bug-gnubg
[Top][All Lists]
Advanced

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

[Bug-gnubg] Analysis Settings


From: Albert Silver
Subject: [Bug-gnubg] Analysis Settings
Date: Mon, 17 Feb 2003 13:19:29 -0300

I've been running some tests to compare the results of time and
precision of 2-ply analysis. That the current World Class settings are
less precise than the Supremo Settings is clear, but the difference in
times was also very significant.

Just as a reminder, the World class settings analyzes up to 8 moves
within 0.160 equity of the top 0-ply move. Supremo, the next notch up in
the presets, analyzes up to 16 moves with 0.320 equity. 

A sample match with 250 plies took exactly 30 minutes (on my weenie
laptop) using the Supremo settings whereas it only took 14 minutes and
47 seconds using the World Class settings. Unfortunately, the Checker
play analysis also presented differences, and I found other cases where
this was true. As a means of comparison BTW, Snowie 4's 3-ply analysis
took no less than 44 minutes 06 seconds. Ouch.

I went about trying to find a less time-consuming setting that didn't
sacrifice quality. My best result so far has been one with 2 forced
moves (I think one must at least have one other move to compare with for
a 2-ply analysis to make any sense) plus 10 at 0.300 equity difference.
Until now I have not found one analyzed match (I have tested 10 so far
for about 2000-2500 plies average) that presented different results from
the Supremo ones, however it did reduce the analysis time of that same
match above for example to 23 minutes 39 seconds. Unless someone finds
conflicting results, I'd suggest this be the default World Class or
Supremo setting.

                                                Albert






reply via email to

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