Opening roll of 4-3

Jason Hyde

unread,

May 2, 1997, 3:00:00 AM5/2/97

to

In an article that Kit Woolsey wrote he said that on an opening roll of 4-3
that 24/20, 13/10 and 13/10, 13/9 were just about tied and 24/21, 13/9 was
in third. I have been playing the JF beta, and it plays the latter on all
of its opening 4-3s. What is the best opening 4-3 as far as roll outs go?

Jason Hyde

Morten Wang

unread,

May 2, 1997, 3:00:00 AM5/2/97

to

["Jason Hyde" <jh...@ix.netcom.com>]

As far as my $.02 goes I would say that is mostly depends on playing
style. I play an opening 4-3 13/9, 24/21 instead of 13/10, 24/20
because I do not like to have all my builder 2 points apart on my side
of the table. I also do not play 13/10, 13/9 since that can quickly
give me a stripped 13-point, which I do not like either. And that's
about it.

Morten!

--
anima sana in corpore sano

Geoff Oliver

unread,

May 3, 1997, 3:00:00 AM5/3/97

to

Jason Hyde <jh...@ix.netcom.com> wrote in article
<01bc5714$05b2d620$507c...@jhyde.dal2.dttus.com>...

> In an article that Kit Woolsey wrote he said that on an opening roll of
4-3
> that 24/20, 13/10 and 13/10, 13/9 were just about tied and 24/21, 13/9
was
> in third. I have been playing the JF beta, and it plays the latter on
all
> of its opening 4-3s. What is the best opening 4-3 as far as roll outs
go?
>

> Jason Hyde
>
I don't think there is ONE best move.
I think that part of the answer lies in which way you want the game to go.
ie How important are gammons to each player ? Steering a game in a
particular direction - even from the opening roll - is very important and
JF appears to bear this out.
I''ve been looking at all the openings at most (but not all - yet) match
scores for a 5-point match and the 4-3 opening you ask about is a
case in point where JF varies it's opening :

At level 7 JF plays 13-9, 24-21 for scores of 0-0, 1-4 (post-Crawford), 4-1
(post-Crawford), or Single game (with Jackoby) but 24-20, 13-10 at scores
of
0-4, 4-0, or 4-4. As I was expecting a difference between plays at
post-Crawford scores of 1-4 and 4-1 if any I'll have to leave this to the
gurus to interprete....

This is very different from JF 2.01 which plays 13-9, 13-10 on all the
above match scores so I doubt we can rely on rollout data from the old JF
as
much we'd like.

re My FAVOURIVE opening roll of 6-4..... My friends will know how relieved
I am that JF still makes the 2-point with an opening 6-4 !!! Well, at 1-4
post-Crawford, anyway !

PS I'm finding the new JF considerably more difficult to beat in matches
than 2.01. Is it really THAT much better or is it me ? Has JF 3.0 played
2.01 ? And if so is there any data on it ?
--
Geoff Oliver
Ealing, London, UK

Stuart Katz, MD

unread,

May 3, 1997, 3:00:00 AM5/3/97

to

"Jason Hyde" <jh...@ix.netcom.com> wrote:

>In an article that Kit Woolsey wrote he said that on an opening roll of 4-3
>that 24/20, 13/10 and 13/10, 13/9 were just about tied and 24/21, 13/9 was
>in third. I have been playing the JF beta, and it plays the latter on all
>of its opening 4-3s. What is the best opening 4-3 as far as roll outs go?

The current popular wisdom is that the moves are indeed a tossup for
all practical purposes. Bringing down 2 leads to more gammmons and a
thin edge in net equity while the splitting moves are thought to lead
to a marginally better winning percentage at the cost of minimal net
equity. I was under the impression that between the splitting plays
13/9 24/21 was preferable.

Maybe someone will commit the time to rollouts but I would guess you
would need a huge number of trials to generate a significant result
given the likely minimal differences.

For historical purposes I have reprinted the opening 43s from the
first 200 5-point matches between JF and ML which I think were played
in 1995. JF always brought down 2 except when it was -1 where gammons
its way had no extra value; it made no change when ML was -1. It never
encountered a 43 opening at -2/-2. When JF split it played 24/21
13/10.

ML always brought down 2, even when it or JF was -1. The sole
exception was a match at -2/-2 where it played 24/20 13/10.

* Crawford game; ! post Crawford

Stuart

Chuck Bower

unread,

May 7, 1997, 3:00:00 AM5/7/97

to

In article <01bc5714$05b2d620$507c...@jhyde.dal2.dttus.com>,
Jason Hyde <jh...@ix.netcom.com> wrote:

>In an article that Kit Woolsey wrote he said that on an opening roll of 4-3
>that 24/20, 13/10 and 13/10, 13/9 were just about tied and 24/21, 13/9 was
>in third. I have been playing the JF beta, and it plays the latter on all
>of its opening 4-3s. What is the best opening 4-3 as far as roll outs go?

