Jump to content

Elo Worthless


298 replies to this topic

#221 Odin

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 498 posts

Posted 17 November 2013 - 10:20 AM

View PostMotroid, on 17 November 2013 - 10:12 AM, said:

I think keeping us in the ELO middle is "working as intended".
Didn't someone say that the goal of MM is 50/50 w/l ratio for everyone?
If that is true, MM works for me because after more than 10000 drops (solo, 2-4 and 12) my w/l ratio is 5490/4538.
It has always been the same cycle. First a massive win streak followed by a massive loss streak. And so on.



50/50? For everyone? Are you sure? If thats intended lol, this game never can be considered a competitive game.
Would be a nightmare for me.
Had around 4700 drops Wins / Losses 2,682 / 1,929. So, there should be room for player skill and improvement.
Not a matchmaker messing up all and every drop, to keep it civil for the average.

#222 Sug

    Member

  • PipPipPipPipPipPipPipPipPip
  • The People's Hero
  • The People
  • 4,629 posts
  • LocationChicago

Posted 17 November 2013 - 10:21 AM

View PostMotroid, on 17 November 2013 - 10:12 AM, said:


Didn't someone say that the goal of MM is 50/50 w/l ratio for everyone?
If that is true, MM works for me because after more than 10000 drops (solo, 2-4 and 12) my w/l ratio is 5490/4538.



That's a bit skewed from 50/50, probably from the group drops. Your ratio is closer to 45/55, with almost 1000 more wins than losses.

Compared to my pure pug ratio of 2,105 / 2,140 or 49.6/50.4

View PostOdin, on 17 November 2013 - 10:20 AM, said:

50/50? For everyone? Are you sure? If thats intended lol, this game never can be considered a competitive game.


Yeah it's basically a coin flip if you pug.

Edited by Sug, 17 November 2013 - 10:23 AM.


#223 Odin

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 498 posts

Posted 17 November 2013 - 10:38 AM

View PostSug, on 17 November 2013 - 10:21 AM, said:

[/size]


Yeah it's basically a coin flip if you pug.


No wonder players love their cheese-builds so much ;)

Edited by Odin, 17 November 2013 - 10:38 AM.


#224 Macbrea

    Member

  • PipPipPipPipPipPip
  • The Hunter
  • The Hunter
  • 270 posts

Posted 17 November 2013 - 10:57 AM

View PostOdin, on 17 November 2013 - 10:20 AM, said:

50/50? For everyone? Are you sure? If thats intended lol, this game never can be considered a competitive game.



That is how the Elo system is suppose to work. To give a quote from the wiki page on it: "The difference in the ratings between two players serves as a predictor of the outcome of a match. If two players with equal ratings play against each other, they are expected to score an equal number of wins (50% each). A player whose rating is 100 points greater than his or her opponent's is expected to score 64%, if the difference is 200 points the expected score for the stronger player is 76%."

So, the goal of the system is to get you to your Rating. At that point, you should generally sit for quite awhile. And at that point, you should win/lose as many matches as you play. Because, it doesn't account for changes in mechs of the same weight class. Your mileage may vary.

For example: I am much better in my Catapults then I am in my Dragons. Though, each match I play is at the same Elo Score.

#225 Sug

    Member

  • PipPipPipPipPipPipPipPipPip
  • The People's Hero
  • The People
  • 4,629 posts
  • LocationChicago

Posted 17 November 2013 - 11:05 AM

View PostMacbrea, on 17 November 2013 - 10:57 AM, said:

So, the goal of the system is to get you to your Rating. At that point, you should generally sit for quite awhile. And at that point, you should win/lose as many matches as you play.


Man, how boring would sports be if they used this f'd up system...

#226 Roadbeer

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 8,160 posts
  • LocationWazan, Zion Cluster

Posted 17 November 2013 - 11:12 AM

View PostSug, on 17 November 2013 - 11:05 AM, said:


Man, how boring would sports be if they used this f'd up system...


