99 replies to this topic

[Homeless Bill]

Grits N Gravy, on 19 September 2013 - 05:10 PM, said:

Actually the specifics are in there.

Spoiler

i.imgur.com/J5ZiG.png

There's no direct correlation to win rate and Elo scores not because one team is expected. In fact the system takes that into account and rewards you more points if you upset a better team. But because the Elo scoring has nothing to do with win and loss ratios. Elo does not equal win loss.

A person can end up with an Elo ranking of 2000 if they beat enough people with high ranks early in his career, due to the K factor. His K factor will eventually go away and he'll sit on a 50/50 win ratio at 2000 elo over a course of another 50 games. Meanwhile a player who maybe an 1400 Elo. Maybe he didn't quite pick up on things early on, but he's really developed a lot of skill, thus allowing to gain some elo points. So he might have a win/loss of 3/5. Because he's at the top of his bracket, he needs to keep up that win loss to stay there as his losses cost him more points than his wins generate, aka elo hell.

Can you start to see how there is no strong correlation to win loss and Elo rankings? Even if we looked at Elo after 20 games when K factors are still in play; People with the same win and loss will have wildly different Elo scores based on the competition they faced, as their Elo scores are derived directly from their opponents. Within the Elo scores it's reasonable to assume a totally random distribution of win loss, that trends to 50/50 the longer the person stays at that score.

I'll phrase it more clearly: there's no way we can do any of the calculations ourselves. I've poured over all their posts and fully understand how it works, but we're unable to do anything with the math they've given us.

Again, I still think that over a statistically significant period there will be a fairly narrow range of win-loss ratios for a given Elo ranking. As your Elo begins to set, I'd imagine you even get a very similar number of "expected to lose" and "expected to win" matches as those with a comparable Elo (because the matchmaker is trying to normalize good players down and bad players up).

Your examples are all fine and well for a tiny sample, but once you get 500+ matches, I'd bet money the win-loss ratio distribution for any given Elo looks a lot like a standard bell curve. Pretending there's no correlation between WLR and Elo is simply ridiculous; it's clearly not direct, but certainly you'd agree that someone with a 1.1 WLR is more likely to have a higher-than-average Elo than someone with a 0.9 WLR.

[Xyroc]

Grits N Gravy great write up. Made me stop and look to see what my win rate was and have to say quite happy ;D

[Robertthedread]

I pretty much think that what I do or don't do matters very little. The match was decided when it was put together cuz I can have a 800 pnt game and lose or I can have a 0 pt game and win. When I win I win 4 or 5 in a row. When I lose its 4 or 5 in a row. My win lose barely changes. Since it matters little what I do, I have loaded nothing by machine guns on a heavy. Didn't bring any weapons in one fight, but that was boring just running around trying not to get hit.

[Grits N Gravy]

Homeless Bill, on 19 September 2013 - 05:34 PM, said:

once you get 500+ matches, I'd bet money the win-loss ratio distribution for any given Elo looks a lot like a standard bell curve. Pretending there's no correlation between WLR and Elo is simply ridiculous; it's clearly not direct, but certainly you'd agree that someone with a 1.1 WLR is more likely to have a higher-than-average Elo than someone with a 0.9 WLR.

The more games played, the more likely that a person's win loss ratio will be 50%. So the there would be a Gaussian distribution of win loss ratios with a significantly large sample of games played. But, the mean win loss ratio would be 50%. The outliers would represent transient people passing through that specific Elo score, or people stuck in Elo hell. The larger the sample the strong correlation.

If you lock in your Elo after a large sample of games you should only play people of your same score. Thus giving you a 50/50 win rate. Elo measures the likely hood of one person beating another. If a large group of people have the same Elo score and a 60% win rates it would mean that matchmaking is broken by allowing to wide of a Elo variance between participants. Or, that the Elo system is broken by adding to many points for wins against lesser opponents. Anyone maintaining a win rate above 56% would be in actual Elo hell, meaning their elo score is to low.

Wins over losses does not determine Elo scores, therefore trying to correlate the two is worthless, cause there is no causal effect. At worst the sampling would be total random, at best the distribution would mimic prototypical distribution of skill aka Gaussian.

If you were to select any person at random, with any Elo score, with a statistically significant amount of games, you are overwhelming likely to select someone with a 50/50 win ratio.

If Elo scores are not determined by win loss ratios, there is no correlation between the two. If there not causal factors the distribution of win rates amongst Elo scores is going to likely distributed in a Gaussian manner. With 50/50 being the mean score, because the Elo system biases towards that as a function of it's design.

So if we took a sample of people with 90% win rates and 100% they would both be most likely to have have a score that is the mode of the aggregate scores. Which judging by the graph i posted is 1250-1260. Simply put a person win loss ratio doesn't give you enough information to determine their Elo score.

