This means that someone who is under-rated (common for new players) or over-rated (common for a new player with a couple wins) will take a lot longer to reach their true rating. I don't think that's a great method.
I think the solution has to involve either anonymity or creating some in-game dynamic to change the incentives.
On playdiplomacy.com, the top players mostly play anonymous games because of this.
@kevinkirkpat you asked «what's the point of a rating system?». I would like to know the answer too! :-)
Really, what's the point of having a rating system? Why do we need one? Morphy, Anderssen, Steinitz, Capablanca, Alekhine and every other player before 1950 knew nothing about ratings. Was chess of their era less enjoyable than chess of our era of ratings? I honestly don't know. More than that: I am honestly confused!
The rating system gives a rough measure of your strength of play as compared to the other people in the same rating pool. For example, I've never played RSK_Asherz, but I can tell by looking at our respective ratings that he's better than me and would probably win if we played, but he's not so much better than me that I would never have a chance. I can also watch how my rating changes over time to see if I'm improving.
The rating system also lets you find opposition at your level. There are a huge number of players between 900 and 1400; that's the vast majority of the rating pool. Without rating-restricted seeks, a grandmaster who tries to play a game online would usually find himself trouncing some patzer who doesn't know the difference between the Sicilian and the Queen's Gambit Declined. Neither player would enjoy that much, and it wouldn't be meaningful practice for the grandmaster. Ratings give you a way to only play people who are close enough to your level that there's a good chance of a reasonably competitive game.
i found the solution, i used to be high rating and first on the leaderboard and had the problem. Then i just lost some rating points, now i dont have the problem anymore! you are welcome 
I considered the anonymity approach also, but seeing the rating helps to potentially plan strategy a bit, but ratings can be deceptive, especially with everyone starting at 1200 in an ELO based system.
Yeah, there are downsides to anonymity too. It might be nice to eventually have the option to play anonymous (with identities revealed after the game), but I'm not sure it should be enforced for all games.
same thing with @chineseroom i found the solution too. back to 1400 - 1600 and blend with them. most of the player they know already what to do specially around 1300+. but if we want have somebody to reach 1800, 1900, or (2000 hoping after 6months from now) rating pool. we can do this thing. but for me in this current system and pairing its hard to reach 1800 or 1900 unless we have around 100 players around 1700 (this is my only humble opinion and base on my observation base on the top player who can reach 17++ and 18++.)
If rating is determined by how you match up against individual player's rating (win or lose), beating the top rated is much more important. The problem with not knowing the ratings is that the rating system depends on it at the moment.
Maybe another way to rate games should be based on the total rating of all the players and your respective finish, instead of individually against every opponent. Say you have 4 players rated 1200 1300 1400 1500. Average 1350. AT the end only finish position 1-4 matters and you evaluate your score based on the average rating. This eliminates the "kill the higher rated guy" from the standpoint of rating.
If rating is determined by how you match up against individual player's rating (win or lose), beating the top rated is much more important. The problem with not knowing the ratings is that the rating system depends on it at the moment.
Maybe another way to rate games should be based on the total rating of all the players and your respective finish, instead of individually against every opponent. Say you have 4 players rated 1200 1300 1400 1500. Average 1350. AT the end only finish position 1-4 matters and you evaluate your score based on the average rating. This eliminates the "kill the higher rated guy" from the standpoint of rating.
I agree. It needs to be points based. Also, not starting at a generic rating, well above your possible strength either, or well below.
Yeah, there are downsides to anonymity too. It might be nice to eventually have the option to play anonymous (with identities revealed after the game), but I'm not sure it should be enforced for all games.
Another thing I considered with anonymity is that, how hard it is look up a guy in a separate window, while waiting to move ? Having said that, I would love to see how anonymity would affect a person's rating, as opposed to not having it.
Yeah, there are downsides to anonymity too. It might be nice to eventually have the option to play anonymous (with identities revealed after the game), but I'm not sure it should be enforced for all games.
Another thing I considered with anonymity is that, how hard it is look up a guy in a separate window, while waiting to move ? Having said that, I would love to see how anonymity would affect a person's rating, as opposed to not having it.
That's easily solved by not even displaying the name until the game is over.
Yeah, there are downsides to anonymity too. It might be nice to eventually have the option to play anonymous (with identities revealed after the game), but I'm not sure it should be enforced for all games.
Another thing I considered with anonymity is that, how hard it is look up a guy in a separate window, while waiting to move ? Having said that, I would love to see how anonymity would affect a person's rating, as opposed to not having it.
That's easily solved by not even displaying the name until the game is over.
