Forum Archive :
GNU Backgammon
I use Snowie, but since everyone is using GNU now, whether due to its
play & ability to play to match scores and other bells & whistles, or
because of the lack of price, I loaded it on a couple of the computers
at my house. I'm not finding it as 'user friendly' as Snowie, I'm
getting by, but wondering what are the standard 'ideal' settings for a
rollout?
Stick
|
|
happyjuggler0 writes:
I'm sure others will chime in but some basics here anyway.
Gnu 2ply is the exact same thing as Snowie 3ply, and Gnu 0ply is the
same as Snowie 1ply.
I recommend rollouts in multiples of 1296. I personally don't use the
jsd cutoff but when I am paying attention after about 100 trials I will
manually stop the rollout any time jsd's exceed 3 and eliminate that
move. Then I will continue the rollout.
Most rollouts do indeed take a while when there is a lot of contact
left, but are rather fast with minimal contact.
Rollouts use the rollout tab under Settings. There are several types of
rollouts you can do: 1)cubeless, 2) truncated, 3) stepped, 4) full. If
you are doing rollouts you should be searching for some kind of "truth",
and usually truncated rollouts are a bad idea. Make sure the box labeled
truncation under general settings is unchecked if you are not using
truncated rollouts.
Use cubeless when the cube is dead. Enable the two-sided and one-sided
databases, they will save time and also be a tiny bit more accurate.
Stepped rollouts are good when you think the next several moves will be
tricky but after that will be easy. You then use 2ply for the first
several moves and then 0ply after that. Use first play both for setting
the first several moves at 2ply. Use later play both for the later
moves. To determine the number of moves using your first setting go back
to the general settings tab and click enable seperate evaluations.
Full rollouts are basically the same as stepped rollout but you never
use the later play both tab. An typical example would be user defined
settings at 2ply, then for move filter use modify, then click on 2ply
and click the 0ply enable this level and unclick the 1ply enable this
level box. "Always accept 0 moves and add 8 moves within .160" is a
fairly standard setting on GOL.
2ply cube is slower than 0ply cube, but not terribly so and seriously
affects the outcome. 2ply checker play is much slower than 0ply checker
play. For this reason you will often see 0ply checker/2ply cube rollouts
on GOL.
Use variance reduction and quasirandom dice.
|
|
Peter Grotrian writes:
One think to consider, once you are familiar with the general rollout
procedure, is cutting down the move filter for special scores. The most
prominent example is DMP where "world class" with 8 moves within 0.05
didn't choose one single move different from pure "world class" in a
test match (including a complex backgame) from Michael Depreli. If you
don't cut down the filter at DMP, the rollout will indeed be as fast as
running in a swamp. Some are more conservative, e.g. I think Zorba uses
(among other settings) 8 moves within 0.08 at DMP.
The reasons for cutting down the filter is that errors comparable to
0.16 at money are much smaller at DMP because the lack of the cube and
gammons.
Do you use an SSE build of GNUBG? If not, start now! The speed
improvement is almost twofold! You will like this when doing massive
rollouts :-). http://www.gnubg.org/media/gnubg-setup.exe
Peter
|
|
Andreas Graf writes:
I downloaded the SSE Version from your link but during installation (Win
XP SP2) there always comes an error message:
C:\Programme\gnubg\gnubg_ts0.bd The source file is corrupted.
Has anybobdy else this error also? Thanks
|
|
Ian Dunstan writes:
I got that error message and one other similar message involving both
gnubg_os0.bd and gnubg_ts0.bd. I ignored them both and let Gnu build
these databases for me (automatically) at the end of the initial
download. My installation seems to work ok.
--
The latest versions of Gnubg utilise a "pruning net" and in my
experience provide much faster rollouts than the equivalent Snowie4
settings.
Rollout settings: Here is a quick overview from my perspective. You
almost always want to use full rollout settings, not truncated. The
following settings are based on full rollouts.
0-ply move/2-ply cube. This is about as low as you want to go for a
"good indication" rollout. 0-ply cube is generally to be avoided, it's
unreliable and the rollout time overhead between 0-ply cube and 2-ply
cube is usually quite reasonable anyway. Admittedly, 0-ply move/0-ply
cube rollouts really fly, so if you're checking to see whether move B
really is a blunder vs move A, they can be a useful first approximation
to the truth.
2-ply move/2-ply cube. Gnubg's "World Class" settings (8 moves, 0.16
filter) are very strong, but do take a long time. If I'm doing a
straight 2-ply move rollout, I tend to relax the filter settings a
little and use a (6 moves, 0.12 filter). This is ok IMO because (6,
0.12) did just about as well as (8, 0.16) in Michael Depreli's tests
(GoL archive), so I sometimes choose to sacrifice a tiny amount of
accuracy for an improvement in trials/per hour.