Your gut is telling you better people win more games, better people will have higher Elo scores. Therefore a person with a higher win to loss ratio will have a high Elo score. That is not the case because the purpose of Elo scoring matchmaking is create games with a 50/50 win chance. The first 30 or so games are going to be what determines your Elo score due to the K factor. After that your win loss should stabilizes in the 50's as you play people of equal skill. The nature of Elo score computation makes it very difficult and time consuming to raise your score more than 100 points on a long term basis.

Edited by Grits N Gravy, 20 September 2013 - 06:59 AM.

[MustrumRidcully]

AntiCitizenJuan, on 18 September 2013 - 07:51 PM, said:

This has been a very educational thread

Except it doesn't tell us about the steering wheel underhive.

[Joseph Mallan]

MaddMaxx, on 19 September 2013 - 09:14 AM, said:

Yes, but now I have to figure where the hell I fit in that Jungle. LOL!

P.S. Oh wait a minute. I am there already. It is Me and Kaffe.

Why try to fit in the Jungle when I'll just burn it down with some Napalm? Just to be sure!

[Mister Blastman]

MustrumRidcully, on 20 September 2013 - 06:57 AM, said:

Except it doesn't tell us about the steering wheel underhive.

They are very real! I have PROOF! SHH! Be quiet!



*whispers*

Here you see an Underhiver in his native habitat. His fingers firmly clutch the steering wheel out of sheer terror as battle wages around him. He knows deep down inside it is only a matter of moments before his chariot of junk is melted into a fine ingot of noobidium but it doesn't stop him. He charges headlong into battle, opening his side doors to shout at the enemy, without ever checking his radar or rear-view mirrors!

Of course, that's just this guy. Many of the underhive grip their wheels and cower behind cover indefinitely, afraid to even move. Most of them can be found hiding in the shadows behind tall rocks waiting for the inevitable. It's almost like there is relief in their eyes when they see you as they know their fear and suffering is almost over.

Be careful out there, guys. The underhive is everywhere. And stomp quietly, please. We don't want to add to their already heightened anxiety.

[MustrumRidcully]

Mister Blastman, on 20 September 2013 - 07:53 AM, said:

They are very real! I have PROOF! SHH! Be quiet!

This picture.. I've never seen this composition before, it's amazing, I ... I am at loss for words.

[Mister Blastman]

MustrumRidcully, on 20 September 2013 - 08:00 AM, said:

This picture.. I've never seen this composition before, it's amazing, I ... I am at loss for words.

You're welcome. It took me a little while to put together but it was oh so worth it. I can now make any variation of it I want.

[MustrumRidcully]

Mister Blastman, on 20 September 2013 - 08:01 AM, said:

You're welcome. It took me a little while to put together but it was oh so worth it. I can now make any variation of it I want.

Bring this to the meme thread and ask for captions.

[The Boz]

...I can imagine that sight as being the other eye cockpit of an Atlas

[Frisk]

MadCat02, on 18 September 2013 - 06:55 PM, said:

The difference between elo hell and elo heaven is only 10% win rate ( League of legends 900-2500 elo is 10% difference) but a hell of skill gap

What elo does is push people who persistenly try to win to a higher level of competition

The more you win , the more likely you are to loose!

So if it seems like you winning only 55% of the games elo really isn't broken . Infact 55% is considered decent

If i learned anything from playing LOL and DOta the only way to get to the top 2% of the community is to work very hard

People who cry about loosing give up or rage and don't actually learn anything

You are a factor in every game and that 10% win rate difference will reflect your performance .

I approve of this message.

[Chrithu]

Elo is a lie.

[Homeless Bill]

Grits N Gravy, on 20 September 2013 - 06:55 AM, said:

If you lock in your Elo after a large sample of games you should only play people of your same score. Thus giving you a 50/50 win rate. Elo measures the likely hood of one person beating another. If a large group of people have the same Elo score and a 60% win rates it would mean that matchmaking is broken by allowing to wide of a Elo variance between participants. Or, that the Elo system is broken by adding to many points for wins against lesser opponents. Anyone maintaining a win rate above 56% would be in actual Elo hell, meaning their elo score is to low.

If you were to select any person at random, with any Elo score, with a statistically significant amount of games, you are overwhelming likely to select someone with a 50/50 win ratio.

Your gut is telling you better people win more games, better people will have higher Elo scores. Therefore a person with a higher win to loss ratio will have a high Elo score. That is not the case because the purpose of Elo scoring matchmaking is create games with a 50/50 win chance. The first 30 or so games are going to be what determines your Elo score due to the K factor. After that your win loss should stabilizes in the 50's as you play people of equal skill. The nature of Elo score computation makes it very difficult and time consuming to raise your score more than 100 points on a long term basis.