I will answer this and the following (how to play 63...) post together.
First off, MOST of the "correct" plays for opening rolls are undecided,
still. That is because alternatives are very close, and because the
software isn't yet good enough to determine the best move that early in
the game (given that it's a close call). Based on JF v2.01 level-6
cubeless rollouts, here is a summary:

"The jury is IN":

31, 42, 61, 65, 53 (8/3, 5/3). Although it is possible that making the
3-point with 53 could SOME DAY be shown to be wrong, it will likely take
a MAJOR change in BG styles. I'd bet against that. Same comment applies
with playing 24/18, 13/8 with 65.

"Split vs. slot":

21, 41, 51. Here it is too close to call between splitting (24/23) and
slotting (6/5). Also, there are other alternatives (24/21 with 21,
24/20, 24/23 with 41, and 24/18 with 51) that are still in the running.
Split AND slot plays (for example, 24/20, 6/5 with 41) are long shots,
at best (similar to the 65 comments above).

"Split + builder" vs. "double split" vs. "double builders"

32, 43, 52, 54. The "double split" (for example, 24/20, 24/21 with 43)
is running in 3rd place, but still alive and kicking. The other two
styles (for example, 24/21, 13/11 vs. 13/10, 13/11 with the 32 opening)
are close. For the 43, both split+builder plays (24/20, or 24/21 and
then the other checker off the 13-point) are candidates. 13/6 looks
dead.

"Run or harass"

62, 63, 64. Close here, too. Making the 2-point with 64 is doing quite
well, also. Starting the 5-point with 62 is on its last breath...
Slotting the bar-point (ala Becker)--pressurized oxygen isn't enough.

Chuck
bo...@bigbang.astro.indiana.edu
c_ray on FIBS

JCDill

unread,

May 7, 1997, 3:00:00 AM5/7/97

to

:bo...@bigbang.astro.indiana.edu (Chuck Bower) mentions:

ż62, 63, 64. Close here, too. Making the 2-point with 64 is doing quite
żwell, also. Starting the 5-point with 62 is on its last breath...
żSlotting the bar-point (ala Becker)--pressurized oxygen isn't enough.

Boy, am I *out* of it!

What is the "popular" 6-2 move, if not to start the 5-point? You run? You
can be hit with a 4 either way, but if NOT hit the blot on the 5-point is
far more valuable to building, why start a running game?

jcd
--

My first name has 2 letters, There's only one letter "J".
My last name's a bit longer, It just worked out that way.
I hate spam, and UCE, and all that unwanted mess.
So to reply, you have to edit my return address.

Chuck Bower

unread,

May 7, 1997, 3:00:00 AM5/7/97

to

>:bo...@bigbang.astro.indiana.edu (Chuck Bower) mentions:

>
>>62, 63, 64. Close here, too. Making the 2-point with 64 is doing quite

>>well, also. Starting the 5-point with 62 is on its last breath...

>>Slotting the bar-point (ala Becker)--pressurized oxygen isn't enough.
>

In article <3371ba6c...@news.concentric.net>,
JCDill <jjjj...@ix.netcom.com> wrote:

>Boy, am I *out* of it!
>
>What is the "popular" 6-2 move, if not to start the 5-point? You run? You
>can be hit with a 4 either way, but if NOT hit the blot on the 5-point is
>far more valuable to building, why start a running game?

Here are the answers, but whether you believe them is another matter...

1) "Popular" among humans is 24/18, 13/11.

2) "You run?" being 24/16 is also under serious consideration, based on
rollouts by Jellyfish v2.01 (and other motivations as well). 13/5 is
still seen occasionally (by human players). Yesterday my FIBS opp.
played 13/7, 13/11. Could s/he have been the elusive Becker?????

Here are Jellyfish v2.01 level-6 cubeless rollout equities (for opener):

equity standard deviation statistical chances

24/16 +0.018 0.013 85%
24/18, 13/11 -0.001 0.013 15%
13/5 -0.034 0.014 <0.1%

(NOTE: "statistical chances" means that if all 3 candidates were rolled
out "an infinite number of times" (instead of the actual 504 times), this
is the probability that the given play would come out "best".)

I could have done more rollouts, but feel that is pointless. JF
doesn't play "perfect" backgammon, so better statistical significance
will still leave systematic uncertainty. I'd rather wait for JF v3.0
and use the computer cycles for something better in the meantime.

The argument "both 24/16 and 13/5 get hit by a 4" is unfortunately
only qualitative. When hit, the 13/5 play is considerably worse, and
if missed, considerably better. (The above argument glosses over the
considerably worse branch.) If you let JF do the QUANTITATIVE analysis,
combining all 36 outcomes and weighting them accordingly, 13/5 comes
in a distant 3rd, as seen above.

Could 13/5 be the best play in the long run? As always, it does
matter how you define "best". But, if you are more specific and ask:

If supercomputerbackgammonplayer (the winner of the Backgammon
Championship of the Universe--open to humans AND robots) plays
itself a trillion games with each of the opening 62 plays, which
play comes out best?

My guess is that 13/5 will NOT work out best. But it could. Right now
it doesn't seem to be in favor by the robots or most of their more fickle
charcoal brained cousins.