That is a great point. While I would prefer to play an anonymous version, I feel like this is spiraling out of control fast for the developers. I actually stopped playing it for now. I feel like I have been ganged up on the last few games I have played and I am not into playing three guys at once. It is hard enough to beat one guy at a time.
icystun:
I am not sure that is true from a theoretical standpoint.
Of course, you gain more points by beating a high rated player, but you also lose less if you finish below the higher rated guy. Vice versa you dont get as many points by beating an equally rated opponent, but you lose more when you finish below him.
There is no inherent advantage of playing destructive against 1 player if that comes (and it always does, otherwise it is not destructive, but self interest) with a disadvantage to your own score/position. Since you are increasing your chance to beat 1 guy, while decreasing your chance to beat the other 2. Of course gangbanging on 1 player is beneficial in theory for the 3 guys (if they invest and gain somewhat equally), but there is no advantage of ganging up on 1 high rated guy as opposed to ganging up on 1 low rated guy. It is just an easy target since the high rated guy is easily distinguishable from the others and they envy his superior playing strength and like the idea of beating him and seeing him suffer.
So the argument "ganging up on the high elo guy is good" made by the original poster is only partially correct. Of course ganging up gives you an edge, but there is no advantage of ganging up on a higher rated guy than ganging up on a lower rated guy.
The extreme of this is, if top player is sufficiently higher-rated than other 3, her rating cannot go down at all (though, naturally, a victory is still to her credit - as per usual "beating weaker players" math).
I like @kevinkirkpat's solution, except for the part above. Points should not be added to the system, otherwise rating inflation may occur. So my correction is:
The extreme of this is, if the top player is sufficiently higher-rated than the rest, his rating cannot go up or down at all.
My reasoning is: Such an extreme case implies that the matching algorithm came up with an extremely unfortunate matching, highly displeasing for the top level player. By making the result of the game important for his rating we add insult to injury. Just make the result irrelevant, and everyone will be reasonably happy.
Asthings stand right now, much as I can sympathize with complaints of top players feeling that other players gang up on them; I can also sympathize with the rational-game-theory side of the equation. If I'm a 1200 player, and two of my opponents have ratings of ~1200, and the fourth player has 1450, there's an obvious incentive for me to try working with the other 1200 players to beat the high-rated player.
The math is simple: if we team up and eliminate 1450-player, I'm left with a
1/3 chance of first place (and gaining substantial rating points from finishing over two equal players and a 1450 player)
1/3 chance of second (again, still getting decent rating points due to finishing ahead of 1450-player).
1/3 chance of third place (possibly breaking even, losing to 2 equal-players might offset beating a 1450 player)
If we don't team up, and I assume the 1450 player has a 2:1 skill edge (would win 50% of the time in an alliance-free match), then my odds are
1/6 (1/3 of the 50% that 1450 player loses): first place. As above, lots of points, but far less likely.
5/18 (27%) second place, gaining few points for beating 2 low-rated players
5/18: (27%) third place, losing a few points for losing to high-rated, with offsetting win&loss against 2 equal-rated players
5/18: (27%) fourth place, losing lots of points for losing to 2 low-rated players
I suggest that the point system be modified to offset these incentives. If left unchecked, nobody will become GM-level high player; the further someone pulls ahead in ratings, the more they will be sabotaged in their games. While it might be argued, "Hey, that's the way the cookie crumbles in 4PC", I'd retort that
1) Obviously certain players are incredibly skillful, and in a blind matchup would trounce the majority of opponents.
2) If the point system is set up / enforced in a way that precludes such high-skill players from obtaining high ratings, then... well, what's the point?
What I'd suggest is an adjustment to the point system to account for skewed match-ups. Basically, the bigger the gap between best player and average of other players:
* The fewer points top-player loses for getting 3rd/4th place. The extreme of this is, if top player is sufficiently higher-rated than other 3, her rating cannot go down at all (though, naturally, a victory is still to her credit - as per usual "beating weaker players" math).
* The fewer points bottom-3 players can win by finishing ahead of top-player. The extreme of this effect would be all 3 bottom players receiving points based only on their relative finishing place (independent of top player)
The effect should be:
A) In a matchup between 3 low-rated players and 1 high-rated player, the high-rated player is still trying to win.
B) The low-rated players are most concerned with defeating the other low-rated players (e.g. defending themselves from "dangerous" high-rated player, still seeking to get points from attacking high-rated player's pieces, but not overtly focusing on high-rated player at the risk of exposing them to attack from other 2 low-rated players).
C) In a circumstance where high-rated player neverthelss feels unfairly targeted, he/she can simply resign, losing few (if any) rating points.