I think that whole first paragraph is a lot of unbacked assumptions. You should only play people of a similar skill, but due to small the player pool and the matchmaker's lax standards, you'll see top-tier players in with complete scrubs. The Elo hell bit is what comes of it, but that's only because the playerbase / matchmaker can't find enough competent opponents. Even though anyone with a high win-loss ratio has an Elo that is "too low," it's still probably one of the higher Elos in the game.

Again, I entirely disagree. Though the way points are distributed based on assumed win/loss messes things up, I see no situation where maintaining a good WLR wouldn't get you an increasing Elo. The more you win, the more "assumed loss" matches the matchmaker will try to give you, the more "assumed loss" matches you win, the higher your score is pushed.

The way you talk about things, it seems like you think there are "tiers" of players that you move between, and those tiers are completely excluded from each other. That's not how it works in this game based on all empirical evidence. The matchmaker just tries to match up average Elo ratings - it doesn't set up a Tier 1 match and a Tier 2 match. We've all seen matches that have PEEFsmash, kaffeangst, and some guy that doesn't know how to set up his weapon groups. The longer it waits, the lazier it gets about finding matches.

Again, what should happen is irrelevant. What does happen is that good people continue to win more because there aren't enough other good players to normalize them downwards. I find it hard to imagine that anyone with a 1600+ Elo rating has below a 1.5 win-loss ratio.

I would also like to thank you for being competent, civil, and enjoyable to debate.

[Homeless Bill]

kaffe, what the hell is your win-loss ratio?

[Vodrin Thales]

It should also be noted that ELO is only really in full effect in the pug queue. 12 mans are a whole other ballgame and can greatly affect won-loss in either direction. My won-loss in pug matches seems pretty close to 50-50, but my overall is about 2 wins for every loss. Most of the difference is because I frequently will go 10-2 in an evening of 12 man play.

[kaffeangst]

@Homeless Bill

For my main mechs, or the ones I primarily use, my W/L is over 4.0 or 80% W/L. Overall W/L is slightly lower, but not by much. Honestly, the majority of my losses are from base caps or cap resources, not from team elimination.

From my highest point, six-seven months ago (or the introduction of ELO), my W/L has only dropped 5.75%. Most players have fallen considerably more than that - same with their other statistics. For example, my K/D ratio has only dropped 12% (from peak), whereas others' have lost 25-50% or more.

I've played MWO for an hour or two a day since Closed Beta, and my statistics have remained consistent throughout all the various balance and game-play changes.

[Homeless Bill]

kaffeangst, on 20 September 2013 - 12:44 PM, said:

@Homeless Bill

For my main mechs, or the ones I primarily use, my W/L is over 4.0 or 80% W/L. Overall W/L is slightly lower, but not by much. Honestly, the majority of my losses are from base caps or cap resources, not from team elimination.

From my highest point, six-seven months ago (or the introduction of ELO), my W/L has only dropped 5.75%. Most players have fallen considerably more than that - same with their other statistics. For example, my K/D ratio has only dropped 12% (from peak), whereas others' have lost 25-50% or more.

I've played MWO for an hour or two a day since Closed Beta, and my statistics have remained consistent throughout all the various balance and game-play changes.

Thanks for the info. It gives some perspective to hear from one of the best players in the game (based on my observations through ~5000 rounds). 4.0 is not as ridiculously high as I was expecting, and that makes me happy.

[MadCat02]

Robertthedread, on 19 September 2013 - 06:03 PM, said:

I pretty much think that what I do or don't do matters very little. The match was decided when it was put together cuz I can have a 800 pnt game and lose or I can have a 0 pt game and win. When I win I win 4 or 5 in a row. When I lose its 4 or 5 in a row. My win lose barely changes. Since it matters little what I do, I have loaded nothing by machine guns on a heavy. Didn't bring any weapons in one fight, but that was boring just running around trying not to get hit.

You can't know if the game will be close or not . If you want to win more games you have to give it your best everytime .

[Homeless Bill]

Robertthedread, on 19 September 2013 - 06:03 PM, said:

I pretty much think that what I do or don't do matters very little. The match was decided when it was put together cuz I can have a 800 pnt game and lose or I can have a 0 pt game and win. When I win I win 4 or 5 in a row. When I lose its 4 or 5 in a row. My win lose barely changes. Since it matters little what I do, I have loaded nothing by machine guns on a heavy. Didn't bring any weapons in one fight, but that was boring just running around trying not to get hit.

Be thankful. I wish I had more of those rounds; as it is, I feel like I have to carry or die. Even when I perform poorly, I'm almost always in the top four on my team. Showing up in fluffly Elo land when I bought my first medium was a dream come true for the first 25-or-so rounds.

I like that Elo grinds against you and makes it harder, but some nights I really wish I could just steamroll bads in a non-ranked queue. I'm not good enough to carry, but not bad enough to escape the responsibility. #firstworldproblems