It's used in the BCS, and we all see how well THAT works out.

#227 Odin

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 498 posts

Posted 17 November 2013 - 11:19 AM

@ Macbrea

;) since I, by now, do more solos then premade drops, I tell you, I sit there already lol.

Edited by Odin, 17 November 2013 - 11:20 AM.


#228 MischiefSC

    Member

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

Posted 17 November 2013 - 02:01 PM

View PostAbivard, on 16 November 2013 - 09:07 PM, said:


It is not working because it is not possible for it to work. No surprise there.

Elo = A system to rank chess players by individual ability. Can be used for many things where one on one play is involved.

MWO= Multiplayer team based game, with dynamically changing teams.

What part of 'Elo is not applicable to MWO' do you not understand?

What you are implying is the same as the statement about monkeys randomly typing out the complete works of shakespeare. it may be possible in theory, but not likely to ever happen.

It most certainly is not something that can be done consistently!

You ask for empirical evidence to support my position but in return I must accept your interpretation of these theories as they may apply to MWO with no evidence what so ever?

Yet the mathematical theorems you present refute your position. You seem to be just trying to baffle us with walls of text as well as hoping to prey upon the ignorance and pride of others.


Mathematical formula are not theories. Statistics could be viewed as a theoretical science I guess but one played out by clear mathematical principles.

Your position seems to be that 'Elo doesn't apply to MW:O because it's too complicated'. That's not an argument. What is your mathematical basis for saying that unlike any and every other statistical variable win/loss in MW:O is simply too complex to be accounted for.

Given that I deal with statistical variables literally dozens of times and in large data aggregates even thousands of times less significant than your impact on win/loss in MWO every single day your assumptions are ones that I disprove 8 or 9 times every morning by about 9:30 am PST.

It's human nature to believe that human behavior can not be predicted or accounted for. This is not correct. When you roll a six sided dice there are countless variables. Force with which it is thrown, twist on release, air viscosity, friction on impact, gravity, imperfections on the dices surface, would be hard to list them all. Yet regardless of them all the odds of any number coming up on a dice, no matter who throws it, are 1 in 6. The impact of your fellow teammates can be largely distilled down to a dozen or so criteria and the probability of them playing out in a match with you is the same as their playing out in a match with someone else. You know what is the same in your matches? You. Your impact is what can be reliably measured from all your matches.

It's not a tough concept.

I'm not asking you to create a mathematical formula, I'm saying you want to challenge the viability of Elo then track 100 matches and show that it somehow defies probability theory. I'd love to see it. I already know you're wrong but why don't you show us.

All you've got right now is bad feelz and a philosophical attempt to make that mathematically true. It's not. Not to be offensive here but you're demonstratively wrong. That's okay but if you want to demonstrate that you're right, put some actual data here.


View PostNightfire, on 16 November 2013 - 11:21 PM, said:

*snip polite and well thought out response*


Slicing metrics more finely is possible but provides diminishing returns. It also increases the probability of including a mistaken or skewed variable. Most people in statistics will tell you simplify simplify simplify, the more fundamental the result you're trying to find the more reliable it is. I'll add the caveat that it depends what you're trying to do. If I'm trying to identify how you'll vote in an election or if you prefer scented or unscented deodorant then I do want a lot of variables because I'm not solving for a performance point but creating a consumer mosaic.

What you're wanting to do is pin down precise behaviors or instant metrics to tie to how you affect winning or losing. I absolutely get the impulse but it's a whack-a-mole objective. The current system makes sense because you get awarded XP and cbills for such things for an immediate reward for performance but measured on win/loss as an aggregate over time.

