Even with Gregorios suggestion we need a sensible translation from GO ratings to ordinary ratings. Also, it will be greatly inflenced by single excellent players in certain games. I must say I prefer the original suggestion - but GO RATINGS MUST BE TRANSLATED DIFFERENTLY! I will explain why:
As stated earlier: For the top 50 players on the MONSTER list, the average GO ratings are 2180, whereas average ratings in other games are 1597. With 1 Dan = 2000, the average GO rating would still be high (1760). This should be more than enough proof that the current translation is wrong, and that 1D=2000 is a much better estimate.
Richyfourtytwo states that we start with 20k in GO and 1500 in other games. That is wrong. We start with a self-proclamed rating between 20k and 1k in GO, and 1500 in other games (in chess, we used to start with a self-proclaimed rating between 1000 and 2000). Thus, all we can say from this is that 1500 translates to something between 20k and 1k.
Also, GO players will (obviously) give themselves a higher rating than 20k - probably close to 1k. These are the players who play the most games (there are obviously exceptions, but still). Thus, most games are played between players with a much higher starting rating than 20k.
Look at chess: There, one could give oneself a rating between 1000 and 2000. If we were to translate this the same way you do for GO, we would say that a Chess rating of 1000 corresponds to a rating of 1500 in other games. But look at the chess ratings - they are very similarly distributed to that of other games as it is! Obviously, for the same reasons I have stated for GO.
Some may claim that good GO players should have a higher rating than a good player in other games. Possibly true to some extent, but definitley not for GO9! Hex19, for example, is far more complex than GO9, and should also distinguish more between good and bad players than GO9.
Why does this matter? Well, GO players get far more points than they should this way. Compared to what I believe is a sensible translation, good GO players get 1500 points more than they should have, compared to poor GO players. That is more points than many players have in all other games combined!
Why is 1D=2000 a good translation? Well, primarily because you get rating distributions similar to other games. The top players then get around 2300 rating, which seems natural. Also, this means putting 1500=10k, which is the median start rating (right in the middle of 20k and 1k), but this is to me less important.
One can also see that the current translation is bad when looking at the low end of the GO players - you'll hardly find a player with ratings below 1400!
Just because people who hardly play GO start with 20k, does not mean AT ALL that this corresponds to 1500 rating. Look at what people who DO play LOTS of GO games start with. They start with ratings from 10k to 1k.
I know that this rating is "just for fun", but I really like the idea. However, it is only meaningful if one use a sensible translation between GO ratings and other ratings, so to me, now it makes no sense at all.
As for "Ray is still on top, so GO is not that important": Please. Ray is ONE person - he's the MONSTER! This only means that DESPITE the heavy dependence on GO games, Ray is so great at most other games that he still manages to come out on top. That says a lot about Ray, but not much about the system.
Whether GO, Hex and Reversi should be weighted like they are in the 2nd Monstership, that's another issue. But if the rating is to have a "Monster" relation, that might be a good thing. This is not that important to me, although it might be if these ratings are to be used as seeding in a later Monstership.
So, I ask again: Please, change the translation between GO ratings and ordinary ratings, so this "Monster rating" actually does have a little meaning.
Little Golem