Assume that when player A and player B start a game, player A's rating is x. During the game, player A wins a few other people and his rating increases to x+y. Now, he also wins his game with player B. Is it x or x+y which is counted to calculate player B's new rating after he loses the game?
If it is x, then it's fair; but if it's x+y, then it opens ways to exploit the rating system.
As a concrete example, a player who was checking his opponent's other games with other people and knew that his opponent was going to win those games within a few days, simply stopped playing (using the timebank) and waited (and is still waiting) for the opponent's rating to increase further. The rationale is that he wants to postpone "resigning" to a later time (to when his opponent's rating is increased), because he wants to lose fewer rating scores after losing the game.
I think that this exploit should be blocked and the way to block it seems to be taking the ratings of players at the time they start their games, not when they finish them.
EDIT: He finally resigned and lost only 7 rating scores (using the abovementioned exploit).
Originally posted by EminentStudentRatings are calculated at End with good reason.
Assume that when player A and player B start a game, player A's rating is x. During the game, player A wins a few other people and his rating increases to x+y. Now, he also wins his game with player B. Is it x or x+y which is counted to calculate player B's new rating after he loses the game?
If it is x, then it's fair; but if it's x+y, then it ope ...[text shortened]... : He finally resigned and lost only 7 rating scores (using the abovementioned exploit).
If a player drops to a LOW rating and starts 100 games they can easily play games that are almost guaranteed wins. I believe Ragnorak had a nice description of a 1900 player dropping to 1000 or so and by the end of winning most games gaining 3800 rating or more.
You can manipulate most anything, but ratings based on start of game is too easy to manipulate so after game ratings is the much lesser of evils.
P-
Fair point!
Both systems can be exploited. The current system (which takes the END ratings) can be exploited by the player who loses the game and the alternate system (which takes the START ratings) can be exploited by the winner (in the way you described). But can't we have a hybrid system which takes the END rating for the winner of the game (to add to his rating) and the START rating for the loser to decrease his rating? This hybrid method seems to be able to block both exploits (unless it leads to certain inconsistencies which I've not thought of).
Originally posted by EminentStudentRead your first example... but now waiting for a players rating to go UP, you see they are some lost games and wait for it to go down.
Fair point!
Both systems can be exploited. The current system (which takes the END ratings) can be exploited by the player who loses the game and the alternate system (which takes the START ratings) can be exploited by the winner (in the way you described). But can't we have a hybrid system which takes the END rating for the winner of the game (to ...[text shortened]... to block both exploits (unless it leads to certain inconsistencies which I've not thought of).
Ratings go up and down all the time.
No one is going to break the bank 'gambling' that any given user's rating may go up or down.
A rating is just a number to gauge how well (or poorly) you play chess. Having this number inflated will only bring it tumbling back where it belongs in the next few finished games.
Don't let it bother you, regardless how you do it there is a way to exploit it if you like... just remember it WILL go back where it belongs eventually.
P-