  Difference Win Lose
400 +29 3
350 +28 4
300 +27 5
250 +26 6
200 +24 8
150 +23 9
100 +20 12
50 +18 14
0 +16 16
50 +14 18
100 +12 20
150 +9 23
200 +8 24
250 +6 26
300 +5 27
350 +4 28
400 +3 29

  This was posted a long long time ago, I don't remmber who figured it out, I cant take credit for it, but I saw people wondering how the rating works so I reposted it.

  that isnt right

  Lanky, are you wearing a dog collar in your picture?

  No, Lanky is a priest  that is his outfit. I would smoke a fatty spliff too if I had to sit through one of his sermons.

  Could you please explain that chart in more detail?

  What tap showed is how the formula incorporates an extension of the gaussian curve that this site uses as the formula:
32*(1(1/(1+10^((rxry)/400))))
That is amenable to chess and other games that relate to the gaussian distribution but it lacks a corrective factor for greater than 2 players and is not amenable in this simplistic form for the game of risk.
The placing in first versus last are not inverse relationships to incorporate the binomial aspect of how the above rating formula applies rating changes in inverse relationships for first and last place.
The standard deviation of the distribution of rating changes should be compressed for a multiplayer game where there is much luck involved. This is not simply a game of skill like chess.

  In a 3player game, suppose your rating is 1500.
If your opponents are 1500 and 2000 and you place first, you will improve by 46.
If you take first with opponent ratings of:
1500,3000, you add 48, just 2 more than above
If you take first with opponent ratings of:
1000,2500 (similar to first example with same mean of opponents but greater dispersion), you improve by just 34.
Strange. Something funky is going on.

  The current rating system is extremely progressive and for no care of the number of games of experience. You cannot see so well in the formula I devised because I condensed it to include all possibilities of 3,4,5,and 6player games but I have an accelerated rating aspect that will cause rating changes to be more drastic upon the first few games and then tapering off.
NEWRATE = OLDRATE + MAX(1, 4(GAMES^2/200)) * 8 * ((2PLACEFACT(PLAYERS1)/(2*FACT(PLAYERS3)))/
((FACT(PLAYERS1))/(2*FACT(PLAYERS3)))
(TRUNC((PLAYERSPLACE+1)/PLAYERS))/((FACT(PLAYERS1))/
(2*FACT(PLAYERS3))))*((1PLAYERS)*TRUNC((PLAYERSPLACE+1)/
PLAYERS)+1)*((TRUNC((PLAYERSPLACE+1)/PLAYERS)*(1/(10^((OLDRATE  AVGRATE)/1600)+1))(TRUNC((PLAYERSPLACE+1)
/PLAYERS)1)*(11/(10^((OLDRATE  AVGRATE)/1600)+1))))
You can see the similarity with the exponential function with my equation and the one that is used for this site.
To have a rating it should exhibit some stability, and not vary 10% from a single game after a rating has been established.
Imagine if such an equation was used for chess ratings. It would be way too wild. Instead of a player having a rating range of perhaps 2455 to 2535 over the course of a year, it would be more like 2100 to 2800 and when all other players are exhibiting the same type of wild variation, it would be difficult to tell who ranks where.

  The truncated formula took a long time to create. I did it so this site could easily incorporate it instead of having to use a "case" statement in the programming.

  dont you mean yes  lanky is a priest

  In a 3player game, suppose your rating is 1500
at the other two players are 1500 and 1900 rated.
Say for some strange reason I get 1st.
Would my rating go up the same no matter who gets 2nd and 3rd?


Reply to this discussion
