Dice Analogy to ELO Hell

Comment below rating threshold, click here to show it.

DuffTime

Senior Member

06-14-2011

Quote:
Originally Posted by herdsheep View Post
for example look at this. on this page of this thread there is a person saying solo queue works (feed lag repeat). there two people saying it doesn't work (razzishi and vpdas). this is after 4 pages of discussion on it, there is no approaching consensus despite like half the people claiming to be some sort students/grads/professors/god-kings of the subject. o btw i didnt include dufftime cause he seems to just agree with whoever made the last point.
Lol. No, I'm just right in the middle.

I believe a good player will never go to Elo hell.

I don't buy into all the math and such as much as everyone else does, but I find it interesting to read what people are saying on the subject, and interesting to offer my two cents and see how people reply.

But to imply I just automatically agree, I don't. Mathematics unfortunately do not translate into this game well because individual skill is such a great factor.


Comment below rating threshold, click here to show it.

Ulanopo

This user has referred a friend to League of Legends, click for more information

Senior Member

06-14-2011

Quote:
Originally Posted by Razzishi View Post
If E(o) means expected value, that means that 'o' must be a random variable. In my work, random variables are always capitalized.
You have to remember that different fields can use the same symbols for wildly different purposes. Heck, physics is awful all by itself for this (Greek letter Mu). If I remember my statistics correctly, E stands for event. E sub 1(p) would then refer to the probability of the "first event."

In any event, we shouldn't even be talking about traditional statistics, because there are too many factors that can drown our data set. The Elo system will always be choppy with regard to team play unless it is the team that has the rating and the team never changes. Obviously, this cannot be the case

The only solution I can see is to reduce the impact on your rating of losses that are clearly due to forces outside your control. If you have a leaver, for example, losing should not cost you much, if anything, assuming the proper precautions are taken to prevent abuse (the leaver has to be gone >50% of the match, for example).


Comment below rating threshold, click here to show it.

ONLY MID THX

This user has referred a friend to League of Legends, click for more information

Junior Member

06-27-2011

Quote:
Originally Posted by vpdas View Post
This is a terrible justification for the poor match making. You can't win as a '6' if your '1's are enabling your enemy to win with their awesome 0 10 0 scores.
But if you're in a rating where people are "1," then so is the other team, and therefore you, being a "6," will be able to influence the game to your advantage.

Quote:
Originally Posted by DuffTime
People are getting stuck in the numbers game, when really a player's individual skill and good champion choices will have a huge impact.
People's individual skill and ability to pick are included in their skill in this analogy.

Quote:
Originally Posted by Loki the Wicked
Just because something is "oki" does not say it can't not be improved in 1000 ways or more.

The system is not worth mentionen as fair when you only gain lose ELO depending on the entire teams performance.

Elo would be 1000 times better if you instead got judged on your own performance after each game and not as a team.

Gain/lose differnt amounts of Elo depending on your performance.

I can come up wtih 1000 new ways to improve the game, but Riot is to focused on spitting out new champions and new skins..... saz alot where Riots focus is.
There are too many variables in this game that determine the victor. The only way to tell if you're having a positive impact is to look at your win/loss record. There are a lot of things you can do that will help your team to win that can't be tracked by the game.


Comment below rating threshold, click here to show it.

Tomego

This user has referred a friend to League of Legends, click for more information

Senior Member

06-27-2011

Very logical. Yes, matchmaking sucks at times but the game is run by RNG, logic driven not emotion.


Comment below rating threshold, click here to show it.

ASpaceOdyssey

Junior Member

06-28-2011

Quote:
Originally Posted by CyraLoL View Post
So… I'm guessing you have a high school education.

Anyways, if you ever actually study the subject, you'll know both what I'm talking about where my confidence comes from.

The fallacy that most people are suffering while reading my posts is that this is a perfect statistical system, which, simply put, it is not.

All the trolls are having fun with generalization I made in my first post which is a reference to a statistical probability called E(o). While I'm not going to try to explain it further than I already have because forum trolls that have at best taken high school stats think they are clever by linking wikipedia articles and talking about trolling and incorrect numerical theories…

That aside, do people realize why the Law of Large Numbers is flawed in this? At first I assumed that people were trolling, but I begin to wonder. First of all, the variance in this game is massive - the sheer number of x-factors in any given game is fairly staggering. I got a bit into this in a previous post, and I won't repeat myself but, for variance to play out to a significant degree we would literally be talking about more games than anyone in this thread has played times two. And between my accounts I've played thousands.

Second of all, do you understand what a bell curve is? This is very simple. While a majority may fall in a given region (less true in this case for the above listed reasons) no matter how many games some plays (short of an infinite number) they could very well never win a game. People two standard deviations form the norm are, in a game of millions of players, still incredible frequent.

Lastly, (now that we have established that there are cases in which people have been completely screwed over by MMR beyond a shadow of a doubt), the only thing left is to establish the frequency. Since few of the people posting here actually understand E(o) probability, lets simplify this a lot.

