26 replies to this topic

[Riverboat Sam]

I've read a bunch of conflicting things about ELO, most recently that if you're in a match and your team wins your ELO goes up even if you sucked. True? How does ELO calculation work?

[IraqiWalker]

If your team was PREDICTED to win, and you win, your Elo stays the same. If your team is predicted to lose, and wins, then your Elo goes up.

[Egomane]

Enlil09, on 13 October 2014 - 01:04 AM, said:

I've read a bunch of conflicting things about ELO, most recently that if you're in a match and your team wins your ELO goes up even if you sucked. True? How does ELO calculation work?

It's Elo! It's numerical skill measurement invented by Arpad Elo and therfore named after him. Originally invented for chess, Elo has found it's way into many more applications today.

Elo is based on an assumption. If you have a higher score then your opponent, then you are more likely to win. All users, within an Elo system, start at the same value. Usually an average of the total numerical range used.

In some Elo formulas, your value increases, no matter if you won against a lower ranked player or one of equal or higher rank, or decrease if you lose against anyone. In others you will simply rise because you are playing a better opponent, or fall for playing a worse one, additionally to the match result change. Not so in MWO. Here your value only changes if you win against a higher rated player, or lose against a lower rated one. Yes, Elo is a two way system. Your score can rise and it can fall. Depending on the difference between the players, the rise/fall can be nearly insignificant or it can be massive. Usually an Elo system has a cap on how much the value can change in a single instance.

In MWO you have 4 Elo values. One for each weight class of mechs: Light, medium, heavy and assault
Each of these scores starts at average and rises or falls, with you playing in each weight class. If you never play an assault mech, your assault Elo will always stay at average value.

During your first 25 matches you are actually playing with a significantly lower Elo value. The intention behind this, is to keep you from playing against far superior opponents. The matchmaker is trying to create matches with all 24 players being within the same Elo range. After about 2 minutes of searchtime however, this range increases, to a maximum at about 5 minutes search, that can pit you against almost anyone.

If you drop as a team, instead of playing solo, the Elo of all players in the team is used as an average, for matchfinding.

Once you are done with the match, the average Elo value of both sides is calculated. The winner will then rise and the loser will fall, unless the winner had a better total Elo average to begin with. So if your team is predicted to lose and you lose, nothing changes. If it is predicted to win and it wins, nothing changes either. But if you don't perform to the expected result a change will happen.

Edited by Egomane, 13 October 2014 - 01:52 AM.

[The Basilisk]

Much as Ego explained.
Sadly its only a fraction of the truth.
The Problem with elo is that it is ignoring your equipment.
Elo just takes the *rating* of a player in a certain wightclass without taking into account what mech or what equipment he is driving.
So there always is the possibility your high elo assault mech player drives a banshee M with only flammers ( exaggerated ) and not a single basic trait unlocked on his mech not even carrying dhs and or any modules.
The other extreme is the sesoned heavy mech player with a low assault mech rating driving a fully fleshed out Daishi.

In reality this means elo gives you the *fitting* players for a team but you may get 3 assault mechs with only LRMs, newb heavys and your mediums and lights will have to tank.

Thats how stomps happen.

And thats the reason why the matchmaker may be a pice of programming art but next to useless at building teams for public matches.

[Karl Streiger]

The Basilisk, on 13 October 2014 - 02:16 AM, said:

In reality this means elo gives you the *fitting* players for a team but you may get 3 assault mechs with only LRMs, newb heavys and your mediums and lights will have to tank.

Thats how stomps happen.

Not the full truth either. Stomps happen because failure is cascading. Every player makes mistakes - the quantity and quality of this mistakes - is cumulative. if you have a close finish - the number of failures done by both teams is equal - if you have a stomp on team did more mistakes. Of course competitive players don't do that much mistakes.

The W/L ratio is an aggregate it includes your Mech, the loadout of that Mech and the impact you have with that Mech.
Anyhow - a single player only influence 8% of his team - and his Mech is only another 50% of this.
So if you run all the time a Full leveled Dire Whale and take a Warhawk - your 8% team impact may be negative - you don't know your Mech you don't have a good idea about the build.
So maybe your team will loose - and you ELO is reduced - but you ELO will rise again, as soon as you Mech is leveled.

I think its a good system - because ELO is a prediction system. It predict an outcome on your former battle performance. If you behave unpredictable - and to take a new Mech is unpredictable - it won't work accurate in the beginning - but soon it will adjust.

Even with equipment values - unpredictable - or even abusive behaviour will be possible.

[Kin3ticX]

heres some information I gathered up

https://docs.google....g27342656c_2125

[The Basilisk]

Karl Streiger, on 13 October 2014 - 02:33 AM, said:

Not the full truth either. Stomps happen because failure is cascading. Every player makes mistakes - the quantity and quality of this mistakes - is cumulative. if you have a close finish - the number of failures done by both teams is equal - if you have a stomp on team did more mistakes. Of course competitive players don't do that much mistakes.

