Jump to content

Why Elo Doesn't Work Here


633 replies to this topic

#21 Bhael Fire

    Banned - Cheating

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 4,002 posts
  • Twitter: Link
  • Twitch: Link
  • LocationThe Outback wastes of planet Outreach.

Posted 22 January 2014 - 04:19 PM

View PostSandpit, on 22 January 2014 - 03:50 PM, said:

Isn't ELO based solely on the individual performance? If you do 600 damage an get 3 kills isn't that going to push your individual ELO up?


Nope.

Arpad Elo's method for ranking player skill is based entirely on wins vs losses. It was meant to be used with 1 vs 1 and fixed teams where winning was actually indicative of player/team skill.

In a game like MWO, where teams are randomly put together Elo does not work properly.

#22 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 22 January 2014 - 04:21 PM

View PostSandpit, on 22 January 2014 - 03:50 PM, said:

Isn't ELO based solely on the individual performance? If you do 600 damage an get 3 kills isn't that going to push your individual ELO up?

No. It's based solely on wins and losses and the relative ratings of the opponent you faced in those games.

#23 Corvus Antaka

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Knight Errant
  • Knight Errant
  • 8,310 posts
  • Twitch: Link
  • LocationInner Sphere

Posted 22 January 2014 - 04:33 PM

it doesnt work because not every match can be calculated by a computer & score to be a perfect thriller. further lack of tonnage limits in the equation make it impossible to give truly more balanced matches.

stomps happen. even the best players make mistakes, and every tactic can have a counter, a bad day, or a million other factors.

all things considered average match balance these days considering we don't have tonnage limits isn't bad at all.

Once tonnage limits come in it should be even better.

#24 Sug

    Member

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

Posted 22 January 2014 - 04:37 PM

View PostMischiefSC, on 22 January 2014 - 02:48 PM, said:

Please provide mathematical modeling to back up this point as it effectively implies that all statistical analysis and the entire field of statistical mathematics is incorrect. A bold assertion but it deserves to be backed up in some form.


View PostAlcom Isst, on 22 January 2014 - 03:02 PM, said:

Okay, only one of you can be correct. I have an 'F' in college probability and statistics, (an A- in Discrete Math though, so I'm not a complete failure at math) so far-be-it for me to judge whether or not MWO's ELO works.


You just can not take random players in random mechs, average out their Elo scores and then give rewards weighted by how likely the matchmaker thinks it is that you will win or lose. That just is not how Elo was meant to be used. On top of that we have organized groups allowed to drop against solo players skewing their Elo out of proportion to their skill.

View PostBhael Fire, on 22 January 2014 - 04:19 PM, said:

Arpad Elo's method for ranking player skill is based entirely on wins vs losses. It was meant to be used with 1 vs 1 and fixed teams where winning was actually indicative of player/team skill.

In a game like MWO, where teams are randomly put together Elo does not work properly.


ty yes.

Edited by Sug, 22 January 2014 - 04:38 PM.


#25 A banana in the tailpipe

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • 2,705 posts
  • Locationbehind your mech

Posted 22 January 2014 - 04:42 PM

There are some clan mates I consider not dropping with simply because their ELO is so high we're forced to go against 8+ syncdrops when we're just trying to duo.

#26 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 22 January 2014 - 04:44 PM

View PostColonel Pada Vinson, on 22 January 2014 - 04:33 PM, said:

it doesnt work because not every match can be calculated by a computer & score to be a perfect thriller.

The thing is, Elo is working just fine. It's providing ratings of player skill, and those ratings (assuming Elo is implemented correctly) get more accurate with every game played.

The problem is that people mistakenly believe that Elo is some magical system that can be used to create thrillingly awesome down-to-the-wire matches. That's just wrong. That's not what Elo does. Elo rates player skill only.

The matchmaker is what's screwed up. Your Elo rating is just one of many things that the matchmaker has available to help it create reasonably balanced matches. But it fails. Often dramatically.

