Gammonfever BG Quiz - Problem #4
Micke Nilsson
good or too bad)?
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O X X X | | X |
| X O O | | |
| O O | | | S
| O O | | | n
| O | | | o
| |BAR| | w
| | | | i
| | | | e
| | | |
| X X X X | | |
| X X X X O | | X |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 163 O: 41 X-O: 0-0/Money (1)
Men Off X: 0 O: 5
CubeValue: 2, O owns Cube
paulde...@att.net
equity.
5 men off means that X's closeout (a very likely possibility) would
lead to a strong double (to 8) for X which would either be a marginal
take or marginal pass for O (depending on the precise position at the
time).
Not at all good enough to double. I feel sure that X should beaver if
O doubles. (But I've been sure so many times and yet been wrong.)
Actually, even with O's cube ownership, I slightly prefer X. If I was
X, I would not want to settle even, and would need to receive payment
(admittedly very slight) to abandon the game.
Paul Epstein
Jim Segrave
Gnubg thinks it's a huge error not to redouble (estimated error 0.199
EMG) and that it's a small pass (0.023 EMG). Winning chances are only
58%, but of those 40% are gammons. I've started a rollout, but it's
way too early to tell - after 22 games it's just 0.050 in favour of
doubling, but with a congidence interval of about 0.100.
--
Jim Segrave (j...@jes-2.demon.nl)
David C. Ullrich
>X's 4 point board and huge chances for shots outweigh O's gammon
>equity.
>5 men off means that X's closeout (a very likely possibility) would
>lead to a strong double (to 8) for X which would either be a marginal
>take or marginal pass for O (depending on the precise position at the
>time).
>
>Not at all good enough to double. I feel sure that X should beaver if
>O doubles.
You may be right. I'm not sure of anything here, but the idea that
O is not good enough to double seems wrong to me. I have no idea
whether it's too good to double - maybe not, because X does have
good chances to win, and I have no idea whether X should take.
But (although I could be wrong) the idea that O is not good enough
really doesn't seem right to me - this is the one aspect of it
where I actually have an opinion<g>. Because all O needs is one
good roll, and it seems to me he's going to have several chances
to roll that good roll, because X doesn't have material in place
to finish the closeout all that soon.
If X had his 6 point, or if O were on the bar already, or
if it were X's roll, various things might make it different.
But here it just seems to me too easy for O to win: Maybe O
rolls a 6 this time and X can't hit back because of the duplication
(X needs a 1, 4, 5, or 6 to come in, which may well be what
he needs to hit - he rolls a 2 or a 3 and he probably
misses.) Or O comes out and X misses just once, should be
enough to get O home.
************************
David C. Ullrich
paulde...@att.net
Yours is a good news/ bad news posting.
Good because it shows a real weakness in my game. My assessments of
these positions is wildly wrong apparently, and knowing this gives me
something to focus on to improve.
The bad aspect is that it doesn't respect the explicit request of the
original poster: "No bot solutions please..." (A paraphrase not a 100%
exact quote.)
I think that that request should have been honoured. Of course, the
bots get the final word, and such a posting would be o.k. after 2 weeks
or so, but others had very little chance to form an opinion for
themselves before you gave an authoritative answer.
Paul Epstein
paulde...@att.net
posting. Clear double, marginal take/pass.
Such positions are vital knowledge in the money game because the cube
tends to be at quite a high level in these positions.
You must be quite deadly in chouettes!
Paul Epstein
Micke Nilsson
any bot answers right away after I posted it. The reason I'm putting this
series of problems out is to give people a chance to think them over and
come to their own conclusions. When you know what the bots say it's right,
it's hard to argue for something else. One of the reasons for posting this
series is to see if you all think that these are positions many players
would get right at the board, like Gammonfever did. This goes back to my
thread "proof of online cheating?" where Gammonfever was accused of cheating
by a top player after playing 32 money games without making one single cube
error.
I have already rolled out the position with Snowie 4, but will not post the
result here for a few days. After that I will publish the next position in
the series.
Sometimes It's extremely grateful to get a bot answer, but this time I
wasn't looking for one because I needed help to know if I made something
wrong or right.
/Micke
"Jim Segrave" <j...@nl.demon.net> skrev i meddelandet
news:11i84ri...@corp.supernews.com...
Micke Nilsson
time like i did on quiz #3. I'll remember this when I post #5 though. I
suppose Jim has not followed my "Proof of online cheating?" thread and the
posts connected to that, which would make me to perfectly understand his
willingness to help out with the "correct" solution.
Without posting Snowies results (direct 3-ply and rollout) I can say that
they are pretty interesting (at least I think lol) and they don't completely
agree with this fast gnu answer.