*Your impact on this game is being one player out of ten. The weight of your skill in this impacting odd varies depending if your playing tank, support, or carry - which is largely determined by the order in which you picked your champion, something which you have little control over.

*Than you have to factor in that you really cannot expect to be outside the top thirtieth percentile - anyone assuming better than that is probably letting their ego do the talking.

*Than you have to factor in the natural game weight to fifty-fifty created by the game purposefully imbalancing teams (to give purple - I believe it is - the advantage).

*Than you have to factor in the stronger tendency towards randomization of created by leavers. Assuming your never a leaver, this is a sixty-fourty.

*Than you have to factor in the odds that your playing against someone that similarly doesn't belong in the ELO bracket. This is much more complicated variable, but a particularly problematic one created by the large influx of new players at 1200 and the high random variance. This means the closer to 1200 you get, the more noise this variable creates.

*Than you have to factor in the odds of you personally having internet, real life, or computer troubles. While some people might suffer sever lag or wife/girlfriend/housemate/kid agro one out of ten or fifteen games, this is another variable that varies significantly. Not that while if you have to AFK you may deserve to lose, one might assume the infrequency with which this happens makes its impact minimal, I'm merely pointing out that its impact is cumulative with many other factors.

*Further, you have to acknowledge variance created by Riot's servers.

*Than some big game play factors. You have limited control of what your team bans and picks. There is a lot of situations where even if your team picks a more or less suitable team, the odds are decreased by the other team via planning or chance picking a team with actual synergy.

*Than you have to factor in the odds that if your an early pick, your team may wind up with a completely screwball composition. While its still possible to win with no AP or no tank or what not, it decreases the weight of your impact further.

*Additionally if you pick early and are the substantial weight of your team's chance to win are fairly hard counter picked without your team mitigating it, your weight on the odds is further decreased - this is particularly prevalent if you are forced by your teams picks to lane vs your counter pick.

*And quite a few more factors, but I'm getting bored.

While I including no numbers this time around as trolls that apparently have a high school education and wikipedia will decide themselves clever and realize that you can just say someone is wrong and on the internet this is considered an argument, I don't think anyone can seriously or realistically contest that all the things I listed above contribute to the situation, and that exactly one of them is your control as the player.

This alone tends to demonstrate that the amount of games for a single relatively small factor to show ANY weight at all is mind-blowingly small.

Of course, there are straight up numerical proofs that people would be stuck in Elo hell, but I'll save that for when you take a class in this or two.
With the approach of trying to break it down to really fundamental factors the sheer approach of putting a number as an equivalent of your theoretical performance in your next game is ridiculous. For the system to be correct you'd have a rating system with an infinite amount of factors just as it is in the real world. I don't really see the poinf of the discussion, ELO is broken beyond repair. It does not contain really important factors like emotional condition, server stability etc. just as you stated.


Comment below rating threshold, click here to show it.

Nightlynx

Senior Member

06-28-2011

Quote:
Originally Posted by Mormidon View Post
Where the hell did you learn probability from?
I think they're quoting some specific law in probability, can't remember which.

Its the one in which:
Flip an unbiased coin;
50% chance to heads and 50% chance to tails.

Flip a second unbiased coin;
Still 50% chance to heads and 50% chance to tails

The two coins are independent of each other, thus you have the same chance of getting heads/heads as you will of heads/tails as they do not affect each other regardless.


Did this thread just degenerate into a math discussion thread? o.0


Comment below rating threshold, click here to show it.

ONLY MID THX

This user has referred a friend to League of Legends, click for more information

Junior Member

07-15-2011

Quote:
Originally Posted by ASpaceOdyssey View Post
With the approach of trying to break it down to really fundamental factors the sheer approach of putting a number as an equivalent of your theoretical performance in your next game is ridiculous. For the system to be correct you'd have a rating system with an infinite amount of factors just as it is in the real world. I don't really see the poinf of the discussion, ELO is broken beyond repair. It does not contain really important factors like emotional condition, server stability etc. just as you stated.
And the fact that there are such a vast number of influences on the outcome of a game means that the only way to tell if you're an asset to your team is whether you win or lose more.


Comment below rating threshold, click here to show it.

Hasjustbeenpwned

Junior Member

07-15-2011

I think "ELO hell" comes from the idea that once you start loosing due to the 1's and 2's on your team it's hard to redeem your ELO, though like this post says, you will eventually and I find it very true.

Also, I find it more as ELO Purgatory, because *eventually* you will be able to get yourself to your ELO, it might just take time, I know I find it very frustrating as I mainly play support characters or tanks which causes me to have a rough time trying to "carry" a team who doesn't take advange of my initiation to adjust my ELO correctly, though eventually I'll be playing with my caliber of players and just have to wait for that day to come.

+1'd


Comment below rating threshold, click here to show it.

Ute

Senior Member

07-15-2011

I just want to say that cyralol is wrong about the 50/50 and 99/1 example. The odds of getting a SPECIFIC roll of 50/50 IE: hthhtththt.... is the same as getting any 1 of the 100 ways to get 99/1. There are, however, many thousands of thousands of ways to flip a coin 100x and get 50/50, while there are only 100 ways to flip the coin and get 99/1.