Don't blame Elo when it fails, though. Elo is a proven system. Blame the matchmaker.

#27 Sandpit

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 17,419 posts
  • Facebook: Link
  • Twitter: Link
  • LocationArkansas

Posted 22 January 2014 - 04:44 PM

View PostBhael Fire, on 22 January 2014 - 04:19 PM, said:


Nope.

Arpad Elo's method for ranking player skill is based entirely on wins vs losses. It was meant to be used with 1 vs 1 and fixed teams where winning was actually indicative of player/team skill.

In a game like MWO, where teams are randomly put together Elo does not work properly.

View PostRoadkill, on 22 January 2014 - 04:21 PM, said:

No. It's based solely on wins and losses and the relative ratings of the opponent you faced in those games.

I'm not asking how the theory works, I'm asking how it's implemented here in MWO. ELO could easily be applied to an individual skill and effort. It's based on a 1v1 theory so that means it could be applied to the individual effort as opposed to the team effort and results.

#28 Khobai

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 23,969 posts

Posted 22 January 2014 - 04:45 PM

ELO should use match score instead of win/loss.

Someone with a high match score on the losing team should get more ELO points than someone with a low match score on the winning team.

It should average the match scores of all players in the game, then give you an ELO bonus/penalty based on how much higher/lower your match score was than the average.

Edited by Khobai, 22 January 2014 - 04:47 PM.


#29 Sandpit

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 17,419 posts
  • Facebook: Link
  • Twitter: Link
  • LocationArkansas

Posted 22 January 2014 - 04:49 PM

View PostRoadkill, on 22 January 2014 - 04:44 PM, said:

Blame the matchmaker.

OR

More players can start taking up responsibility for their own skill level instead of looking to blame any and everything for losing.

Other than the occasional tonnage discrepancy (which incidentally will be resolved once tonnage limits are implemented) the MM does a pretty good job of matching players by skill. Not once have I seen a single player jumping up and down about MM, ELO, premades, weapon balance, etc. say "Yea I was in a match last night and we got stomped. We got stomped because we played like {Scrap} as a team"

I DO, however, see a lot of those players talking about how they got 3 kills and did 900 damage and still lost. That shows me they're less concerned about teamwork and more about their personal KDR. never mind that they got that 3 kills and 900 damage because they waited in the rear launching lrms at random mechs (instead of focusing on a target), then rolled up after the match has become 12v1 and took out a couple of heavily damaged mechs (that their teammates had incidentally chewed up in the earlier melee) and then claim they were great while the rest of the team "sucked"

#30 A banana in the tailpipe

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • 2,705 posts
  • Locationbehind your mech

Posted 22 January 2014 - 04:49 PM

View PostRoadkill, on 22 January 2014 - 04:44 PM, said:

The thing is, Elo is working just fine. It's providing ratings of player skill, and those ratings (assuming Elo is implemented correctly) get more accurate with every game played.

The problem is that people mistakenly believe that Elo is some magical system that can be used to create thrillingly awesome down-to-the-wire matches. That's just wrong. That's not what Elo does. Elo rates player skill only.

The matchmaker is what's screwed up. Your Elo rating is just one of many things that the matchmaker has available to help it create reasonably balanced matches. But it fails. Often dramatically.

Don't blame Elo when it fails, though. Elo is a proven system. Blame the matchmaker.



One of these days I'd love for you to make a second account just to experience how bad the average joe has it when it comes to how ELO currently works. PGI decided to cater to the elitists again by speeding up their queue times and dropping them in with low ELO players. It's a blunder that will cost them fresh blood if not corrected soon.

#31 Sandpit

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 17,419 posts
  • Facebook: Link
  • Twitter: Link
  • LocationArkansas

Posted 22 January 2014 - 05:00 PM

I have YET to be stomped or win by stomp and see it as the MM's fault or a premade or any other reason listed. I have seen poor teamwork, joe derps derping out in the open and getting gunned down early on, I've seen rambos charge off and then spend the rest of the match moaning about how they would have done better if their team hadn't sucked, and I've seen lots of players that were just simply better than I am.

