.

You would not see on Solaris very often With or with out the Law.


I guess that my point is that the dynamic is different between the two queues. The Solo queue....where, technically, more people play (especially new players) has been ignored for so long now that it's virtually unplayable most of the time.

It'd be nice if the solo queue would get a little love.

#60 DaZur

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 7,511 posts
  • LocationWisconsin

Posted 09 October 2014 - 04:30 AM

This fricking gem again.... :rolleyes:

It's pretty amazing how Elo protagonists keep falling back on the fact that Elo pairings genesis came by way of chess yet fail to appreciate outside of BV and CV, that were actual BT canon metrics, most other match-makers used today also found genesis in equally dissimilar competitive events.

What I find even more entertaining is the flawed premise that there's a magic MM algorithm out there that can eliminate roflstomps...





21 user(s) are reading this topic

0 members, 21 guests, 0 anonymous users