I get that Merc Corps will want stats and I'm all for providing that but again, that's different than Elo. Elo measures how well you impact your teams odds of winning no matter how you do it. Stats say what exactly someone is good at. A Merc team can build a squad with specific metrics to give them the right synergy to win matches. Elo doesn't see or account for synergy - even with premade teams it treats everyone like a pug. Right now that's what we want and for any pug environment that's what you want. I don't think Elo should be involved in CW matches, I think instead you'll have a ranking with factions (driven by your win/loss on missions for them no doubt) which is similar but without the pugging concept behind it.

A team is made up of individuals. Each individual contributes to the team. You can look at a teams overall metrics and see how each teammate influences the total but the easy metrics (in MW:O it's score, KDR, damage) don't actually paint as clear a picture of overall performance.

Each individual impacts their team statistically and measurably. To pretend otherwise is to abdicate personal responsibility for your performance. If I have a team of 15 people and only 1 works hard they will still get more work done over 1,000 hours than a team of 15 in which nobody works hard. I won't notice it much after 5 hours, but after 1,000? Absolutely I do. To pretend otherwise is disingenuous.

I don't disagree at all that perception is important - that's why I'm all for changes that make the matchmaker tie bands of Elo players more tightly, though I suspect the outcry then will be that it puts people with someone who is too hard for them to beat or that even if I could show you that everyone on your team was generally as good or better than the other team you'd still say the rest of your team sucked because you lost.

There in lies the rub. Nobody likes to lose and when they do nobody likes to say 'I should have done more'. Especially in a team. I'd bet that in more than 9 out of 10 losses there was someone on your team who 'did better' even though your team lost. I'm all for managing expectations but the reality is that we're simply more comfortable blaming a group of strangers than a group we know. That's the only real difference between dropping in a 12man and losing a bunch and dropping in pugs and losing a bunch.

The math behind Elo works. It may not create an environment everyone enjoys - some people don't want things to get more challenging as they progress, they want to progress so that they have an advantage over others. Someone prior called that 'being competitive'. I'd say it's the opposite but whatever. Personally I like having my matches continue to be challenging. If it's not why don't I just play an offline game in godmode?

#229 Kubernetes

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Blazing
  • The Blazing
  • 2,369 posts

Posted 17 November 2013 - 03:11 PM

View PostOdin, on 17 November 2013 - 06:33 AM, said:

If ELO in MWO ever should be a plausible, true measure of your abilities, it will be based only on your personal performance during the match and NOT on its outcome.

LOL its THAT simple.

ELO based on win/loss is in fact a measure of the random groups (PUG) or premade drops, your a part of.
Since this game only offers team-death matches, lets ignore capping, there are only 2 factors to be considered:
  • Your kills
  • Your damage done

Currently there are no other measurable factors in game, which could be taken into account.


This is my main issue with it. Win/loss is a secondary statistic. Yes, it's true that if you're a good player your contributions will lead to more wins, but you're only one mech in a field of 24. A huge number of games end with a result that is totally at odds with your individual performance. I've died in 3 min with 50 dmg and my team has won. In others I've dominated the match and we still lose.

Does it all even out at the end? Possibly, but it's dissatisfying to think that you're being scored on a secondary statistic that largely depends on the performance of the other 91.7% of your team. Full disclaimer--I have no idea if I'm good, bad, or average. The whole thing is so opaque to me. Did I just kill 3 hapless noobs or did I happen to catch 3 skilled guys in a vulnerable spot? Who knows?

#230 Victor Morson

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 6,370 posts
  • LocationAnder's Moon

Posted 17 November 2013 - 03:41 PM

View PostOdin, on 17 November 2013 - 10:20 AM, said:

50/50? For everyone? Are you sure? If thats intended lol, this game never can be considered a competitive game.
Would be a nightmare for me.


It works great.. in games that are actually super competitive. For example you might have a 1:1 win rate, but that's only because it's begun pairing you up against people just as good as you, so every game is tense. That's the theory and how it works in, say, StarCraft 2 for the most part.