The W/L ratio is an aggregate it includes your Mech, the loadout of that Mech and the impact you have with that Mech.
Anyhow - a single player only influence 8% of his team - and his Mech is only another 50% of this.
So if you run all the time a Full leveled Dire Whale and take a Warhawk - your 8% team impact may be negative - you don't know your Mech you don't have a good idea about the build.
So maybe your team will loose - and you ELO is reduced - but you ELO will rise again, as soon as you Mech is leveled.

I think its a good system - because ELO is a prediction system. It predict an outcome on your former battle performance. If you behave unpredictable - and to take a new Mech is unpredictable - it won't work accurate in the beginning - but soon it will adjust.

Even with equipment values - unpredictable - or even abusive behaviour will be possible.

Yea, yea you are right saying there are ALSO other factors to winning/loosing a match.
But even if you get a perfectly even battle elo whise I say equipment, Mech tier and combination of certain weapons and equipment pieces completely ofsets the elo rating as it is used atm.
Pair 4 players Elo whise:

Map: Forest colony
Team 1: ECM Raven with tag and beagle, Stormcrow custom, Ilya 2Gauss 3Mlaser, Direwolf 2xGauss+ laser vomit
Team 2: Puma prime, Nova prime, Orion 1M, Banshee 3M

Who do you think will win ?
And this is not even an extreme situation given the assumption ALL players have their mechs on the same skill level.

And thats where ELO realy realy is failing.
As you said there can be now cascading factors of advantages and disadvantages for either side.
Unskilled Mechs, trollbuilds, weapons that have been nerfed or bugged to oblivion, toxic weapon combination,

None of this factors are caused by deliberately beeing unpredictable.
They are caused by a Matchmaker just ignoring all those factors.
And thats why solo queue matchesseem to be just random and imbalanced in their outcome.
The factors going in those matches ARE unpredictable by definition.

Personal skill ( given the player mastered at least basic skills in moving and shooting ) without coordination in procedure and equipment is almost negligible because you are only representing 8% of your team.
Its the combination of the team that creates the overall effect and this combination is not taken into account by the current elo matchmaking system.
It trys to build teams of mechwarriors instead of building teams of warriors in their mechs.

What we need is a system of combat worthnes modified by player elo.
We already have some official statements on Mech Tiers.
So what? Assign combatvalues to weapons, weaponcombinations and chassis and let the playerelo be represented as combat worthnes rating adding to your mechs rating.
( Hey does that sound familiar to you ?...Battletech combatvalues ? )

Edited by The Basilisk, 13 October 2014 - 03:50 AM.

[Karl Streiger]

Hm - its becoming a discussion topic.
Not the idea of this forum.

- for the OTs question:
here are the formulas:
http://mwomercs.com/...ost__p__1626065

In the topics regarding the game mode system Russ stated that the average difference in ELO in PUG games is 50.
So the chance of winning is 42% ELO decided that you have to loose - that means victory is really possible - and will result in a change in ELO of 21.
On the other hand - it will also reduce your ELO as fast.

Ok - i think i see the point.

First regarding the "lance layout" - ELO works with the assumption that players know what they are doing.
ELO for chess wouldn't be very representative if you have 2 players that discuss if you have to play Chess with 2d6 or 1d10.

So the second lance should know that their Mechs are not the current optimal builds - matched vs a almost equal enemy in ELO terms they may loose. They will loose until their Mechs are matched vs less talented players with Meta Mechs - or equal minded players - using "Mechs" that feel like BattleTech.

Hm - lets figure it out:
I have a Mech with a hypothetical ELO of 1800. I'm matched vs players of ELO 1700-1900

Without BV - and a "bad" build

• 1800 vs 1900 loose - keep elo
  
• 1800 vs 1900 loose - keep elo
  
• 1800 vs 1900 loose - keep elo
  
• 1800 vs 1700 loose - because of Mech slight reduce
  
• 1775 vs 1800 loose - keep elo
  
• 1775 vs 1700 loose - reduce elo

OK now with BV -1000 and 1800 ELO vs 2000 BV and 800 ELO - would be adequate?

Hm - i think the BV is less important: so try 500 + 1800 vs 1000 + 1300 - could work

• 2300 vs 2200 win - no change
  
• 2300 vs 2400 win - change
  
• 2325 vs 2400 win - change

OK there are problems - first for the ELO only part - you only loose ELO if you are predicted to win and loose - because of your build or the others build. So you will see a long and bloody way of defeats - until your ELO is settled down.