2-3 Joe derp/rambo players that die early on basically snowballs into a stomp usually. yet it's easier to blame PGI, MM, premades, etc. for that than look at the real causes

Is there a problem with the MM? Sure, there's 2 that I see.
1.) Tonnage discrepancies can be a royal pain sometimes but I don't see them often to be honest and the fix to that is coming in tonnage limits

2.) New players getting dropped in with vets. New players aren't going to do as well as vets in customized mechs regardless of ELO. ELO evens the field as far as player skill, not equipment. If they would simply put all new players into their own queue while they're earnign their cadet bonus and allow vetted and approved veterans to drop with them as "drill instructors" to help them out and offer advice, tips, information,etc. it would curb the noob stomp and help them learn the game faster.

If you're not a noob, you have customized mechs, and have been playing for a while? Chances are the stomp happened because you and your team simply did not play very well and I have no sympathy for you and don't see why PGI, MM, etc. should be blamed or changed just because you're not quite as good as you thought you were.

View Postlockwoodx, on 22 January 2014 - 04:49 PM, said:



One of these days I'd love for you to make a second account just to experience how bad the average joe has it when it comes to how ELO currently works. PGI decided to cater to the elitists again by speeding up their queue times and dropping them in with low ELO players. It's a blunder that will cost them fresh blood if not corrected soon.

uhm no sir, it was just posted that the ELO discrepancies you're talking about are purely fictional. The factual statistical data proves that wrong. The average joe isn't going to play against upper tier elo players. Fact.

Giving in to elitist.....

BWAHAHAHAHHA

If that were truly the case? There would never have been any limits on grouping up, 90% of the nerfs that have been implemented would never have been, and we'd be running around in mechs with TT value armor

#32 Thimble

    Member

  • PipPip
  • Philanthropist
  • 48 posts

Posted 22 January 2014 - 05:03 PM

I think Elo *can* work ok in MWO.

I do not think that it *is* working well. Primarily due to the scattershot matchmaker.

MM needs to do the following to get good data from a match (and thus be able to adjust Elo afterward):

*Balance based on mech composition... at the very least weight matching and should be some form of BV
*If matching solo vs group, inflate group Elo to account for increased ease of communication/coordination.
*Ensure the match is comprised of players from a relatively narrow Elo band.

If we had those 3 things, matches would be much better and I would have some confidence in its implementation.

Honestly, though, I think what we've got is as good as it gets from PGI.

#33 SuperNobody

    Member

  • PipPipPip
  • Veteran Founder
  • Veteran Founder
  • 84 posts

Posted 22 January 2014 - 05:12 PM

I don't think the implementation of Elo is terrible. It's much better than what we had without it. Sure, drop weight balance will help a lot, but Elo has overall been a positive change. I have quite a bit over 5000 games played and my W/L is 1.10. Ninety-five percent of that is solo pugging.

Rolls happen quite often, but I don't think you can always attribute that to Elo or even the weight balance being off. People do though because we don't have the data to either confirm or deny what exactly is happening so people are going to use any match results to support their own opinion. A lot of it is also confirmation bias.

I think Elo is doing it's job as well as it really can. I'd rather have it than not have it and hope that further improvements will come as drop weights are added and more changes in the future.

#34 SamsungNinja

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 224 posts

Posted 22 January 2014 - 05:17 PM

View PostAlcom Isst, on 22 January 2014 - 02:03 PM, said:

Wouldn't the sheer probability behind many games drive your elo up or down based on your skill? Luck can easily drive individual matches or a day of play, but over a long duration, your skill should have its influence on your Win/loss ratio.


This.

/thread

#35 MischiefSC

    Member

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

Posted 22 January 2014 - 05:17 PM

View PostSug, on 22 January 2014 - 04:37 PM, said:




You just can not take random players in random mechs, average out their Elo scores and then give rewards weighted by how likely the matchmaker thinks it is that you will win or lose. That just is not how Elo was meant to be used. On top of that we have organized groups allowed to drop against solo players skewing their Elo out of proportion to their skill.



ty yes.