View PostOdin, on 17 November 2013 - 10:20 AM, said:

Had around 4700 drops Wins / Losses 2,682 / 1,929. So, there should be room for player skill and improvement.
Not a matchmaker messing up all and every drop, to keep it civil for the average.


The problem in MW:O however is that way too many factors are out of your control, so most people's improved win rate comes from 4 man games with their group. Otherwise, MW:O averages around 50/50 not because the games are balanced, but because the games are total {Scrap} shoots if you end up on the winning/losing team.

#231 Sug

    Member

  • PipPipPipPipPipPipPipPipPip
  • The People's Hero
  • The People
  • 4,629 posts
  • LocationChicago

Posted 17 November 2013 - 08:05 PM

View PostVictor Morson, on 17 November 2013 - 03:41 PM, said:

For example you might have a 1:1 win rate, but that's only because it's begun pairing you up against people just as good as you, so every game is tense.


That would awesome. Right now 1 game out of 20 is an actual tense, good, 12 - 11 game.

#232 Ghogiel

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2021 Gold Champ
  • CS 2021 Gold Champ
  • 6,852 posts

Posted 17 November 2013 - 08:13 PM

View PostSug, on 17 November 2013 - 10:14 AM, said:


You can affect the outcome of a game, but eventually you will be averaged out with worse players, and are unable to significantly affect your Elo because you only change your Elo signifigantly by winning against a team you have low odds of beating, or losing to a team you had high odds of beating.

I affect the outcome of every game.

Even those I have low odds of winning.

#233 Odin

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 498 posts

Posted 17 November 2013 - 11:38 PM

View PostKubernetes, on 17 November 2013 - 03:11 PM, said:

This is my main issue with it. Win/loss is a secondary statistic. Yes, it's true that if you're a good player your contributions will lead to more wins, but you're only one mech in a field of 24. A huge number of games end with a result that is totally at odds with your individual performance. I've died in 3 min with 50 dmg and my team has won. In others I've dominated the match and we still lose.

Does it all even out at the end? Possibly, but it's dissatisfying to think that you're being scored on a secondary statistic that largely depends on the performance of the other 91.7% of your team. Full disclaimer--I have no idea if I'm good, bad, or average. The whole thing is so opaque to me. Did I just kill 3 hapless noobs or did I happen to catch 3 skilled guys in a vulnerable spot? Who knows?


The same here. Absolutely my experience so far.

#234 KinLuu

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,917 posts

Posted 17 November 2013 - 11:47 PM

View PostSug, on 17 November 2013 - 10:14 AM, said:

You can affect the outcome of a game, but eventually you will be averaged out with worse players, and are unable to significantly affect your Elo because you only change your Elo signifigantly by winning against a team you have low odds of beating, or losing to a team you had high odds of beating.


I disagree with this conclusion.

Even slight differences on the side of the individual player, can have massive differences - not only on individual performance, but also on the teams performance. The most obvious example might be having ECM vs. not having ECM.

But there are other examples as well. Just to give some food for thought, here is a funny example out of my personal stats:

These are my global winrates, since the stats are tracked:

Posted Image

While this is my winrate with the advanced sensor range module:

Posted Image

As you can see, there is a significant difference. Of course, with such a small sample size (only 800 games), this could be a statistical anomaly. But I would argue, that the reason for this difference could also be having the module. The ability to lock onto a target at a longer range, can give your team a significant advantage in information warfare. And this allows your teammates to make more informed decisions. Resulting in a higher winrate. And a higher winrate should result in a higher elo, over time.

I believe that the individual performance of each player has a massive influence on his elo. At least after a couple of hundred games.

#235 Sug

    Member

  • PipPipPipPipPipPipPipPipPip
  • The People's Hero
  • The People
  • 4,629 posts
  • LocationChicago

Posted 18 November 2013 - 12:00 AM

I feel like i'm taking crazy pills.



View PostKinLuu, on 17 November 2013 - 11:47 PM, said:

Just to give some food for thought, here is a funny example out of my personal stats:


Though to be fair that was a funny example.