The BV part will match you vs new or less experienced players in good builds. I'm not sure how you can't figure exact values - for equipment. Look in BatteTech TT BV - works because all players are equal able to use those weapons.
In MWO - a player that don't know that his LRMs may need a lock on - and deal less damage below 180m - would not be able to get full advantage of a LRM 100 build - so BV is only the edge. The more important part is the player.

Edited by Karl Streiger, 13 October 2014 - 04:14 AM.

[Egomane]

Please keep BV out of this discussion. This is no thread to promote this idea, we have plenty of others for that, but one to answer a question.

Edited by Egomane, 13 October 2014 - 04:52 AM.

[Rear Admiral Tier 6]

ELO is mix of arcane sorcery and high level chaos theory,only guys who claim to know their ELO are the ones losing.

[L Y N X]

Elo was intended for 1v1 matches, while it has been used in team vs team competitions, it has been revoked or obsoleted in many of those situations for a better team vs team rating system.

[Mad Ox]

Been covered pretty well but in-general I call it EGO as the people freaking over it are using it to stoke their uber awesome greatness.

In the end game can only create team with what people are playing so some nights will suck big some will be good. It can't compensate for the morons been out all night so drunk there is blood in their alcohol or butt heads being butt heads along with billion of there human related variables.


Though I will say I have much much better matches on MWO then with WOT. Of course WOT I think at least half the players just log in to suicide then spend rest of battle in chat yelling at team for being noobs and not supporting them letting them die.

Edited by Mad Ox, 13 October 2014 - 01:43 PM.

[Xtrekker]

Elo is a placebo number designed to invoke a self-appointing caste system as part of a decade-log social experiment spearheaded by the UBC Department of Psychology under their social behaviours program (http://psych.ubc.ca/...al-personality/). It primarily demonstrates the relationships between the "id" the "ego", and in the case of MWO, how this relates to digital social interaction and "the online persona", particularly among youths 16 to 24 years of age as well as choice examples beyond those parameters. A premise on the subject, initially submitted as a relevant precursor to the MWO experiment, can be found at http://prezi.com/-bq...y-presentation/. If you view the presentation, you'll note how the MWO experiment currently mirrors the expected outcome in a number of rather disturbing ways, e.g. "white knights" ("mind guards" - self-appointing members who shield the group from dissenting information), illusions of unanimity, direct pressure as "disloyalty", and stereotyping opposition as ignorance. The current pro-PGI trend is the beginning of the next phase of the experiment, conformity, then ultimately obedience to authority (as a followup to Yale's "Milgram Experiment", inspired by the war trial of {Godwin's Law} war criminal Adolf Eichmann). This is, of course, fully sanctioned by the Canadian government as part of their continuing mission of world domination.

So, yeah, that's Elo.

(edit, typo)

Edited by Xtrekker, 14 October 2014 - 11:47 AM.

[TheCaptainJZ]

Suffice it to say, matchmaking is a hot topic in this game

[RazorbeastFXK3]

What does ELO stand for? Estimated.. Luck Outcome? Load Outcome? Lick Out?

[Alaskan Nobody]

RazorbeastFXK3, on 13 October 2014 - 05:01 PM, said:

What does ELO stand for? Estimated.. Luck Outcome? Load Outcome? Lick Out?

It doesn't stand for anything

It is the name of the person who put the idea together

Edited by Shar Wolf, 13 October 2014 - 05:12 PM.

[RazorbeastFXK3]

And here I thought everyone was just saying 'allo.

Shar Wolf, on 13 October 2014 - 05:06 PM, said:

It doesn't stand for anything

It is the name of the person who put the idea together

[VoodooLou Kerensky]

Electric Light Orchestra

[Riverboat Sam]

Wow. I had no idea. Thanks guys. This is very helpful. It also explains a lot.

[Tim East]

Xtrekker, on 13 October 2014 - 02:30 PM, said:

("mind guards" - self-appointing members who shield the group from dissenting information)

Egomane, on 13 October 2014 - 04:44 AM, said:

Please keep BV out of this discussion. This is no thread to promote this idea, we have plenty of others for that, but one to answer a question.

I laughed.

In answer to the question, Elo is basically the method by which people of hypothetically equivalent abilities are matched against one another in an attempt to make the most entertaining matchups possible. From what I've heard, yes, your Elo can rise even if your personal performance was a bit lacking, though that seems a bit unlikely since it would require your team to have been predicted to lose based on previous performances AND to carry, since your own performance that match was less than ideal.

I will say that even though it is not ideal in the sense of giving uniformly close matches as yet, the matchmaker IS greatly improved in comparison to the days when large groups would drop into the pug queue and just wreck faces. The game is a work in progress, and while sometimes mistakes happen, generally each iteration is better than the last, and that's good enough for me.