Quote:

**isobold**:

1) This affects also players that were on their Elo for the most time and just dropped through a streak. To be precise: this is the most common cause for being in this 1% section, because the odds to be in this section decrease exponentially with the length of a streak involved that is needed to keep a player there.

Quote Lyte where he said that, please.

Quote:

**isobold**:

2) Players aren't talking about Elo-Hell complaining to be 125 Elo from their True-Elo. People tend to complain about being several hundred Elo away from their true Elo. But the amount of players effectively hit by such event decreases exponentially with the Elo gap involved.

From my experience, players tend to consider a particular elo range (that being 1200 to 1400) to be elo hell, given that it has a very large distribution of skill level between players.

Quote:

**isobold**:

So the number of actual players being a) more than 100 Elo away from their Elo and b) being there since they started playing isn't 1%. When we talk about players playing 300 games and more, the number of players affected is pretty sure 0.

"Pretty sure", also known as "I'm going to make up numbers to support my point and assert that they're true without providing any of the supporting evidence I would normally have to provide to justify that claim".

Quote:

**isobold**:

The 1% Lyte is referring to aren't representing Elo Hell.

No, but they're representing people who aren't at their True Elo. My question has always and forever been what I asked in my original post: how many people are there and by what magnitude are they affected?

Quote:

**isobold**:

So you say you got a single counter-example, but you don't feel like showing it.

I provided it in the post you quoted.

Quote:

**isobold**:

Fun fact: since my mathematic proof only works for an infinite number of games, I'm pretty sure you are just a bragger.

Fun fact: people don't play an infinite number of LoL games, therefore your mathematical proof is irrelevant to this discussion.

Quote:

**isobold**:

Unless I'm talking to god here, you won't be able to present a single example of a player having played an infinite number of games.

Fun fact: I don't need to, because I'm providing a counter-example to a

__mathematical proof__. We are discussing mathematical theory and therefore theoretical examples suffice. Please go back to school and relearn the definition of "counter-example" as compared to "counter-evidence".

Quote:

**isobold**:

Also you not understanding the mathematic proof that's behind the comic doesn't mean there is none.

I assumed you had a mathematical proof that was relevant, i.e. for a finite number of games, preferably at most 2000.

The 314+ page thread? No thanks. I'd like to save myself 20 hours of work if you don't mind. How about, since you posted it, you go quote it?

Quote:

**isobold**:

I've already said several times that I keep the entire ladder of all servers and queues for absolutelegends since more than a year. It's the base of the Elo-Graph. One is able to run statistic analyses on this data. None of the data I have so far suggests that Elo would be oscillating the way it would, if Elo-Hell could exist. I hereby repeat my offer I gave to readers of this thread previously to let you do all the analyses you want to run on the data to prove me wrong.

I have no idea where to get this data you supposedly have. Is it on

http://www.absolutelegends.eu somewhere? If so, where?

My requests are quite simple:

Provide me the mean and standard deviation of the number of trolls (by whatever metric you want to measure what a "troll" entails, preferably realistic though) per game at a particular elo (say 1200).

Then provide me with the approximate impact said troll has on their sides chance of victory, as well as the standard deviation of this chance.

Then provide me the mean number of players per game who are not at their true elo at that particular elo, as well as their standard deviations. Both players better than that elo, and players worse than that elo.

Finally, provide me with the approximate impact said player has on their sides chance of victory, as well as the standard deviation of this value.

If you can't, then all your data means jack **** to me because it doesn't actually address the problems I presented.

Quote:

**isobold**:

If there was a single one that failed, Elo Hell fanatics would have presented it the day it exists.

Maybe they did and you just missed it? Threads don't live forever bud.

Quote:

**isobold**:

I never said otherwise. But I don't actually care for statistics or experiments, I only did those to bolster the mathematic claims I made, since people who are bad at math tend to not trust math ...

You prefer mathematical proofs that rely on ridiculous assumptions like an infinite number of games played to actual statistical analysis? No wonder you came to the conclusion you did. Smells like confirmation bias in full effect here to me.

Quote:

**isobold**:

So what's your base to say player X is in Elo-Hell then?

Elo-hell is a loosely defined concept that's based on people's subjective evaluation of their own skill and the skills and intentions of their teammates. I can't stick a definitive number on such a broad concept since there are of course many different scenarios that could arise where a player justly and rightly concludes that his teammates cause him to lose more than he wins.

