Open Side Menu Go to the Top
Register
FIBS Ratings FIBS Ratings

02-28-2009 , 01:45 PM
Could someone who knows more about FIBS ratings than I do classify these player types and give them what you think would be a typical FIBS rating?

Complete beginner
Casual player (plays with friends, knows the rules but nothing more)
Guy who has read an introductory book
Guy who has read a couple books and put in some time playing against w/ GNU or Snowie tutor mode
Guy who is good and studies a lot
Very good human players
World's best human player
The best bot (Snowie 4 I presume)

I'm just trying to get a feel for the rating system. Feel free to make up your own categories if you like.
FIBS Ratings Quote
02-28-2009 , 05:55 PM
Quote:
Originally Posted by kitaristi0
Could someone who knows more about FIBS ratings than I do classify these player types and give them what you think would be a typical FIBS rating?

Complete beginner
Casual player (plays with friends, knows the rules but nothing more)
Guy who has read an introductory book
Guy who has read a couple books and put in some time playing against w/ GNU or Snowie tutor mode
Guy who is good and studies a lot
Very good human players
World's best human player
The best bot (Snowie 4 I presume)

I'm just trying to get a feel for the rating system. Feel free to make up your own categories if you like.
GamesGrid started the use of rating backgammon players online, and developed a system where everyone started at 1500 and moved up or down from there. Over time, the pool of players sorted themselves out roughly as follows:

Casual player (plays with friends, knows the rules but nothing more) = 1100-1200

Guy who has read an introductory book = 1200-1400

Folks with some experience and knowledge = 1400-1800

Very good human players = 1800-2000

World's best human player = 2000-2050

The best bot (Snowie 4 I presume) = 2100 or so.

GamesGrid isn't around anymore, but I think most of the sites that use ELO ratings would have a breakdown something like this.

Notice the relatively small range between the bottom of the heap and the top, only about 1000 points. Chess has a range of greater than 3000, demonstrating its greater complexity. The range for Go, the most complex game of which I'm aware, would be even greater, probably around 6000 points.
FIBS Ratings Quote
02-28-2009 , 06:00 PM
Thanks
FIBS Ratings Quote
02-28-2009 , 09:31 PM
Quote:
Originally Posted by Robertie
GamesGrid started the use of rating backgammon players online, and developed a system where everyone started at 1500 and moved up or down from there. Over time, the pool of players sorted themselves out roughly as follows:

Casual player (plays with friends, knows the rules but nothing more) = 1100-1200

Guy who has read an introductory book = 1200-1400

Folks with some experience and knowledge = 1400-1800

Very good human players = 1800-2000

World's best human player = 2000-2050

The best bot (Snowie 4 I presume) = 2100 or so.

GamesGrid isn't around anymore, but I think most of the sites that use ELO ratings would have a breakdown something like this.

Notice the relatively small range between the bottom of the heap and the top, only about 1000 points. Chess has a range of greater than 3000, demonstrating its greater complexity. The range for Go, the most complex game of which I'm aware, would be even greater, probably around 6000 points.

have you got any idea what the equivalent snowie error rates are?
FIBS Ratings Quote
02-28-2009 , 11:09 PM
Snowie's error rate is around 1.3 to 1.5. Since Snowie can't detect errors in its own play by evaluation alone, you have to do rollouts on all the interesting moves that occur during a Snowie session, which results will then override the Snowie evaluations.

The best human players seem to play in the 2.5 to 4.0 range, on average.

Average Open tournament players (those who play in the top section of major events) include the world-class players above as well as players as far down as the 7.0 to 8.0 range.

A few years ago I played in an online tournament sponsored by the WarpGammon group, which runs big online events with long matches but no money. It was great for practice, but since it was a draw of 128 players with each round being best 3-of-5 11-point matches, took quite a while to finish. Some of my early round opponents would certainly have been considered weak intermediates (at best) by the standards of live tournaments, and I recall that their error rates were in the 13.0 to 15.0 ballpark.
FIBS Ratings Quote

      
m