You know I should probably not post this. Just got back from physical therapy and I'm on so many pain meds I am high as a ******* kite right now but I'm still going to respond.

I get that people want to believe that any time they lose it's not their fault. I get that. Totally get it. I get that when you see someone on your team do something stupid when you're spectating it's easy to draw the conclusion that 'if they'd just X, we'd have won!' That, however, has nothing to do with your Elo and I'll explain to you why.

Elo (not ELO, it's a last name) is not 'made for 1v1'. It's not 'made for' anything. It's a system for representing in a point value the results of a 1/0 (win/loss for example) contest. 1v1, 100 v 100, it doesn't matter. The only difference how many people involved in each match creates is how many total matches you need to play to get a comparatively accurate reading. To refine it more simply in a 12 v 12 environment you need to play 11 times as many matches to get the same results as 1 v 1.

The other 11 people on your team are irrelevant. Completely and totally and absolutely 100% irrelevant in the aggregate measure of win/loss as it relates to you.

Why you ask? That's impossible! They make you lose when otherwise you would win! It totally undercuts your performance! CUZ REASONS!

Every single other person playing MWO is in the same boat though. They are 100% as likely to be in the same boat as you. Over 100, 200, 300 matches those anecdotal events wash out because they happen to everyone.

What doesn't wash out is your performance. How much you help carry your team to a win. That's not reflected in damage or kills or assists, it can be good tactical recommendations. How well you support your other team instead of hanging back and kill-stealing. At the end of the day is your team more likely to win because of how you played. Every single other factor is inherently less reliable because it's contextual. Win/loss is not. You either help your team win or you don't.

Sometimes I take my Jag and throw on 2x LB10x and 4xSRM4s. When my team does well I can go in and pull a score of 120, 130 or so because I get a ton of component destructions, assists and the like and it's not hard to do 800+ damage. Yet I'm not really helping win, am I? I can take an M3 with 2 PPCs and lurk back sniping and get 4 or 5 kills with reasonable consistency but honestly most of those kills were people my team could have finished without my presence. Conversely I can play a D-DC and play human shield for my team, keep up and use my ECM to shelter them when they get attacked and always run back to help allies caught alone by lights. I can finish with ~300 damage, 0-2 kills and maybe 4 assists but have markedly improved my teams odds of winning.

Win/loss is all that matters. Everything else you do, damage, kills, it's ego-fapping.

Elo works, works correctly and with reasonable accuracy will reflect after a good 200 or 300 matches how likely your presence is to help your team win (or lose) a match against comparably skilled opponents.

Now, the matchmaker (which uses Elo to make matches) needs to have pug and premade Elo split and will provide better matches by matching everyone with the same range rather than matching high/low for a target but that's a separate issue.

Elo and measuring win/loss is the only worthwhile and reliable method of ranking players.

Here are some links on why math works and isn't an lie of THE MAN to just keep you down:
Statistics
Probability
Probability sampling

Here are some links on why you don't want to believe that it works:
Anecdotal experience

Confirmation bias

Self-serving bias

Persistence of discredited beliefs

Illusory Correlation

I totally understand that people want to believe that statistical mathematics are wrong or just somehow math doesn't count when it applies to this but it's not wrong and it does count. I could, with enough samples, model not just the odds of your behavior promoting a win or loss for your team but I could predict your political affiliation, religion, familial status (married/unmarried/living at home) and financial stability with enough samples of how you play MW:O.

Elo is accurate and about as correct as possible. Like with any sport it obviously can't predict in the immediacy your mood and likelihood of winning a specific game but taken as an average how likely having you there will promote a win over a loss against a given skill of adversary? Yes, it works like all other probability sampling and statistical analysis works.

Edited by MischiefSC, 22 January 2014 - 05:19 PM.