I will say that being 100 elo less than your true elo after 100 games is a problem if you aren't a serious player, particularly if your current elo is at or around the entry point (1200-1400 elo). That's the range where you're likely to see the greatest deviation in player skill and thus the biggest perception that your teammates are holding you back.

Quote:

**isobold**:

As for the errors in the player base I already told you to do the math yourself and see for yourself how small the number of affected players actually is.

And I already explained how problematic such an analysis is, something you apparently decided not to quote or respond to.

In short, the analysis of deducing the reliability of elo of a particular person depends on the reliability of elo of their teammates.

Quote:

**isobold**:

You misunderstood what I wanted to say. Forgive me if it was my fault, English isn't my mother tongue. What I meant is: the longer the streak needed, the less likely it is going to happen. If you need a streak of 300 games bad luck, you most likely won't see it happen once on the entire player base of LoL.

You don't actually need a streak of 300 games of bad luck. You just need your occasional bad luck of being matched with a troll not counteracted by good luck where the other team has a troll, and the resulting downward pressure to meet or exceed the upward pressure that results from you being better than your opponent.

The example I provided in the original post thoroughly outlined how this could happen. But just to provide another: if you lose on the net 8% of games you should have won due to trolls and bad players, and win on average 2% of games you should have lost due to trolls and bad players over the course of 400 games, then at any elo where your chance to win should be 53% because you're better than the players around you, your actual chance to win will be around 50%, i.e you will tend to stay at that elo even though you are slightly better than the people around you. That is because increasing in elo will lower your effective chance to win and decreasing in elo will tend to raise it.

After 400 games of a 53% expected and 50% actual (assuming you started at 1300 elo) the resulting elo displacement would be close to 100. That leads to a pretty negative experience if your elo is 1300 when it should be 1400, as there is a drastic difference between the quality of player in those two elos.

Quote:

**isobold**:

Of course, because that's the actual reasoning behind Elo-Hell for most players. More intelligent players who don't refer to trolls spiraling them down most of the time understand the mathematics why Elo-Hell can't exist and understand the difference between a streak of bad luck that's limited in time and an actual Elo Hell that keeps you down several hundred Elo below your true Elo for-ever (or even long enough to feel like forever).

Creating a false dichotomy that paints all people who support you as scholarly intellects and all who oppose you as near-drooling tards is pathetic. Don't try to pull that **** on me again.

Quote:

**isobold**:

I tell you that from a mathematical point of view, it's pretty unlikely to be kept for example 500 Elo below your true Elo for 300 games.

That's not a mathematical point of view, it's a statistical point of view. It relies on data like standard deviations and means. You can't make a statement like "x is unlikely" and still be talking mathematics, because strictly mathematical proofs aren't concerned with statistical evidence like the probability of x occuring.

Quote:

**isobold**:

You don't need any data to do those maths. Just assume the worst case that every single game is decided entirely by external influences and go from there. Real world numbers are going to be less catastrophic and thus supporting my reasoning even more.

Okay, worst case scenario is 100% of games are decided by trolls, afks, dcs, and the presence of feeders whether intentional or not.

Worst case scenario is that 100% of the time you get 4 feeders, trolls, afks, dcs, etc. on your team and your opponents have none for every single game you play from now till you stop playing.

As a result, your elo should drop to 0 and stay there.

What exactly was I supposed to conclude after making a worst-case analysis here?

Quote:

**isobold**:

Wrong. If you think that, you haven't understood the concept of infinity. Any streak is mathematically guaranteed to end.

No, actually it's not. A streak can theoretically go on for infinity.

Quote:

**isobold**:

No, they are NOT even. Read and understand the comic. It's really simple. I dumped it down so that sixth-graders can understand it. The odds for ANY external event, by it advantageous or detrimental to the outcome is happening in 5 out of 9 cases in the enemy team and only in 4 out of 9 cases in your own team. This is the mathematical prove that Elo will eventually converge, which also is the mathematical prove for the nonexistence of an Elo-Hell.

I read and acknowledged your comic and it was a waste of time. I already addressed this **** in my very first post on the topic about how averages alone aren't really that relevant to discussions of statistics and probability over finite time periods. Note though that I said "about even", implying that they are close to the same. 4/9 is about the same as 5/9: they're both 50% +/- 5.555...%.

Quote:

**isobold**:

Elo-Hell doesn't require you to spiral down-ward, it's enough to stay at a lower Elo. But what it requires is to persist into eternity. Anything else isn't the Elo-Hell I'm discussing here.

All it requires is that it persists until you stop playing the game, which could be after just 100 ranked games if you have a handful of bad experiences.