/Micke
<paulde...@att.net> skrev i meddelandet
news:1126440732....@z14g2000cwz.googlegroups.com...
paulde...@att.net
Yes, good point. I hope I wasn't too critical of Jim. I thought there
was a lot of good in Jim's posting, and some valuable info and
analysis.
I would probably have been less critical if I had remembered that your
"no bot" request was from a different thread.
Paul
Grunty
think O must make him pay more for trying his chances. I mean, he's not
at all too good to double.
If O hadn't a direct shot at a blot, he should certainly refrain from
doubling.
Now regarding X's action.
X would have a clear take if there weren't so many gammons. Gammons are
increased by the fact that the 6 point is open. So, hitting doesn't
guarantee X a win for a long time, while O may have further chances to
reenter and scramble home. I vote for X to pass.
Jim Segrave
<paulde...@att.net> wrote:
>Hi Micke,
>
>Yes, good point. I hope I wasn't too critical of Jim. I thought there
>was a lot of good in Jim's posting, and some valuable info and
>analysis.
I was not upset. I admit that I hadn't looked at the original posting
too closely and thus, even if the original had had a "no bots, please"
request, I might well have not registered it. I actually only started
paying attention when I looked at your response and wondered about it
- I was less than confident that I knew the answer. And, as I pointed
out in my first posting, gnubg rollouts leave things well open to be
decided - at the moment it's now up to 212 games and there's no
definitive answer, the differences are about 1/2 of a joint standard
deviation, which to all intents and purposes means that it's
undecided.
--
Jim Segrave (j...@jes-2.demon.nl)
David C. Ullrich
Just the opposite - the reason I doubt that it's not
good enough for O to double is all the times I've been
killed by an early redouble after hitting in a back game:
I say aha, O's position has deteriorated to a one or two
point board, no way he can win this, and it seems like
O "always" manages to roll one good number before I
close him out (especially when O is gnubg or Jellyfish,
haha.) I've slowly decided that the problem is as I
think I said: X doesn't need any miracles to win, but
he does need a substantial number of fairly reasonable
things, and he doesn't win with just one miracle here.
On the other hand O needs just one very good roll to
go back to gammoning X, and there are a fairly large
number of very good rolls available right now, will
probably be a few for the next few rolls... (I almost
posted a list of all the things O could roll,
classifying each as bad, medium, good, or very good.
You can do that yourself - you'll find that O really
does have a lot of ways to get back to a "win unless
a miracle happens" position right away, and it does
seem that that will stil be true for several rolls
to come. Or rather that it will be true several
times in the future - maybe X hits and O bounces,
but when O comes in he's again probably going to have
lots of winners, or so it seems to me.)
(All of which is saying that I _really_ think it
would be too early for X to redouble if he owned
the cube, which of course is not the same as the
current question. In my limited experience it
seems more likely that X would be owning the cube
here - I wonder whether there wasn't a position
earlier where O should have cashed instead of
playing for the gammon.)
Ok, I'll admit to being slightly dangerous in
our local chouette, but that's just because all
but one of the other players are much weaker,
not understanding many of what we here all think
of as standard plays.
>Paul Epstein
************************
David C. Ullrich
Micke Nilsson
in to this too much, but I can say that O has not missed to cash earlier in
the game. The reason I feel confident in this without having the whole match
is that O did not make one single cube error in the 32 matches they played.
No missed double or wrong double/take/drop.
/Micke
"David C. Ullrich" <ull...@math.okstate.edu> skrev i meddelandet
news:c5kai1dsf7f8itbc1...@4ax.com...
Raccoon
Micke Nilsson wrote:
> come to their own conclusions. When you know what the bots say it's right,
> it's hard to argue for something else.
And what if you know what the bot says is wrong? Or don't know if it's
right?
If you were suspicious that an opponent were cheating by using a bot,
you might find positions where only -- you think -- a bot or a world
class player would have found the right move. Or you might find
positions -- and they need not be difficult positions -- where your
opponent makes the same mistake that a bot makes in its evaluation.
In your position 4:
> >>> O on roll, cube action
> >>> +-1--2--3--4--5--6--------7--8--9-10-11-12-+
> >>> | X O O X X X | | X |
> >>> | X O O | | |
> >>> | O O | | | S
> >>> | O O | | | n
> >>> | O | | | o
> >>> | |BAR| | w
> >>> | | | | i
> >>> | | | | e
> >>> | | | |
> >>> | X X X X | | |
> >>> | X X X X O | | X |
> >>> +24-23-22-21-20-19-------18-17-16-15-14-13-+
> >>> Pipcount X: 163 O: 41 X-O: 0-0/Money (1)
> >>> Men Off X: 0 O: 5
> >>> CubeValue: 2, O owns Cube
> >>
I redouble. O is sometimes 2 rolls away from a no-brainer 8-point
gammon. But O is not too good. I also take. X has 7 checkers well
positioned to contain O, can sometimes hit and make the 6 point in two
or very few rolls, has excellent recube vig since O has only 5 checkers
off, and in-to-then-end acepoint vig.
