103 replies to this topic

[MischiefSC]

First rule of the Gen Disc forum here - if the title isn't insulting nobody reads it.

So to the point though I've seen a lot of threads lately with people talking about their relative scores in win/loss, KDR and the like. I see in this a lot of misconceptions I thought I'd help (hopefully) clear up.

#1 - In Elo Rankings Quantity > Quality

If you've only played 50 matches and it wasn't 50 straight wins or 50 straight losses all in the same weight class then you probably haven't affected your ranking much at all. If you've got less than 100 or 200 drops in the same weight class (Light, Medium, Heavy, Assault) then you're probably still dropping at newbie Elo. If you've only dropped 10,20,30 matches in a chassis/weight class and wonder why your amazing 5.0 win/loss ratio still has you dropping with newbies it's because your 5.0 means NOTHING in that small a sample. Same with your KDR. You're dropping in n00btown still. Have a 5.0 with 300 drops in the chassis, that means you're rockstar Elo and deserve it.

500, 600 drops in a weight class will start to really show where you sit in Elo. Even at 200 or 300 it's possible that you've only shifted a bit in the rankings, because....

#2 - In Elo Not All Wins Or Losses Are Equal

If the Elo system predicts that you've dropped in a match where your team has a distinct advantage the win will mean little to your ranking. Conversely if you drop against opponents it figures will beat you then the loss means little. The converse is also true; winning against higher ranked people will net you a bigger improvement. Losing to lower ranked a bigger loss.

What does this mean? It means that you'll never know how much any win/loss affected you or that your actual win/loss ratio accurately reflects your Elo. If you drop with friends and you all have high Elo rankings you'll probably end up in some really, really tough games - except when the matchmaker has to drop you against inferior opponents due to lack of similar ranked folks around. So you may win a lot but those wins have a smaller total impact. Conversely if you have a low win/loss rate but most your losses are to people way above your skill and most your wins against comparable or even slightly higher ranked folks you may be GAINING in Elo even if your win/loss is 1.0 or even less.

#3 - Dropping In Premades Skews Your Results

Most the people you group with have a disparity of Elo scores. This means that while you'll drop against opponents of a similar average the impact of those wins/losses will affect you all differently. Is your Elo higher than your teammates? The wins will mean little to your score but the losses will cost you. You're going to be dropping against mostly lower Elo opponents. Are you the sandbagger on the team? When you guys win it'll help you a lot - when you lose it won't affect you as much. This is BAD. It means that when you pug you're going to be at an Elo score boosted by the help of your more skilled teammates.

TL;DR

If you haven't had hundreds of drops in a weight class (light/medium/heavy/assault) your win/loss, KDR mean pretty much NOTHING and you're probably still at n00bzor Elo

Winning/losing affects you based on your opponents. Your win/loss is NOT a directly accurate representation.

Dropping in premades will probably skew your Elo widely one direction or another and result in you being out of Elo balance for pugging.

Edited by MischiefSC, 01 April 2013 - 01:58 PM.

[Kin3ticX]

until it matches by class i consider it all broken

[AceTimberwolf]

Seems someone is Mad about their poor ELO Score LOL. Its nice up here at the top with many GG to go around. You should come join us.

[Royalewithcheese]

Telling an online multiplayer community that an objective measurement of their e-peen exists but will not be made available to them is a beautiful way to troll them.

Edited by Royalewithcheese, 01 April 2013 - 02:08 PM.

[deforce]

MischiefSC, on 01 April 2013 - 01:58 PM, said:

First rule of the Gen Disc forum here - if the title isn't insulting nobody reads it.

So to the point though I've seen a lot of threads lately with people talking about their relative scores in win/loss, KDR and the like. I see in this a lot of misconceptions I thought I'd help (hopefully) clear up.

