Tesunie, on 06 January 2016 - 06:27 PM, said:
Elo as I understand it:
- Is based on W/L.
Elo ratings are derived from the results of playing the game, using win or loss as the only input. That's not the same as saying it is "based on W/L."
Quote
- It gives a rating/number to a player based on their individual W/L rate.
Sort of. It isn't based on the player's W/L rate, it's based on their results against other players. That sort of
sounds like W/L, but it actually isn't.
For example, you can't look at a player's stats and say "this player has a W/L rate of .500, so his Elo rating is 1600." His Elo rating will depend on the ratings of the players he beat and lost to, not simply on his W/L rate.
Quote
- It then tries to pair that player with another player (single player related) who has a similar if not exact rating. (Or for teams, tries to create teams that all player Elo scores equal a certain total per team.)
False. Elo is a rating system only, it doesn't have anything to do with matchmaking.
A
matchmaker can use Elo ratings as input into its matchmaking process, but that's not at all the same thing.
Quote
- If there is a disparity of their Elo rating, then the formula predicts (by what PGI said of their own formula) which player (or Team for MW:O) is likely to win.
An expected outcome can be derived from relative Elo ratings, yes. But that's not really "predicting" the outcome of the match like people think. It simply gives you a percent chance that side A will beat side B.
PGI's matchmaker attempted to keep the aggregate Elo ratings of the two teams equal. If the two sides' Elo ratings are equal, then the expected outcome is 50/50. I.e. both sides have an equal chance to win.
Quote
- If the player who is predicted to win wins, then there is little change in their rating. If the player predicted to win losses though, then changes happen accordingly.
This is where PGI deviated from a proper Elo system, and their deviation made the system less responsive. Their change still works, it just works even more slowly than normal.
In a proper Elo system, your rating will change after every match (with some minor corner-case exceptions) in line with the expected outcome.
What that means is that if you and I play, and I have a 90% chance to win, then if I win there will be very little change in either of our ratings. But if I lose there will be a very large change in both of our ratings. If I only have a 50% chance to win, then win or lose both of our ratings will change by a moderate amount. The winner's rating always increases while the loser's rating always decreases. (Again, except for some corner cases that aren't worth worrying about.)
Quote
- As stated before, is a system designed best for solo PvP (as in, one player on each side, not a team of players per side) styled games, such as chess (or checkers, or boxing even, or Battleship, or...). In that area, it is very effective.
It was first used for chess, but it isn't correct to say that it was "designed for Solo PvP" in any way. Elo ratings are simply a mathematical formula that works in any 2
sided system. Each side can have as many players on it as you want.
Quote
- Doesn't work so well for Multi-player, as each player could have vastly different actual skill levels within the (randomly) assigned team (for PUGs).
False. Elo works just as well for multi-player as it does for 2-player, it just takes longer to reach equilibrium. PGI's Elo ratings took even longer because of the change they made. However, players with thousands of games easily had accurate Elo ratings even with PGI's modified Elo ratings. (The math says that a couple hundred games at most would be plenty, and possibly as few as 50.)
Quote
Did I miss anything major? Did I mess something up here?
You fell into the typical gamer trap, that's all.
Elo ratings are a very simple, very accurate mathematical formula. All of the rest of that baggage really belongs to the PGI matchmaker and it's use (or misuse) of Elo ratings as input.