Elo ratings in games

I've just recently started getting 'into' stats, at a very basic level, but am becoming ever more curious. Another interest of mine is videogames, and the two combine quite well.

My question is on the Elo ranking of players in games, especially in DoTA-style games. The ongoing discussion in the community is on the existence of Elo-Hell: a bracket of Elo rankings where people are consistently being teamed up with people below their level and being dragged down/held back in rankings because of it. I'm wondering how legitimate this complaint is.

As I understand it, a person's Elo ranking is increasingly accurate the more games he or she plays. However, the formulas for Elo ratings are kept hidden by the game producers, and from my reading, Elo was conceived for chess, not for 5v5 or 3v3 games. There are loads of questions I'm curious about really, including one to the people who run the games on their formula's K-factor, but I guess the starting point is:

In a team game of 5v5, where everyone begins at a specific ranking but can move below or above it, how many games will it take for a player's Elo to be 'correct' (this is holding skill-level constant)?

Thanks in advance :)

- if anyone's trying to guess, it's League of Legends I'm talking about.