#1 - In Elo Rankings Quantity > Quality

If you've only played 50 matches and it wasn't 50 straight wins or 50 straight losses all in the same weight class then you probably haven't affected your ranking much at all. If you've got less than 100 or 200 drops in the same weight class (Light, Medium, Heavy, Assault) then you're probably still dropping at newbie Elo. If you've only dropped 10,20,30 matches in a chassis/weight class and wonder why your amazing 5.0 win/loss ratio still has you dropping with newbies it's because your 5.0 means NOTHING in that small a sample. Same with your KDR. You're dropping in n00btown still. Have a 5.0 with 300 drops in the chassis, that means you're rockstar Elo and deserve it.

500, 600 drops in a weight class will start to really show where you sit in Elo. Even at 200 or 300 it's possible that you've only shifted a bit in the rankings, because....

#2 - In Elo Not All Wins Or Losses Are Equal

If the Elo system predicts that you've dropped in a match where your team has a distinct advantage the win will mean little to your ranking. Conversely if you drop against opponents it figures will beat you then the loss means little. The converse is also true; winning against higher ranked people will net you a bigger improvement. Losing to lower ranked a bigger loss.

What does this mean? It means that you'll never know how much any win/loss affected you or that your actual win/loss ratio accurately reflects your Elo. If you drop with friends and you all have high Elo rankings you'll probably end up in some really, really tough games - except when the matchmaker has to drop you against inferior opponents due to lack of similar ranked folks around. So you may win a lot but those wins have a smaller total impact. Conversely if you have a low win/loss rate but most your losses are to people way above your skill and most your wins against comparable or even slightly higher ranked folks you may be GAINING in Elo even if your win/loss is 1.0 or even less.

#3 - Dropping In Premades Skews Your Results

Most the people you group with have a disparity of Elo scores. This means that while you'll drop against opponents of a similar average the impact of those wins/losses will affect you all differently. Is your Elo higher than your teammates? The wins will mean little to your score but the losses will cost you. You're going to be dropping against mostly lower Elo opponents. Are you the sandbagger on the team? When you guys win it'll help you a lot - when you lose it won't affect you as much. This is BAD. It means that when you pug you're going to be at an Elo score boosted by the help of your more skilled teammates.

TL;DR

If you haven't had hundreds of drops in a weight class (light/medium/heavy/assault) your win/loss, KDR mean pretty much NOTHING and you're probably still at n00bzor Elo

Winning/losing affects you based on your opponents. Your win/loss is NOT a directly accurate representation.

Dropping in premades will probably skew your Elo widely one direction or another and result in you being out of Elo balance for pugging.

tl;dr

summary after reading your first sentence= my ELO>your ELO

[Roadbeer]

I find your theories intriguing and I would like to subscribe to your newsletter

[Vrekgar]

[AceTimberwolf]

deforce, on 01 April 2013 - 02:06 PM, said:

tl;dr

summary after reading your first sentence= my ELO>your ELO

Nah his ELO is pretty low since he only started playing in Janurary. Dont worry Deforce I'm sure ur higher

I'll take Experience over skill anyday

Edited by AceTimberwolf, 01 April 2013 - 02:11 PM.

[Spunkmeyer]

I'd be interested to see where you're gathering your data from, and evidence to back up your assertions.

[jay35]

http://mwomercs.com/...ost__p__2165198

[MischiefSC]

deforce, on 01 April 2013 - 02:06 PM, said:

tl;dr

summary after reading your first sentence= my ELO>your ELO

Hmm? No, I'm probably right about average. I pug almost exclusively and I'm far from a ninja player. If this were chess and assuming all my games were played with/against equal Elo scored people I'd only be at like a +29 or +30. That's not a lot given that your newbie score is 1300 if they're using a regular Elo model.

I do see however posts of people talking about their stellar win/loss or KDR and saying Elo is broken because they should be in the Golden Elite Tier because of it and post screenshots.... that show they've played like 20 matches in heavies. Great! Keep it up for another 100 or 200 matches and see where you are, that's going to be more accurate.

AceTimberwolf, on 01 April 2013 - 02:10 PM, said:

Nah his ELO is pretty low since he only started playing in Janurary. Dont worry Deforce I'm sure ur higher

I'll take Experience over skill anyday

The only score that counts is what's come out since they started seeding Elo in early February - so someone who's been playing a long time but very little since February will have seen *less* change in Elo than someone who started in February but has played 1,000 games since then.

