[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Bug-gnubg] Cubeful Equities in GNU

From: Ian Shaw
Subject: RE: [Bug-gnubg] Cubeful Equities in GNU
Date: Mon, 5 Oct 2009 09:38:11 +0100

Hi Rick,
What you need to do is download the databases you want and put them in the gnubg directory, unzipped of course. Then you rename them as listed below, to govern how gnubg uses each file.
Gnubg uses up to four bearoff databases, and identifies them from the filename.
gnubg_os.bd - One-sided database read from disk as required.
gnubg_os0.bd - One-sided database loaded into memory (by default a 15-chequer, 6-point database)
gnubg_ts.bd - Two-sided database read from disk as required.
gnubg_ts0.bd - Two-sided database loaded into memory (by default a 6-chequer, 6-point database).
This allows you to choose how large a database you want to load into memory, gaining speed at the expense of memory usage. If gnubg can't find the position in memory, it will then look it up in the file stored on disk.
To check what is installed:
Select Analyse...Evaluation Engine (or, from the command line, type "show engine")
I hope this answers your question. 
Finally, I'll just take this opportunity to express my appreciation and thanks for the contribution you've made to backgammon theory and the redevelopment of strong bots.
-- Ian Shaw

From: address@hidden [mailto:address@hidden On Behalf Of Mueller Achim
Sent: 05 October 2009 06:58
To: gnubg-list
Subject: [Bug-gnubg] Cubeful Equities in GNU

Anfang der weitergeleiteten E-Mail:

Von: "Rick Janowski" <address@hidden>
Datum: 3. Oktober 2009 17:29:30 MESZ
Betreff: Cubeful Equities in GNU

Hi Achim,
I recently downloaded GNU backgambmon and was gratified that the programmers have made use of my work on doubling cube models and have made appropriate acknowledgement.
As you may be aware, I did quite a lot of work with Snowie over a short period six or seven years ago in helping them develop cubeful models for matches. The general approach, for both money and match play, was to estimate equities where the cube is fully dead or fully live for both players. The effective equity was then interpolated between these two extremes based on cube-usage factors (dependent on future volatility and other similar factors). The developers of Snowie considered both the simple case of a cube usage value (ie, for both players) and two separate values (ie, separate for each player).

Looking over my work more recently, I came to the conclusion that my way of looking at the refined model (with separate values for cube usage for both players) was flawed to some extent, particularly with regard to cube-centred equities. I had originally considered interpolation only between the two extreme cases of dead-dead and live-live cube values. In reality there are four extreme cases:

Dead-Dead     Cube dead to both players

Dead-Live       Cube dead to Player 1 but fully live to Player 2

Live-Dead       Cube fully live to Player 1 but dead to Player 2

Live-Live        Cube fully live to both players

These four points may be imagined as four corners of a rectangle, with Player 2 cube-liveness being the measure on the axis of Dead-Dead to Dead-Live. Similarly, Player 1 cube-liveness being the measure on the axis of Dead-Dead to Live-Dead.

I would be interested in your ideas on the value of developing this approach.

Could you advise me on how I can incorporate within GNU the following downloads from your website:
* One-sided race database reaches the midpoint - A one-sided race database that extends to the 13-point is now available for download. 
* 11.1.2 Download You may download the two sided database with 6 checkers on 6 points from ftp://alpha.gnu.org/gnu/gnubg/gnubg_ts0.bd.gz and the one sided database with 15 checkers 6 points fromftp://alpha.gnu.org/gnu/gnubg/gnubg_os0.bd.gz.
Rick Janowski

achim mueller, ederweg 3, D-48431 rheine
+49 (0)5971 83767, +49 (0)163 8458340
pgp/gnupg key: 1024D/5DF3A722 (wwwkeys.de.pgp.net)

reply via email to

[Prev in Thread] Current Thread [Next in Thread]