Forum Archive :
GNU Backgammon
> Does anyone know why the evaluations of GNU BG for 2,3,4 ply seem to
> differ, with 2 and 4 ply often seeming to agree closely and 3 ply
> being different.
> Am I just imagining it?
> Is one 'more reliable' than the other?
Is one 'more reliable' than the other?
Several readers (not me) have devoted lots of time to this and could
answer more thoroughly (they have in the past). The answer seems to
depend on
(a) whether you want to find the best play or cube decision or want
reliable reported equities.
(b) whether you're talking about evaluations or rollouts
Checker play:
When gnu plays itself on 0-, 1- and 2-ply, the higher ply wins. I don't
know that similar tests have shown 3-ply to be significantly better than
2-ply, or 4-ply significantly better than 3-ply. However, in reviewing
my own matches I find many positions in which the 3-ply play is either
(a) better in rollout or (b) is in my judgment the better play. Most
examples that I have noticed of (b) are bear-in positions, and I judge
that 2-ply is definitely (and annoyingly) weaker in this phase of the
game. With regard to (a) I should note that I have also found positions
in which the 2-ply play is correct, and some where neither 2- nor 3-ply
find the best play.
Reported Equities:
Generally, 1-ply and 3-ply equities are similar, and 0-, 2- and 4-ply
equities are similar. The odd-plied equities are often strange, off the
wall. My impression is that even when 3-ply finds a better play than 2-
ply, the 2-ply equity tends to be more accurate.
I have not seen, and would like to see, a comparison of 2- and 3-ply
reported equities of opening roll and response positions which have been
rolled out extensively.
Cube decisions:
Because the odd-ply equities are often off the wall, you get much better
cube decisions with 0- and of course 2-ply. In Practice
For playing against the computer, I use only 2-ply checker / 2-ply cube.
This is because 3-ply checker play is too slow for me.
For analysis, I never use anything but 2-ply cube. I usually use 2-ply
checker play with selective 3-ply reevaluation of interesting plays 2-
ply has marked as an error. No doubt I am missing some 2-ply errors
where I have unfortunately made the same mistake as 2-ply. But speed in
analysis is usually important, and also I am usually interested in
observing equities of positions even when there is no cube decision, and
it is my impression that in general I'll get better equity reporting
from 2-ply match analysis.
For rollouts, I use 0-ply with 2-ply cube and 2-ply with 2-ply cube. 3-
ply play is much too slow. There are lots of positions for which 0-ply
checker play is worthless, and there's no good reason other than a
slight difference in speed that I don't use 1-ply play instead. I
probably should, since I'm convinced 1-ply plays significantly better.
4-ply is just too slow to use for analysis or play. From what I've seen
my impression is that 4-ply tends to confirm (a) even-ply cube decisions
and (b) 3-ply checker play, where 2- and 3-ply differ, but with regard
to (b) this impression is purely anecdotal; I haven't seen any rigorous
evidence to support it.
Generally, though, if one were to aspire to play like a bot, my
impression is that 3-ply checker play would be a better goal than 2-ply.
|
|
|
|
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
|
| |
|