Which player are we to be suspicious of here? I see nothing suspicious
in doubling or taking. Not doubling might be suspicious, but Jim has
already revealed (no problem, Jim) that gnubg doubles and drops. If
Snowie does also, then a drop by Player X would seem suspicious to me,
since I'm prop-sure it's a take.
Micke, will you please include Snowie's evaluation as well as rollout
data in the positions in this series?
Jim Segrave
Gnubg's evaluator gave the resutl I initially posted, but, as I noted,
I was already rolling it out. 24 or so hours later, nearing 1296
games, and there's still no answer with two joint standard
deviations. Assuming gnubg plays this well, and it's a rather extreme
position, so it's not impossible that it doesn't play it as well as
some others, then the static evaluator is far too confident in the
value of doubling. If rollouts even make it a double, it looks like it
will be more a 0.030 or so EMG double, not the almost 0.200 or so that
the static evaluator gave. It's still quite possible that it's not
even a double. I suspect that a lot of the play here may require
trying to think 3 or four plies ahead, which is beyond what the neural
nets normally attempt (if I'd tried a four ply rollout, I'd probably
be on the 3 or fourth game now, rather than over 1000).
--
Jim Segrave (j...@jes-2.demon.nl)
Micke Nilsson
results on quiz #3. I do not know if the original accuser found every single
one of these positions possible cheating ones, but I do know that when some
of the top players in Sweden sat down to say how they had done in these
positions they got between 12 and 15 correct answers, according to Snowie
(and I guess this was the 3-ply eval because I don't think the accuser
rolled out 32 complete matches).
I will post the Snowie 4 results directly after my original post in this
thread and then let you have a go at #5.
/Micke
"Raccoon" <racg...@yahoo.com> skrev i meddelandet
news:1126550543.6...@g47g2000cwa.googlegroups.com...
Micke Nilsson
or too bad)?
3-ply Evaluation:
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O X X X | | X |
| X O O | | |
| O O | | | S
| O O | | | n
| O | | | o
| |BAR| | w
| | | | i
| | | | e
| | | |
| X X X X | | |
| X X X X O | | X |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 163 O: 41 X-O: 0-0/Money (1)
Men Off X: 0 O: 5
CubeValue: 2, O owns Cube
3-Ply Money equity: 0,645
9,0% 39,0% 58,3% 41,7% 0,0% 0,0%
1. Redouble, take 0,986
2. No redouble 0,801 (-0,185)
3. Redouble, pass 1,000 (+0,014)
Proper cube action: Redouble, take
------------------------------ End ----------------------------------
Snowie Rollout:
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O X X X | | X |
| X O O | | |
| O O | | | S
| O O | | | n
| O | | | o
| |BAR| | w
| | | | i
| | | | e
| | | |
| X X X X | | |
| X X X X O | | X |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 163 O: 41 X-O: 0-0/Money (1)
Men Off X: 0 O: 5
CubeValue: 2, O owns Cube
Rollout Money equity: 0,725
9,5% 42,1% 60,5% 39,5% 0,0% 0,0%
95% confidence interval:
- money cubeless eq.: 0,725 ą0,036,
- live cube no redouble: 0,858 ą0,044,
- live cube redouble take: 1,089 ą0,091.
Rollout settings:
Full rollout,
324 games (equiv. 4741 games),
played 2-ply (fast), cube 2-ply,
settlement 0,550 at 16 pts,
seed 1, without race database.
Evaluations
1. Redouble, pass 1,000
2. No redouble 0,879 (-0,121)
3. Redouble, take 1,161 (+0,161)
Proper cube action: Redouble, pass
Live cube
1. Redouble, pass 1,000
2. No redouble 0,858 (-0,142)
3. Redouble, take 1,089 (+0,089)
Proper cube action: Redouble, pass
------------------------------ End ----------------------------------
Grunty
> 3-ply Evaluation:
> Proper cube action: Redouble, take
> Proper cube action: Redouble, pass
OMG, I'm a cheater...
Raccoon
Micke Nilsson wrote:
> Snowie Rollout:
> >
>
> Rollout Money equity: 0,725
> 9,5% 42,1% 60,5% 39,5% 0,0% 0,0%
> 95% confidence interval:
> - live cube no redouble: 0,858 ±0,044,
> - live cube redouble take: 1,089 ±0,091.
