bug-gnubg
[Top][All Lists]
Advanced

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

[Bug-gnubg] Strange cubeless rollout result?


From: Michaeldepreli
Subject: [Bug-gnubg] Strange cubeless rollout result?
Date: Wed, 8 Jan 2003 03:32:46 EST

Hi
I set up the following position and wanted to roll it out CUBELESS.

Position ID vHsXAADCNhQCNw
Match ID MAEAAAAAAAAA

So I selected:
Settings---Options---uncheck "use doubling cube"
However I *forgot* to uncheck "use Jacoby rule"

First question:

If I'm doing a cubeless rollout, then I want G and B to count, so why do I still need to uncheck "use Jacoby rule" ?

Now for the strange part.

I did 32000 0-ply rollouts for this position. The reason is, I'm doing a study between GNU, SW3.2, SW4 and JF to see how they play this position
at 0, 1 and 2 ply.

Here's the results I got when I forgot to uncheck "use Jacoby rule"
(I adjusted the cubeless equity to score G and BG)

-0.968 32.2/1.2/0.0---67.8/37.9/24.6

Compared to the other three Bots GNU didn't do so well at 0-ply.

It only managed 32.2 % wins versus SW3.2(40.4) JF(40.2) SW4(41.9)

The thing that did appear strange was the BG figure.

GNU(24.6) SW3.2(13.0) JF(13.6) SW4(12.8).

I ran the rollout again this time I unchecked "use Jacoby rule"
Here's the results:

-0.736 34.9/1.0/0.0---65.1/34.3/11.3

These figures look more realistic and makes me suspicious of the first rollout with "cubeless AND use jacoby rule".

Am I missing something obvious here?

Is GNU acting as if the cube is in play and the Jacoby rule is in effect and estimating W/G/BG breakdowns when the game is cashed (due to too good to double) had the game gone to conclusion? Like a kind of pseudo-truncation?

Thanks

reply via email to

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