Snowie Rolout Parameterts?
Daniel Murphy
rollout parameter settings:
All Play decisions:
Truncated rollout, depth 5,
2592 games (equiv. 75720 games),
played 1-ply according to score,
seed 1, with race database.
All Cube decisions:
Mini-Rollout
Truncated rollout, depth 7,
108 games (equiv. 14220 games),
played 2-ply according to score (tiny),
seed 1, with race database.
With these parameters:
Should I believe the results? -- Never, usually, (almost) always?
How might the parameters be improved, assuming I can tolerate:
0% increase in computing time required
50% increase in computing time required
100% increase in computing time required
unlimited increase in computing time required
Thanks,
________________________________________________
Daniel Murphy www.cityraccoon.com/
Humlebæk Backgammon Klub www.hbgk.dk/
Raccoon on FIBS www.fibs.com/
Raccoon on GamesGrid too
Daniel Murphy
Yeah, yeah ... dig those eye-catching spelling errors in the subject
line ....
tony
I have Snowie2 but I cannot run the program because I need Keylib32.dllm
could you please send me that file via mail??
Thanks for your help.
TONY
Daniel Murphy wrote:
> A Snowie owner sent me a match that he had Snowie analyze using these
> rollout parameter settings:
>
> All Play decisions:
> Truncated rollout, depth 5,
> 2592 games (equiv. 75720 games),
> played 1-ply according to score,
> seed 1, with race database.
>
> All Cube decisions:
> Mini-Rollout
> Truncated rollout, depth 7,
> 108 games (equiv. 14220 games),
> played 2-ply according to score (tiny),
> seed 1, with race database.
>
> With these parameters:
> Should I believe the results? -- Never, usually, (almost) always?
>
> How might the parameters be improved, assuming I can tolerate:
> 0% increase in computing time required
> 50% increase in computing time required
> 100% increase in computing time required
> unlimited increase in computing time required
>
> Thanks,
>
Douglas Zare
> Yeah, yeah ... dig those eye-catching spelling errors in the subject
> line ....
However foolish and ignorant someone might appear because of spelling
errors, you make yourself look just as bad by pointing them out.
:)
/\
||
Douglas Zare
Douglas Zare
> [...]
> All Play decisions:
> Truncated rollout, depth 5,
> 2592 games (equiv. 75720 games),
> played 1-ply according to score,
> seed 1, with race database.
>
> All Cube decisions:
> [...]
I can't help with your interesting question, but I have a question about
this terminology. Does "cube decision" include the decision not to
double? Would it be called more than one error not to double any time
during a long sequence in which one should, e.g., while closing out
someone who is primed but has a few off?
While one pays for errors in play (or taking/dropping) seperately,
errors in not doubling don't add up in the same way.
Douglas Zare
Joe Loria
and bring one man down from the 13-point with the 5 or 3 ?
I like the resulting distribution of my men, can build my
board a lot quicker if not hit, and would rather get hit by
my opponent's 3-1 or 1-1 reply than see them make points
on their inner board.
The way I see it, it's best to take these risks early in the game,
and I like to unstack my six point with 1's and 2's. Also, if my
opponent slots their five point on the opening roll, then slotting my
four point duplicates his threes.
Donald Kahn
wrote:
13/8 6/4 ia surely very tempting, and the story at least goes that
Joseph Dwek, a very top player of the 70s-80s, adopted it for awhile.
Also according to the story, he found the results inferior, and
returned to conventional methods. To me it is very playable and maybe
best when opponent has slotted his 5 point and needs the 3 to cover.
with 3-2, 13/10 6/4 is a play I might make when I looking for gammons,
(though two down from the midpoint is considered the biggest
aggressive play) and don't much care if I get gammoned. At ordinary
scores, seems to me that 24/21 13/11, doing useful work on both sides
of the board, deserves its superior rating.
dk
ti...@globalnet.co.uk
The reasoning you give all seems perfectly valid, although I
understand the best computer programs don't play this way ...
But if you're going to play like this, why not (with a 3-2) bring the
2 down from the midpoint and slot on the 5-point with the 3, rather
than slotting on the 4-point with the 2? Isn't the 5-point a more
valuable one to make on the next roll, if your blot isn't hit?
This was the play recommended by B.J. Becker in his book "Backgammon
For Blood" from the 70's, and is what I always used to play.
(Becker's book also contains other interesting recommendations not
validated by computer roll-outs.)
tnn
Joe Loria
On Tue, 26 Oct 1999 ti...@globalnet.co.uk wrote:
> The reasoning you give all seems perfectly valid, although I
> understand the best computer programs don't play this way ...
Yes, Jellyfish doesn't play this way, and I suspect the others
don't either. But in chess, one doesn't always open P-K4
(except for Fisher prior to 1972!), so why not vary your
opening choice in backgammon ? For one thing it puts your
opponent into new territory and they may misplay the position.
> But if you're going to play like this, why not (with a 3-2) bring the
> 2 down from the midpoint and slot on the 5-point with the 3, rather
> than slotting on the 4-point with the 2? Isn't the 5-point a more
> valuable one to make on the next roll, if your blot isn't hit?
Yes the 5-point is more valuable, but I don't think there is
that much difference. I'm going to try for the 5-point eventually,
but not right now, if the dice aren't there. I just want to unstack
my 6-point at this stage of the game, and not strip the 8-point.
I think there's a lot to be said for the "distribution" of men early
in the game, although I haven't read anything about that concept.
Another element of the early game is what I call "sacrificing
a man", as when slotting my 4-point when the opponent has their
5-point slotted. I'd rather get into a blot-hitting contest,
than watch passively as the opponent builds a better board than me.
Another thing I like about slotting vs splitting is that, if
not hit, I may be able to use 1/2 of my next roll to cover
my man, and the other 1/2 to do something else, like split the
back men, or hit, or whatever.
> This was the play recommended by B.J. Becker in his book "Backgammon
> For Blood" from the 70's, and is what I always used to play.
>
> (Becker's book also contains other interesting recommendations not
> validated by computer roll-outs.)
>
Yes, Becker had some radical concepts! His recommendations for
the 6-x openings mostly recommended slotting his bar point. I'm
not that keen on making my bar point, the mid-point gets stripped
early, and the opponent gets to make good use of their sixes
in reply. I somewhat agree with Becker about not splitting the
back men too early, as it seems I've been successfully blitzed
by Jellyfish more ofen than I like. It may be that I don't
have a good feel for that position, and am still stuck in my old
slotting ways. :) Well at least I sometimes make the 2-point
with a 6-4, which I never even considered before JellyFish
and the other bots said it was okay.
Chuck Bower
<ti...@globalnet.co.uk> wrote:
(snip)
>...why not (with a 3-2) bring the
>2 down from the midpoint and slot on the 5-point with the 3, rather
>than slotting on the 4-point with the 2? Isn't the 5-point a more
>valuable one to make on the next roll, if your blot isn't hit?
>
>This was the play recommended by B.J. Becker in his book "Backgammon
>For Blood" from the 70's, and is what I always used to play.
WHOA! B.J. Becker was a championship level bridge player for
many decades. He DID NOT write BACKGAMMON FOR BLOOD. 'Bruce Becker'
was a pen name (that is, apparently not a real name) of the author of
BG FOR BLOOD.
>(Becker's book also contains other interesting recommendations not
>validated by computer roll-outs.)
Not only 'not validated', but more accurately invalidated, at least
at the robot vs. robot level of play.
Chuck
bo...@bigbang.astro.indiana.edu
c_ray on FIBS