I only read that far into the post because I have to go, but this bugged me since he was getting thumbs up while misinforming. His point may otherwise be valid but that was untrue.

"Chance to get 50/50 is like 10^27 times more likely than 99/1." This is exactly right. To put it more simply for those statistically challenged:

For 100 coin flips, the probability of 50 heads/50 tails is 8%, while the probability of 99 heads/1 tail is 0.0000000000000000000000000008%.

The insults about high-school education should be stopped regardless, especially when you are so very wrong on a subject and have so many obvious spelling mistakes yourself.

Quote:
*Your impact on this game is being one player out of ten. The weight of your skill in this impacting odd varies depending if your playing tank, support, or carry - which is largely determined by the order in which you picked your champion, something which you have little control over.

You can pick whatever type of champ you want, at whatever pick you want. Picking to help bad players will not advance you. Pick to advance yourself if you are "better than your Elo and stuck in Elo hell."

*Than you have to factor in that you really cannot expect to be outside the top thirtieth percentile - anyone assuming better than that is probably letting their ego do the talking.

FYI, THEN* Aside from that, what would the thirtieth percentile matter? Is that a magic number where you can start to carry any game at low Elo?

*Than you have to factor in the natural game weight to fifty-fifty created by the game purposefully imbalancing teams (to give purple - I believe it is - the advantage).

The system doesnt intentionally create unbalanced teams, it is simply impossible to create perfectly fair teams. In any case, EVERY game is skewed in your favour because you are higher than your Elo says, right?

*Than you have to factor in the stronger tendency towards randomization of created by leavers. Assuming your never a leaver, this is a sixty-fourty.

fifty-forty, another stat that is aimed at pushing those who deserve it higher in Elo

*Than you have to factor in the odds that your playing against someone that similarly doesn't belong in the ELO bracket. This is much more complicated variable, but a particularly problematic one created by the large influx of new players at 1200 and the high random variance. This means the closer to 1200 you get, the more noise this variable creates.

This factor is based on the assumption the system is already flawed as far as placement matches go. If the system does work and places players near their true Elo, this creates minimal noise in the equation or even no noise at all.

*Than you have to factor in the odds of you personally having internet, real life, or computer troubles. While some people might suffer sever lag or wife/girlfriend/housemate/kid agro one out of ten or fifteen games, this is another variable that varies significantly. Not that while if you have to AFK you may deserve to lose, one might assume the infrequency with which this happens makes its impact minimal, I'm merely pointing out that its impact is cumulative with many other factors.

This factor is silly, every single person p[laying has this, and this is a direct link to your Elo. If you are a person who has to leave often, you will always have a lower Elo. Elo isnt only based on skill, it is based on everything that happens from the second picks start to the second the game ends, including attitudes, skill and distractions among many other things.

*Further, you have to acknowledge variance created by Riot's servers.

This would affect everyone equally, so it can be ignored entirely.

*Than some big game play factors. You have limited control of what your team bans and picks. There is a lot of situations where even if your team picks a more or less suitable team, the odds are decreased by the other team via planning or chance picking a team with actual synergy.

Both teams have similar odds of this happening, except we know for sure that you belong at a higher Elo, so that 20% of the time you get to pick bans or counter-pick you are guaranteed to do it right, and hence have a higher chance yet again of succeeding. This is another point against your entire argument

*Than you have to factor in the odds that if your an early pick, your team may wind up with a completely screwball composition. While its still possible to win with no AP or no tank or what not, it decreases the weight of your impact further.\

This guarantees that your first pick will be the best one in the game, and it guarantees you remove the best champion from the enemy. Outside of your pick, both teams have the exact same chance of picking poorly, so yet again this is only favourable to your Elo.

*Additionally if you pick early and are the substantial weight of your team's chance to win are fairly hard counter picked without your team mitigating it, your weight on the odds is further decreased - this is particularly prevalent if you are forced by your teams picks to lane vs your counter pick.

If the enemy is competent enough to counter pick you, your team is competent enough to counter pick them back. Moot point again except you got the champion you wanted to play and are best at. Yet again another force which should push you higher in Elo.

*And quite a few more factors, but I'm getting bored.

I dont think you made a single point here.
My response to your first point, which I feel important to expand on:
"You can pick whatever type of champ you want, at whatever pick you want. Picking to help bad players will not advance you. Pick to advance yourself if you are "better than your Elo and stuck in Elo hell."

Elo is not only based on skill, it is based on ability to win. You could be a 2500 Elo carry and then pick supports every game in low Elo because your team never does, and lose every game because you are a 600 Elo player while on supports. granted you are skilled, but you are too dumb to use your skill. Hence, you dont deserve that 2500 Elo, you deserve the 600.

If you really feel Elo hell exists check my past game histories, 9 wins 1 loss in ranked. I am running 70+% wins in ranked after dodging to sub 800 Elo, and now Im back near 1200 after only a couple of days.