Spunkmeyer, on 01 April 2013 - 02:10 PM, said:

I'd be interested to see where you're gathering your data from, and evidence to back up your assertions.

Sure! Here -
http://mwomercs.com/...-making-update/
http://mwomercs.com/...79-matchmaking/
http://en.wikipedia....o_rating_system

There have also been some Q&As with Garth and others about it implying that 1300 was the start score and saying pretty much exactly what I've said above. I've got premium time running so I can't go find them for you right now but the above should make it pretty clear.

[AceTimberwolf]

haha I'm not talking about his ELO just his experience. Welcome to MWO

[KitK]

Why troll the OP? Its a good post. It seems consistant with what Paul wrote up when ELO was released. It seems built on math and stats theory. Now that's not my strong suit, so maybe I missed something. But what was said makes sense on the surface.

[Vrekgar]

KitK, on 01 April 2013 - 02:28 PM, said:

Why troll the OP? Its a good post. It seems consistant with what Paul wrote up when ELO was released. It seems built on math and stats theory. Now that's not my strong suit, so maybe I missed something. But what was said makes sense on the surface.

Because NOBODY save the devs has any idea what their ELO score is or should be.

Also the Devs have stated multiple times that the system is not a True ELO implementation. Who knows but the devs what any win or loss will do to your stats. You also do have confirmation that each player has multiple ELO scores to account for Light, Medium, Heavy, and Assault mech classes.

Also there is little info on how playing in a group alters the ELO spread and how any win or loss changes your scores because of that.

Theres a running theory that the "Events" that we have had are veiled attempts to entice players to play as many solo matches as possible to help train and test the ELO system.

[neviu]

Cookies?

[MischiefSC]

Vrekgar, on 01 April 2013 - 02:34 PM, said:

Because NOBODY save the devs has any idea what their ELO score is or should be.

Also the Devs have stated multiple times that the system is not a True ELO implementation. Who knows but the devs what any win or loss will do to your stats. You also do have confirmation that each player has multiple ELO scores to account for Light, Medium, Heavy, and Assault mech classes.

Also there is little info on how playing in a group alters the ELO spread and how any win or loss changes your scores because of that.

Theres a running theory that the "Events" that we have had are veiled attempts to entice players to play as many solo matches as possible to help train and test the ELO system.

I have no idea how the exact details of your Elo is calculated, this is just based off of what the Devs have already said. They did say that it's an 'average' but that Elo is calculated 'individually'. Thus if you drop with a team your opponents Elo will be set off your team average - the impact on your Elo will be individual.

They have also said that at the moment (at least when they last spoke) win/loss was the only calculation, if that's changed since then they haven't said.

That doesn't seriously impact the original points though.

1 - Any sort of Elo system (they have confirmed that 1300 is start so point values seem at least vaguely similar) is going to require a lot of games to make movement in. 100 wins more than losses, for example, to see a significant move up the ladder. So if you play 300 games and win 200 but lose 100 you'd see a pretty huge improvement.

2 - If it's even vaguely based on Elo and has even the pretext of fairness your Elo impact will be adjusted based on your opponents. I believe Garth has already confirmed this.

3 - If your teams average Elo is used then your individual Elo will be affected like normal based on the Elo of who you beat or were beaten by.

Of course it's all speculation, but it's speculation based on statements from the Devs and models of how the system is supposed to work and has been implemented everywhere else.

[neviu]

My elo is over 9000 bite, me

Edited by neviu, 01 April 2013 - 02:48 PM.

[aniviron]

Good thread, OP actually knows how Elo works, and how to capitalize it. Tired of seeing tons of "omg elo" threads by people who don't know how it works.

[Ghogiel]

This is written by the guy who can't even into command chair dev posts.

[MischiefSC]

Ghogiel, on 01 April 2013 - 03:00 PM, said:

This is written by the guy who can't even into command chair dev posts.

Can't even into command chair dev posts....

Sorry man, you lost me. Parse that again?