Seventh Lion2 votesI agree.. people that afk should have a higher LP lose, and people who have to deal with the loss should have a smaller impact on their LP.
Cause even if youre a skilled player, its hard to control the way a game can go when theres an afk or more. But then again theres times where my game crashes, or my internet stop working due to high ping etc. Not like I join a game just to leave, but there are players who do that. But you still have to put in perspective that there are thousands of people who may just have a lower end computer, or internet, or their game just randomly crashes for no given reason. So that would not be far for those who can`t control what their system does.
hh the thinker1 voteslosing team just needs one person to afk so they don't lose out and this can be abused.
AMYS GRAVE1 voteswhat if it still counted as a loss for the afk with perhaps reduced penalties for their team mates instead of complete forgiveness? i suppose to balance this out there would have to be reduced rewards for winning, but if you are winning a 5v4 that probably should be the case anyway.
ValyrianBlade1 votesShould win prevented be given when an opponent afks?
What's to stop people afk'ing as soon as things turn South?
The general idea is, given you don't afk, you have a 4/9 chance of afk on your team and 5/9 chance of it being the other team. So you actually get more wins due to afks than losses, removing the lp loss on an afk loss would just tilt the ladder really hard.
AMYS GRAVE1 voteshopefully a well implemented reporting system or leaver buster would stop people from intentionally afk-ing... but maybe thats a dream...
for the community as a whole people will move up because of afks because of this 4/9 5/9 chances of it being on your / the enemy team, but it does not have to balance out for an individual. if we just look at games with one afk (which excludes normal 5v5 games or server issues where multiple people are kicked etc...) and with 27 million players daily, statistically speaking there are around 3 million players who experience 2/3 of afks on their team which makes getting promoted nearly impossible and on the extreme there is one poor bastard who has had 17 afks on his team with none on the other team.
EDITS: if you are interested in the math behind this i try to explain it a bit better below, but i figured the average reader(no pun intended) wouldn't be that interested. http://en.wikipedia.org/wiki/Binomial_distribution http://en.wikipedia.org/wiki/Gambler%27s_fallacy http://en.wikipedia.org/wiki/Gambler%27s_ruin http://xkcd.com/356/
ValyrianBlade2 votesWhere are you getting your numbers from? Are these real numbers from over season 4's total ranked community?
I ask because, for any individual, for each game with 1 afk they have this 4/9 vs 5/9 chance. By playing more games, you likely increase the number of games with an afk present. Since the 4/9 and 5/9 don't change, then sure if you play 17 Games with an afk you individually could end up with anywhere from 0 to all 17 having it on your team, although it would have an expected number of 4/9 × 17. If you played 9000 Games though, you'd again have from 0 to 9000 Games with the afk on your team, but as the number increases the probability of having more afk against rather than on your team increases dramatically. So long term even for an individual, you'll benefit from others afk'ing, unless of course you exhibit behavior that causes afk's (e.g. toxicity).
AMYS GRAVE1 votesi'm just doing the math. actually to get 17 afks on your team the odds are (4/9)^17 which is ~1*10^-6 or one in a million. with 27million players that means that around 27 people have had this streak happen to them in a given 17 game series with afks (I was imprecise with my wording previously, forgive me). when you get a lot of people playing you get weird statistical fluctuations like this. that doesn't mean that this is necessarily permanent, but to suggest that this losing streak will go away to a particular player and they will now switch to having good luck and the afks will balance out the enemies team is just the gamblers fallacy. it does not have to balance out for an individual, only the collection of all players. it is true that the average player will have the average outcome, but a 1:27 million chance will happen about every day.
ValyrianBlade2 votesThe gamblers ruin problem suggests there is a lower limit that you are eliminated when you hit it. In this case, you can't get kicked out of league for sucking, so it doesn't apply.
The gamblers ruin theory also shows that you must have at least 50% chance of losing to guarantee ruin if you keep playing, since you have over 50% chance of benefiting, one player will with certainty eventually come out ahead. Granted, it may take 16 billion years for a certain person, just statistically you'll come out ahead.
You're also ignoring that most games don't have an afk (at least not an unintentional one, if a teammate rage quits that can be considered part of your losing the game).
The probabilities are quite straightforward to be honest - use any software to determine the probability a BIN (n, 5/9) distribution results in less than n/2 for various n. The value will tend towards 0 as n increases.
AMYS GRAVE1 votesHello, FYI i majored in math and physics and now am in graduate school for high energy physics specializing in statistical analysis. I hope i don't come off too dismissive, that's not my intent, i just hope to explain my original response more clearly. :)
the gamblers fallacy still applies, the heart of the fallacy is that the next independent event is influenced by the previous independent event. the actual fallacy has nothing to do with elimination at all, though it is often colloquially expressed in this way.
you can calculate the odds for an individual, but you can't assume that all people will eventually be brought to the mean. the mean of all player experiences will tend toward a central value and become more accurate with more trials but the variance np(1-p) will grow as you consider more players. a binomial with large n tends toward a Gaussian distribution, not a delta function. so there will still be losing streaks and people who do not come out ahead, even in the long run.
