Talk:Elo rating system/@comment-188.178.187.14-20120126081725/@comment-77.97.134.113-20120127181004

I agree with both of you.;) I think that ELO system is not a right choice. In fact this is used in quite ridiculously different environment it was designed for. There are so many external factors affecting performance of the players that ELO system is not accurate. However, I think that Law of Large Numbers is a valid argument supporting the ELO system. When you have ~200 your average performance gets you closer to where you should be. The Law of Large Numbers can be used specifically to indicate the average of calculations that include a variety of external factors e.g. if you played 1000 games the external factors can be ignored completely as, due to LLN you are, where you should be. That does not mean after 200 games you are exactly, where you should end up but you are most likely 'moving', to ELO you belong to.

English is not my first language so I will try to explain it clearer. Basically you are 25% better than most people at your level of ELO. Assuming all other players, who are on the same ELO, have equal skills you should be moving up constantly. However, you then have to take into account the mass of external (and often completely random) factors e.g. GP shot someone and instantly killed because of Critical Strike. Now this could not happen with out the Crit. It gives him a huge control over the lane and increases the chance of his team to a greater extent that it should (after all it was just a lucky Crit). There are so many factors that influeance the outcome that, although you are 25% better than other, you climb up the ELO ladder VERY slowly. This is what makes LLN appropriate for the calculations but also shows that ELO system is not suitable. After 200 games I may be still climbing up at a rate of 2%. Assuming I only play one ranked a day this is ridiculously low rate. And it is affected by the external factors to nearly the same degree that it is affected by performance of the player.