Edited by Sug, 18 November 2013 - 12:04 AM.


#236 KinLuu

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,917 posts

Posted 18 November 2013 - 12:20 AM

View PostSug, on 18 November 2013 - 12:00 AM, said:

I feel like i'm taking crazy pills.

Though to be fair that was a funny example.


You just fight a war that you cannot win. Like Don Quichotes crusade against the windmills.

This discussion is very old. It was done a number of times... the best example would be LoL.

And it was proven again and again, that over time, variantions of elo work nearly perfect.

The only problem with MWO is, that it takes far longer to work, because there are more players per team, than in LoL or CS:GO.

#237 Odin

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 498 posts

Posted 18 November 2013 - 12:27 AM

View PostVictor Morson, on 17 November 2013 - 03:41 PM, said:


It works great.. in games that are actually super competitive. For example you might have a 1:1 win rate, but that's only because it's begun pairing you up against people just as good as you, so every game is tense. That's the theory and how it works in, say, StarCraft 2 for the most part.


Hi Victor
Dunno Starcraft, I think MWO with its huge pool of different weapons and Mech designs is simply to hard to "frame" into anything
predictable. So each and every match turns out to be open. Seen under these conditions, Kubernetes described it perfect. Your simply to small a factor in regard of the actual outcome of the match. No matter what you, as a single (solo) player can do.

View PostVictor Morson, on 17 November 2013 - 03:41 PM, said:


The problem in MW:O however is that way too many factors are out of your control, so most people's improved win rate comes from 4 man games with their group. Otherwise, MW:O averages around 50/50 not because the games are balanced, but because the games are total {Scrap} shoots if you end up on the winning/losing team.


Exactly.

The overall weapons nerv we've seen over time here, is ment to prevent singly pilots to dominate the outcome, the main factor is the focus, the team work. So win/loss is the result of all 24 players, not in your hand - no matter what you accomplished. Thats why only premades dominate, which in fact is a good thing. Curios only, that team work is ment to be a measure of your individual "score".
Thats IMO is a mistake. I am be no means the best pilot here, but I regard myself as experienced. But no matter what, ending up as one of the top pilots of the drop, more often then not, doesn't seem to have any effect overall.
You either get shot out early, score a number of kills or help with lots of damage done, win or loss is not in your hand alone.
I think individually, your to small a part of it.

So my reasoning is, your individual ELO, should reflect your and only your doings.

#238 Ghogiel

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2021 Gold Champ
  • CS 2021 Gold Champ
  • 6,852 posts

Posted 18 November 2013 - 12:55 AM

View PostOdin, on 18 November 2013 - 12:27 AM, said:


Hi Victor
Dunno Starcraft, I think MWO with its huge pool of different weapons and Mech designs is simply to hard to "frame" into anything
predictable. So each and every match turns out to be open. Seen under these conditions, Kubernetes described it perfect. Your simply to small a factor in regard of the actual outcome of the match. No matter what you, as a single (solo) player can do.

Start AFKing and after 1000 matches come tell me how you maintain a 50/50 W/L

#239 Odin

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 498 posts

Posted 18 November 2013 - 04:00 AM

View PostGhogiel, on 18 November 2013 - 12:55 AM, said:

Start AFKing and after 1000 matches come tell me how you maintain a 50/50 W/L


Ghogiel, effect here would be obvious, but would proof nothing.

#240 KinLuu

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,917 posts

Posted 18 November 2013 - 04:26 AM

View PostOdin, on 18 November 2013 - 04:00 AM, said:


Ghogiel, effect here would be obvious, but would proof nothing.


Oh?

So now you admit, that your personal performance does have an influence on your teams chance to win?

Ghogiel takes aim, Ghogiel scores. 1:0





12 user(s) are reading this topic

0 members, 12 guests, 0 anonymous users