2-ply (Stepped)/2-ply cube. This is my favourite setting: 2-ply(6,0.12)
move for the first 12 moves (6 each player), and then reverting to 0-ply
move for the rest of the rollout. 2-ply cube gets used all the way, of
course. The idea with a stepped rollout is that from any given position,
often much of the equity will be attributable to how well the next few
rolls are played, so beef-up the start of the rollout and settle with 0-
ply moves later. In my experience, this type of rollout will nearly
always provide results somewhere between straight 0-ply and straight 2-
ply. Of course, the main reason to do a rollout like this is to try and
get the best compromise between accuracy and trials/per hour. I think
Stepped rollouts are worth doing, but I'm the only one on this forum who
regularly posts rollouts performed in this way, so some others here will
probably disagree with me.
|
|
|
|
GNU Backgammon
- Analyzing GamesGrid matches (Roy Passfield, Dec 2001)
- Batch analysis tool (Øystein Johansen, June 2004)
- Cache size (Ned Cross+, Mar 2004)
- Compiling for Windows (Øystein Johansen, Jan 2002)
- Edit mode removing checker from bar (Scott Steiner+, May 2003)
- Entering an annotated match (Albert Silver, Dec 2003)
- Error rates: Gnu vs. Snowie (Raccoon, Mar 2006)
- Even-ply/odd-ply effect (Raccoon, Nov 2004)
- Even-ply/odd-ply effect (Tom Keith+, Oct 2003)
- Even-ply/odd-ply effect (Scott Steiner+, Dec 2002)
- Filter settings (Robert-Jan Veldhuizen, Nov 2004)
- Gnu 0.13 versus Jellyfish and Snowie (Torsten Schoop, Aug 2003)
- Gnu 0.13 vs. Snowie 4 (Albert Silver, June 2003)
- Gnu 0.14 vs. Jellyfish (Michael Howard+, July 2003)
- Gnu versus Snowie and Jellyfish (Michael Depreli, Oct 2005)
- How luck factor is calculated (Gregg Cattanach, Aug 2002)
- How rollouts work (Gary Wong, July 1999)
- How to enter an illegal move (Øystein Johansen, Aug 2003)
- Importing .gam files (PAR+, Mar 2005)
- Importing PartyGammon matches (rew+, July 2006)
- Improving your game using GnuBG (D.U.G.+, Nov 2002)
- Installing on Windows (maareyes, Oct 2001)
- Interpreting JSD's (Adrian Wright+, Feb 2005)
- JSD's and confidence intervals (Daniel Murphy+, Jan 2005)
- Logging rollouts (Øystein Johansen, Oct 2004)
- Luck rate (Kees van den Doel+, May 2002)
- MWC versus Equity (EMG) (Ken+, Apr 2005)
- Manually entering first roll (Andreas Graf+, Apr 2005)
- Match equity tables (Raccoon, July 2005)
- Personal reflections (Louis Nardy Pillards, Sept 2002)
- Playing two computers against each other (Stanley E. Richards+, Mar 2008)
- Python scripting (Øystein Johansen+, Nov 2004)
- Quasi-random dice in rollouts (Ian Shaw, Mar 2004)
- Question marks in game list (Jim Segrave, July 2005)
- Questions and answers (Jim Segrave+, Jan 2003)
- Questions and answers (Jørn Thyssen, Aug 2002)
- Restarting a rollout with different settings (Jim Segrave, Apr 2005)
- Restarting a rollout with different settings (Robert-Jan Veldhuizen, Apr 2004)
- Rollout settings (geoff arnold+, Apr 2007)
- Rollout settings (Stick+, Nov 2005)
- Rollout settings (Robert-Jan Veldhuizen, Mar 2004)
- Rollout settings (Ian Dunstan, Aug 2003)
- Rollout settings for the impatient (Robert-Jan Veldhuizen, June 2004)
- Running rollouts in background (Bruce+, Apr 2004)
- Saving rollout results from command-line interface (Jeremy Bagai+, Apr 2006)
- Saving rollouts (Mislav Radica+, May 2006)
- Setting GnuBG's playing strength (JP White, Sept 2001)
- Setting skill level (Jim Segrave, Apr 2004)
- Setting up and saving a rollout (Albert Silver, Dec 2003)
- What's GNU? (Gary Wong, Oct 2001)
- Which player is player 0? (Neil Kazaross+, Oct 2004)
From GammOnLine
Long message
Recommended reading
Recent addition
|
| |
|