Forum Archive :
Ratings
Claes Thornberg wrote:
> Why is it that you (on FIBS) gain more from a 2ptr than from a 1ptr,
> when we all know that in a perfect world a 2ptr is not different from
> a 1ptr? (see kit woolsey monograph 'How to play tournament backgammon')
> Can we design a ratingsystem that works for all different kinds of
> matchlengths? Or is it better to have different ratings for different
> match lengths, and only allow matches of certain lengths?
I've looked into this question a little bit. One idea I
had was to use a different way of measuring the opportunity
for skill in the match. FIBS just uses the match length.
I figure a better indication of the amount of work involved
is the total number of dice rolls in the match. Or, even
better, the number of dice rolls in "contact" positions.
So how many "contact position rolls" are there in typical
matches of various lengths? To answer this, I scanned
through the Big_Brother database and counted up the rolls.
Here are the results.
Average Rolls Per Match in Big_Brother Matches
rolls matches
 
Match length 1: 17329 / 359 = 48.27
Match length 3: 54800 / 639 = 85.76
Match length 5: 116451 / 907 = 128.39
Match length 7: 86353 / 498 = 173.40
Match length 9: 9992 / 44 = 227.09
Match length 11: 8230 / 31 = 265.48
Note that there are less than twice as many rolls in a
3point match as there are in a 1point match. So, if
counting rolls is a good indicator, it suggests that
opportunity for skill is not really proportional to the
length of the match.
The other idea I had was to do exactly what Claes suggested
in another post: Compute a match equity table for players of
different ability. This method is not perfect  the
computation assumes that both players always make efficient
doubles. This will produce some error, but I think the
results are still useful.
The following table is constructed assuming the stronger player
wins 51% of the time and that 25% of all wins are gammons.
Match Equity Table between Players of Unequal Ability
1 2 3 4 5 6 7 8 9
        
1: .5100 .6974 .7636 .8300 .8549 0.9011 .9186 .9431 .9528
2: .3226 .5100 .6067 .6770 .7496 0.8061 .8461 .8788 .9045
3: .2563 .4158 .5124 .5849 .6602 0.7237 .7717 .8129 .8474
4: .1871 .3441 .4394 .5123 .5883 0.6543 .7077 .7529 .7940
5: .1611 .2717 .3651 .4381 .5144 0.5816 .6390 .6890 .7351
6: .1121 .2128 .3003 .3717 .4475 0.5150 .5754 .6280 .6779
7: .0934 .1715 .2513 .3179 .3903 0.4555 .5162 .5700 .6220
8: .0664 .1367 .2082 .2714 .3393 0.4026 .4625 .5167 .5694
9: .0558 .1093 .1722 .2292 .2925 0.3525 .4108 .4647 .5177
If this table is accurate, then a player who beats his opponent
51% of the time in a 1point match can expect to win 51.77%
of the time in a 9point match.
There are some interesting features in this table. It says a
2point match gives the stronger player no more advantage than
a 1point match. This is true if the weaker player always
doubles at his first opportunity. It also says a 4point match
is no better for the stronger player than a 3point match.
This may be an exaggeration (coming from the assumption of both
players making efficient doubles), but it is something to think
about.
To summarize, let me compare the relative skill levels
estimated by each of these three methods:
1. The FIBS method.
2. The counting dice rolls method.
3. The match equity table method.
We'll start by assuming that a 1point match gives the
stronger player 1.00 units of advantage. Here are the
relative advantages predicted by each method for longer
length matches:
FIBS Rolls MatEq
  
Match length 1: 1.00 1.00 1.00
Match length 2: 1.41 1.00
Match length 3: 1.73 1.33 1.24
Match length 4: 2.00 1.23
Match length 5: 2.23 1.63 1.44
Match length 6: 2.45 1.50
Match length 7: 2.64 1.89 1.62
Match length 8: 2.83 1.67
Match length 9: 3.00 2.16 1.77
Match length 10: 3.16 1.84
Match length 11: 3.31 2.34 1.92
Both the countingrolls and matchequitytable methods
give less credit to longer length matches than FIBS does.
Or, another way of saying this, they give more credit to
1point matches.
The question of how much greater advantage a stronger
player has at longer length matches is an interesting one.
I'd like to hear other people's ideas on this.
Tom




Ratings
 Constructing a ratings system (Matti RintaNikkola, Dec 1998)
 Converting to pointspergame (David Montgomery, Aug 1998)
 Cube error rates (Joe Russell+, July 2009)
 Different length matches (Jim Williams+, Oct 1998)
 Different length matches (Tom Keith, May 1998)
 ELO system (seeker, Nov 1995)
 Effect of droppers on ratings (Gary Wong+, Feb 1998)
 Emperical analysis (Gary Wong, Oct 1998)
 Error rates (David Levy, July 2009)
 Experience required for accurate rating (Jon Brown+, Nov 2002)
 FIBS rating distribution (Gary Wong, Nov 2000)
 FIBS rating formula (Patti Beadles, Dec 2003)
 FIBS vs. GamesGrid ratings (Raccoon+, Mar 2006)
 Fastest way to improve your rating (Backgammon Man+, May 2004)
 Field size and ratings spread (Daniel Murphy+, June 2000)
 Improving the rating system (Matti RintaNikkola, Nov 2000)
 KG rating list (Daniel Murphy, Feb 2006)
 KG rating list (Tapio Palmroth, Oct 2002)
 MSN Zone ratings flaw (Hank Youngerman, May 2004)
 No limit to ratings (David desJardins+, Dec 1998)
 On different sites (Bob Newell+, Apr 2004)
 Opponent's strength (William Hill+, Apr 1998)
 Possible adjustments (Christopher Yep+, Oct 1998)
 Rating versus error rate (Douglas Zare, July 2006)
 Ratings and rankings (Chuck Bower, Dec 1997)
 Ratings and rankings (Jim Wallace, Nov 1997)
 Ratings on Gamesgrid (Gregg Cattanach, Dec 2001)
 Ratings variation (Kevin Bastian+, Feb 1999)
 Ratings variation (FLMaster39+, Aug 1997)
 Ratings variation (Ed Rybak+, Sept 1994)
 Strange behavior with large rating difference (Ron Karr, May 1996)
 Table of ratings changes (Patti Beadles, Aug 1994)
 Table of win rates (William C. Bitting, Aug 1995)
 Unbounded rating theorem (David desJardins+, Dec 1998)
 What are rating points? (Lou Poppler, Apr 1995)
 Why high ratings for onepoint matches? (David Montgomery, Sept 1995)
From GammOnLine
Long message
Recommended reading
Recent addition

 