Quote:

**isobold**:

I do of course also discuss the odds for a player to get stuck at a lower Elo with players, but that's a different topic.

...What? I'll excuse you because you claim you're not an english speaker, but...

That's actually the topic I was discussing the whole time.

Quote:

**isobold**:

But the mathematical prove for the nonexistence of Elo Hell only applies to eternity and I never said otherwise. Even the comic I linked you doesn't state or imply anything different.

You sure as hell implied your mathematical proof applied to finite time periods when you claimed it had any relevance to this debate.

Quote:

**isobold**:

No, that's actually not interesting, since twice as much doesn't return an actual usefull result. 2 for example is twice as much as 1. So if you got trolled twice in 300 games, this made you lose 12 Elo on average. If having lost 12 Elo in 300 games is Elo-Hell, I don't mind it exists ...

It's more interesting than the irrelevant strawman example that I was quoting, because it can then be related to an average and used to derive the approximate deviation from your current elo like you just did.

Quote:

**isobold**:

What's actually interesting are the odds of getting trolled down several 100 Elo and then getting trolled significantly often and more than your enemies to be kept down for 300 games.** I admit that this is bit more difficult to calculate**

Well holy ****. Isn't that what I've been saying the whole time?

Looks like you agree that "just do the math" is a bull**** response, because as you admit it's a but more difficult to calculate.

Apparently the only time you can even conceive of the argument

**I've been making the whole damn time** is when you're chastising me for being too simplistic with one of my responses.

Quote:

**isobold**:

but we can make some assumptions, to make the calculations possible without using any data. Just assume 100% of those 300 games were decided by trolls and thus the player has no influence on the outcome. This is the worst case scenario so we don't lose anything by this assumption. In real world even less players will be affected than what you will calculate with this assumption.

So you're saying that if I assume that:

a) all of the 300 games a particular player participated in were decided by the presence of a troll on either team, and

b) that for a particular player, the chances that the troll was on his team was twice as likely as the chance that they troll was on his opponents team,

that I will be surprised by the results?

No, I'm actually not surprised at all. The persons win rate is on average 100 wins to 200 losses and their elo would be around 570. Given that said elo is derived 100% from pure luck, obviously the elo is completely wrong in the majority of cases.

Wasn't surprised one bit.

Quote:

**isobold**:

Since you didn't know how to build a worst case scenario and go on from there, I could say the same about you. But on the other hand I don't remember you to claim any education, so I won't play that card ...

FYI, I didn't build a worst-case scenario because I have more realistic scenarios that also support my position, which I presented in the post you were responding to as a counter example.

Quote:

**isobold**:

I didn't want to imply that we are talking about the same order of magnitude. What I wanted to imply is that both results are so small, they are way out of human imagination.

That's your opinion. Based on some confirmation bias-ridden experiments, I bet, and a mathematical proof that doesn't really apply to the issue at hand.

Quote:

**isobold**:

They are relevant to answer the question if them holding true are to be expected in reality ...

They aren't. It's simply easier to calculate the resulting effects that trolls have on win rates if certain assumptions are made. Adding the additional caveat that the presence of a troll doesn't predetermine the outcome but rather shifts the percentage likelihood of winning by some amount adds unnecessary complexity to a counter-example.

Quote:

**isobold**:

It does matter, since if the real deviation is to small to bring numbers below 50%, your entire reasoning breaks down.

My reasoning doesn't break down due to empirical evidence because it's reasoning, and therefore theoretical. But besides that, my original argument was that there exist some deviation between real and true caused by a normal distribution of trolls, not that the deviation is by necessity significant.

For example, if you have a 50.05% chance of winning based on your real to expected elo, it doesn't take a very strong deviation of trolls to knock that number below 50%. Now, of course the resulting elo displacement isn't very significant, but like I said, I was never arguing that at all.

I was merely arguing that you should expect that there be at some level a deviation (this is mathematically true by the way unless you assume that there is zero deviation in trolls) from your true elo and that it would take statistical analysis on data I don't have to compute the actual effects of it. To determine the true significance of the deviation would require data I don't have. It could be 50 elo, or 5 elo, or 200 elo, or 2 elo over the course of 300 games. I don't know.

Quote:

**isobold**:

No, as the odds for a troll are higher in the enemy team, an increased number of events means decreased odds of events that the sum of the events is at your disadvantage.

That's true on average but not necessarily for individual cases. You would have to correlate an increase in games played with a decrease standard deviation of trolls, which you have yet to do.