You're also ignoring that most games don't have an afk
well i actually said
if we just look at games with one afk
so that's not to say that the one poor bastard has had 17 games in a row with an afk, it is just that in the past 17 games he played that contained an afk (that presumably wasn't him), he has gotten the short end of the stick in all 17. a subtle but important point, sorry if i didn't make that clearer.
ValyrianBlade2 votesI'm glad you mentioned your major, so I can go in to actual math instead of simplifying things as I did last time.
First, I'll apologize for misreading gamblers ruin instead of the gamblers fallacy - although I'll comment that we're in a case of the former rather than the latter.
My comments on the binomial distribution weren't to say that for large n it's impossible to attain k < n/2, rather as I said it was to show that the probability diminishes rapidly with large n. The comment on 16 billion years demonstrates the difference between my theoretical University and your more practical learning (via physics) - I'm branching on to the concept of infinite. Regardless if you are 0 out of a million trials, another 9 million trials would have expectation of 5m successes and 4m failures so that you would break even if you beat the expectation. If not, you apply another 90m trials, then 900m, etc... No matter how bad your luck is, I have an infinite number of chances to succeed on a 50-50 coin flip - I'll win eventually.
Anyways, explaining my previous post aside, time for the real content - this is a renewal theory problem (often known as the gamblers ruin).
The most basic version of the problem is set up as a random walk: starting at 0 you move up 1 with probability p and down 1 with probability (1-p). The walk is finite in the simplest version, and you find out the probability of returning to 0 exactly 0, 1, 2, etc... times for any value of p. The next step is an infinite walk, where the same applies but the walk goes on forever. Again, for given p the interesting problem is to determine the probability of returning to 0 a certain number of times before the walk wanders off to positive or negative infinite. The gamblers ruin problem is the next step. It basically says you start at a certain amount of money, say $k, and move up $1 with probability p and down $1 with probability (1-p). You then find the probability of ruin for various p, I.e. The probability that the gambler ends up at $0. Typically you consider p <0.5 and the gambler doesn't quit, so he inevitably eventually loses everything and ruin ($0) is guaranteed. Next steps involve introducing an upper limit that the gambler decides to quit at, and given k and p you determine the probability of ruin.
Anyways, back to this particular problem, I stated that gamblers ruin does not apply - which is true because there is no lower limit, you can't reach a ruin of $0 because there is no $0. Rather, we're in the second case I mentioned in this analysis, an infinite random walk. With a simple random walk (p = 1/2) it has been proven that given infinite time every value will be crossed an infinite number of times. Thus, even with a 50% chance of the afk being on either team, you're GUARANTEED to not only break even, to not only reach breakeven an infinite number of times thereafter, but also to reach any arbitrarily (but pre defined) large number of extra benefits and extra disadvantages of afks. The flaw being that you have to let it go on for an infinite number of trials (ain't nobody got time for that! ). Alas, I am speaking in terms of infinite (again, see my 16 billion years comment as 'waiting forever') - this isn't wrong as you can always play more games. Anyways, from that point, it should be obvious to see what happens when we increase p to > 0.5, or in this case p= 0.55556. If not, we can look simply at the problem where you want to get ahead. I.e. starting at 0, what is the probability that within an infinite number of steps we'll reach +1? The solution to this is {1 if p> 1-p, p/(1-p) otherwise}. Since p > 0.5, p > 1-p, and thus the probability is 1. In other words, given enough time you are guaranteed to come out ahead - exactly the point I made in the first place.
AMYS GRAVE1 votesOK interesting, it seems we are viewing the problem from different perspectives then.
an individual playing a large number of games vs. a large community playing a smaller set of games each.
if every daily player plays 5 games a day every day for 80 years they get to fewer than 150k games each. the leaver rate on NA and EUW seems to be less than 1% so which means that the most leavers people will possibly see is ~1500. i don't think this sampling is large enough to say with certainty that every player will even see the break even point in their lifetime, let alone to dismiss this as a factor (as many people do) in an individual's struggle to climb ranks.
upon thinking about gamblers ruin vs. gamblers fallacy i think both are commonly used in the context of afks (even if not directly related to the problem we are doing). the gamblers fallacy is often employed to cheer up people who have gone on a bad afk streak, while the gamblers ruin might apply with the ruin being the lower threshold before giving up lol out of frustration.
i think we both earned a couple of points today... http://xkcd.com/356/
ValyrianBlade2 votesFair enough. I would only mention that if your afk rate is 1% it shouldn't affect your ability to climb much at all (considering some will doubtlessly be in your favor, the 99% of fair Games will apply much more to determining your mmr than the 1% that have an afk on one team or another). I would assume the rate changes over tiers and players complaining see it much more than 1% in the low tiers.
Admittedly though I think the reason it's problematic is because individuals too weak to climb intentionally lose when someone on their team is in a promo series (they can look up your lp easily). The system is broken from that perspective - but that's an argument for another thread.
Sukishoo1 votesI think if one person is gone from either side it shouldn't be counted as a win nor a lose.