#36 Entail

    Member

  • PipPip
  • Ace Of Spades
  • 47 posts

Posted 22 January 2014 - 05:38 PM

View PostRussianWolf, on 22 January 2014 - 01:54 PM, said:

2) it isn't taking into account the quality of the win - did your team destroy 12 mechs to their 1?
3) doesn't take into account individual success - If I die in the first 2 seconds and my team goes on to win, I still get credit for the win. And if I rack up 9 kills and 1500 damage but can't carry us to the win, its still a loss.

Simple fact, giving an individual ELO rating in a game where you rate on a team metric simply doesn't work.

You would need to have set teams in order to use a team metric and get any accuracy in the rating, such as an actual league. This simply doesn't work for PUG matches at all.

OR

You need to change the metric to individual performance in order to get a rating that is actually accurate. This would work across the board


I agree, and along with individual performance should be an indirect rating/ranking system explicit to other players.
Because yes, several times I got 6 or more kills and lost lol, and it went against me (was on the team that lost).

Edited by Entail, 22 January 2014 - 05:38 PM.


#37 MischiefSC

    Member

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

Posted 22 January 2014 - 05:54 PM

No. Your value can and should only be tracked via win/loss. In the aggregate it's the only reliable way to determine how useful you are as a member of a team.

#38 Bhael Fire

    Banned - Cheating

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 4,002 posts
  • Twitter: Link
  • Twitch: Link
  • LocationThe Outback wastes of planet Outreach.

Posted 22 January 2014 - 05:55 PM

View PostSandpit, on 22 January 2014 - 04:44 PM, said:

I'm not asking how the theory works, I'm asking how it's implemented here in MWO. ELO could easily be applied to an individual skill and effort. It's based on a 1v1 theory so that means it could be applied to the individual effort as opposed to the team effort and results.


It's how it works in MWO; A player's Elo score is based on their wins and their losses only. Nothing else is factored in.

There's several posts floating about the forum that describe how it work in detail.

#39 Sandpit

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 17,419 posts
  • Facebook: Link
  • Twitter: Link
  • LocationArkansas

Posted 22 January 2014 - 06:01 PM

View PostBhael Fire, on 22 January 2014 - 05:55 PM, said:


It's how it works in MWO; A player's Elo score is based on their wins and their losses only. Nothing else is factored in.

There's several posts floating about the forum that describe how it work in detail.

Posts from the devs?

#40 Necromantion

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,193 posts
  • LocationBC, Canada

Posted 22 January 2014 - 06:03 PM

View PostRussianWolf, on 22 January 2014 - 01:54 PM, said:

I can drop in a hundred games (or a thousand) and my ELO rating is going to be based on how my team in those 100(0) matches comes out in the end. Wins and losses.

Problems:
1) it isn't taking into account the way you win - Cap rush win is a win even if a shot is never fired.
2) it isn't taking into account the quality of the win - did your team destroy 12 mechs to their 1?
3) doesn't take into account individual success - If I die in the first 2 seconds and my team goes on to win, I still get credit for the win. And if I rack up 9 kills and 1500 damage but can't carry us to the win, its still a loss.

Simple fact, giving an individual ELO rating in a game where you rate on a team metric simply doesn't work.

You would need to have set teams in order to use a team metric and get any accuracy in the rating, such as an actual league. This simply doesn't work for PUG matches at all.

OR

You need to change the metric to individual performance in order to get a rating that is actually accurate. This would work across the board



Yep and stop saying games are balanced cause "Elo of each team was within x amount of eachother" because all that often results in is the team with the more median players stomping the team with 2-3 really good players and the rest that are junk.

View PostBeliall, on 22 January 2014 - 02:01 PM, said:

a Win / Loss elo system is not suited well for any team based game. You can perform flawless on your end and then support your team the best you can and still end up in low elo ranking do to very inept team mates. Its irritating in other games as well in this one.



Yep constant 700-900 dmg games with 80% of the kills in the game at times but the games drop my Elo cause i lost. Whoopty do





29 user(s) are reading this topic

0 members, 29 guests, 0 anonymous users