bug-gnubg
[Top][All Lists]
Advanced

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

RE: [Bug-gnubg] Holding game improvement?


From: Ian Shaw
Subject: RE: [Bug-gnubg] Holding game improvement?
Date: Mon, 9 Jun 2003 09:47:43 +0100

> 
> It would be nice to start with a clear definition of a holding game. 
> Then we can see if we can anything about it.
> 

To kick off the brainstorming I'll offer a definition:

At least 1 side has Anchor[3,4,5,6?,7] AND NOT 2ndAnchor[1-7]

This would include holding games & mutual holding games but exclude backgames 
and low anchor games. Kit Woolsey's' new encyclopaedia has ace & two point 
games in a different chapter, so I guess he classes them as a different beast. 
Also, IIRC, Neil's comments were prompted by a four-point game. I don't know 
how he feels about low anchors.  High anchor games can be borderline 
take/passes, so getting the absolute equity right becomes important. Of course, 
we'd like to have low anchor games spot on too, but it may be a trade off. The 
latter are more likely to be drops (excepting certain match scores), so perhaps 
it doesn't matter quite so much.

Some people think of positions with 2 high anchors as holding games rather than 
back games (because your shot comes much earlier & you don't have the same 
amount of timing) so perhaps the last part of the definitions should be AND NOT 
2ndAnchor[1-3]. With this definition, a 45 or 47 game would be classed as a 
holding game and included in the net. Positions such as 51 & 42 games would be 
excluded. (I had a 57 game this weekend and it certainly felt more like a 
holding game than a backgame.)

Six point games are so rare that including them might detract from the overall 
performance.

Another possible criterion is to set a limit on the maximum number of chequers 
back, say 3. Once you get to four chequers on the five point you can still get 
gammoned reasonably often so perhaps they shouldn't count.  If you want to get 
even more subtle, this limit could be ABS(Player0MenBack - Player1MenBack) < n. 

Alternatively, you could exclude complicated games with lots of men back by 
limiting (Player0MenBack + Player1MenBack) < n. Setting pip count < 155 might 
do the job more simply. ( I suggest this number because running both back men 
out to the midpoint with 65 65 is a pip-count of 155, and would well end up as 
a holding game.)

You may also like to consider the outfield. Is it still a holding game if you 
have a second anchor in your opponents outfield. (18 through 14 points) How 
about a single chequer?

Finally, should we consider the other side of the board. Have you still got a 
holding game if your board has crunched, or is it just a mess. I would not 
favour this criterion. If your board is crunched then you are probably advanced 
enough to have decent racing chances.

It might be worth generating a few sample positions using some of these rules 
and see if what comes out LOOKS like a holding game to us.

These are my initial thoughts. I hope they stimulate some discussion. Please 
comment, even if you simply agree.

--Ian




reply via email to

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