Re: [Bug-gnubg] Interesting backgammon scoring system

[Roy A. Crabtree]

Not sure what the "tables" referred to are and why they would be regarded as 
"fair" versus the simple betting match method of counting up the remaining pips and
considering the betting cube.

Albeit the betting match principle of establishing a value per pip at the beginning
of each game, as well as final pip count, do not get called into each game's
or a matches (or tournment's) play, depending on the rules of the tournament.

I found what appears to be an anomaly in gnubg bearoff play, that would be
"fixed" by "training" the nets n such a circumstance for such a weighted
scoring system (betting/straight, or as proposed by this method).

White: 6/1 3/1 1/1
Black: 5/1 2/2 1/1 (20/1 23/2 24/1)

White rolls 64, moves 6/2 3/off (inverted) leaving 2/1 1/1

     rather than 6/off 3/off leaving 1/1

Version: Wndows .14-Mingw current
(I have the game file for those who want it).

Assuming this is confirmed, you can see that there is no difference at the point
after white's move which ever is taken:

  but the net plays the "dumber" roll leaving the "impression" that it is
more fungibly frangible than it actualy is.

And players watching the play would be trained to take the
suboptimal bearoff.

Which, in fact, a betting player might do, ...

   to mislead the oppositon.

But that would be a losing lay within the contexct of a single
betting game or ending game in a match.

THough it might be a winning srategy to play against
the _opponent_ nnetheless earlier in such a match

   (If the opposition were "dumn" or "misleadable"
     should the context be much more complex)

Long term, beting players only continue to succeeed mximally
when they continue to _successfuly_ encourage _losing_
players to continue to bet against them and longer
term betting players only get that when

  a)  they successfully _hide_ their expertise
  b) they succesfully _hide_ which _tell_ they are using
  c) most importantly:

         they train (seed and breed) new players to bet against them

which leads to the most skilled players in betting to
attempt to cull players by misleading them via trainng to use

   _suboptimal moves_ against them in fixed positions;

either on the basis of the rules of gammon or the current context

     (variations such as Jacoby rule, etc.)

or when the lesser player plays against this player in specific:

    playing the player.

The main sequence of the "fair" rules ystem as poposed of

  valid interst

is in the initial opening roles balancing, and alternating, plus
counting pips in trhe last or final position.

Although I might propose some additional changes:

    i) mimic the strength od position coring/weighting
      system currrently accepted as "best"

   ii) Further away means less pay:

       24/1 would be worth less than 23/1 1/1

   iii)   from inner to outer from sinner to doubter:

       pips on your inner board are worth more than your outer:

      24/1 would be worth less than 6/4.

   iv)  ....and so forth ...

Cheers.

   ii)

On 9/14/06, Øystein Johansen <address@hidden > wrote:

Hi all,

I've made contact with Dr. Jakob Garal. He is the inventor of a new
backgammon scoring system. The scoring system mainly differs with normal
backgammon in the way that not only win, gammon and backgammon is the
outcomes of the game, but the "quality" of the win also matters for the
result. You get more points the more of the opponents checkers are left
when you win. A close race is less points than just a close gammon save.

In addition his scoring system lends some principles from bridge. It
uses Match points, international match points (IMP) and victory points.

You can read more about his scoring system, and even play a match/game
in his server at http://www.fairbg.com

Personally I find his scoring system a bit amusing. I believe the
scoring system can introduce new depth to the game, ( in the same way
the doubling cube did about 80 years ago)

Now, it would be really überkool if the scoring system was implemented
in GNU Backgammon! I believe the scoring system can be implemented,
(more or less trivial), but it will be really hard to analyse and
evaluate a position according to this system.

If someone reading this feels this looks like a challenge, or if someone
got curious, go to the site and look at the rules. I'm also quite sure
Jakob will provide you more information if you're interested in
implementing the scoring system.

-Øystein

PS: Analysing according to the scoring system will probably need a full
reimplementation of the neural nets, since there is lots of possible
outcomes of each game.



_______________________________________________
Bug-gnubg mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/bug-gnubg

-- 
 --
Roy A. Crabtree
UNC '76 gaa.lifer#11086
Room 207 Studio Plus
123 East McCullough Drive
Charlotte, NC 28262-3306
336-340-1304 (office/home/cell/vmail)
704-510-0108x7404 (voicemail residence)

address@hidden
address@hidden
address@hidden

http://www.authorsden.com/royacrabtree
http://skyscraper.fortunecity.com/activex/720/resume/full.doc
--
(c) RAC/IP, ARE,PRO,PAST
(Copyright) Roy Andrew Crabtree/In Perpetuity
    All Rights/Reserved Explicitly
    Public Reuse Only
    Profits Always Safe Traded