According to this article in league of legends matchmaking there's an average of 50/50 chances of winning or loosing a game based on how system ranks a specific account (player skill). However this implies that a player is expected to loose at least half of his games (provided that he's not a student of the game) or at most he'll win 75% of the games (that's ranked top rated guys and the highest win rate I've seen in my region, so should be a feasible boundary).
However, the latter is rather rare since not so many people actually achieve that skill level (that can also be observed by going through ranked statistics) and the former is also not a constant since a) it's also in the algorithm and a queue of 1 min is most likely not a 50/50 and b) there are situations where it is extremely obvious that a player was placed against a way stronger team or vice verse.
An example from my practice is that I had a win rate of 88% which severely dropped after about one ore two weeks (to around 71%) and it rose again in a week or so with me learning nothing (I literally didn't). So, during such drops a player reward expectation is not met which results in negative attitude (taken to extreme in some cases, as was also observed). Some people were literally saying: "I hate the game" and they kept playing.
So the question is: why? why is it that with such reward volatility people still keep playing? (please, give a rigorous proof, if any, supplied with references presumably from behavioral psychology)
P.S.: I do not have any empirical data, so I realize this might be only a feeling as well as the fact that mostly I observed it through my account (not even lvl 30) and in my region. I also understand that most likely they are using statistical learning algorithms which obviously require data and will put a player to his actual rating as number of games grows (theory of big numbers).