> Rollout settings:
> Full rollout,
> 324 games (equiv. 4741 games),
> played 2-ply (fast), cube 2-ply,
> settlement 0,550 at 16 pts,
> seed 1, without race database.
> Live cube
> 1. Redouble, pass 1,000
> 2. No redouble 0,858 (-0,142)
> 3. Redouble, take 1,089 (+0,089)
> Proper cube action: Redouble, pass
Well ... I'm still taking. I think 3-ply Snowie must be used for this
containment problem, and that it will get the right answer.
A GNUBG 1-ply (Snowie 2-ply) rollout also has this position as a pretty
big pass, but GNUBG 2-ply (Snowie 3-ply) rollout plays the take side
well enough to make it a close double and very easy take.
(I didn't let my gnubg rollout run long, so I'll look forward to seeing
Jim's gnubg rollout results, also.
Micke Nilsson
were right on any of these problems. What is very suspective however is if
you would have gotten all these 18 positions correct at the table. If you
even further did not make one single cube error (not blunder) during 32
consecutive matches at least some would hold that as proof.
/Micke
"Grunty" <grunti...@yahoo.com> skrev i meddelandet
news:1126562535.1...@g49g2000cwa.googlegroups.com...
Micke Nilsson
evaluation", how Snowie would play at 3-ply if you played against "it" in a
match. This is the most you could expect someone sitting with Snowie during
a match could have time to get. If you more or less have the position in
advance and have a really fast computer, you might have time to squeeze in a
Mini-rollout, if you have long enought time for your decision.
It is not a 3-ply rollout. In my rollout I took the default settings but
added live cube and removed the truncated settings. Maybe I can tweak the
settings more, but I'm not all that aware of the best settings.
/Micke
"Raccoon" <racg...@yahoo.com> skrev i meddelandet
news:1126576160....@g49g2000cwa.googlegroups.com...
Micke Nilsson wrote:
> Snowie Rollout:
> >
>
> Rollout Money equity: 0,725
> 9,5% 42,1% 60,5% 39,5% 0,0% 0,0%
> 95% confidence interval:
> - money cubeless eq.: 0,725 ą0,036,
> - live cube no redouble: 0,858 ą0,044,
> - live cube redouble take: 1,089 ą0,091.
Grunty
Just joking, since I seem to at least have come close in this one (not
like my #3 blunder).
Grunty
> Maybe I can tweak the settings more,
Being these position so tricky to evaluate, I think you should set the
rollouts at its maximum strenght (3-ply), and some 300 games would be
fine (someone correct this number?). Depending on your machine
capabilities, it will take some time to perform (half or 1 or 2 hours?).
Micke Nilsson
it took somewhere 3½-4 hours to make the below rollout:
Money session. Score X-O: 0-0
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O X X X | | X |
| X O O | | |
| O O | | | S
| O O | | | n
| O | | | o
| |BAR| | w
| | | | i
| | | | e
| | | |
| X X X X | | |
| X X X X O | | X |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 163 O: 41 X-O: 0-0/Money (1)
Men Off X: 0 O: 5
CubeValue: 2, O owns Cube
Rollout Money equity: 0,684
9,2% 40,0% 59,6% 40,4% 0,0% 0,0%
95% confidence interval:
- money cubeless eq.: 0,684 ±0,031,
- live cube no redouble: 0,810 ±0,042,
- live cube redouble take: 0,979 ±0,077.
Rollout settings:
Full rollout,
324 games (equiv. 6315 games),
played 3-ply (standard), cube 3-ply,
settlement 0,550 at 4 pts,
seed 1, without race database.
Evaluations
1. Redouble, pass 1,000
2. No redouble 0,839 (-0,161)
3. Redouble, take 1,073 (+0,073)
Proper cube action: Redouble, pass
Live cube
1. Redouble, take 0,979
2. No redouble 0,810 (-0,168)
3. Redouble, pass 1,000 (+0,021)
Proper cube action: Redouble, take
------------------------------ End ----------------------------------
"Grunty" <grunti...@yahoo.com> skrev i meddelandet
news:1126612026.8...@o13g2000cwo.googlegroups.com...
Grunty
> Evaluations
> 1. Redouble, pass 1,000
> 2. No redouble 0,839 (-0,161)
> 3. Redouble, take 1,073 (+0,073)
> Proper cube action: Redouble, pass
> Live cube
> 1. Redouble, take 0,979
> 2. No redouble 0,810 (-0,168)
> 3. Redouble, pass 1,000 (+0,021)
> Proper cube action: Redouble, take
If I understand correctly, the most realistic measure is the "Live
cube" one? : Redouble, take ?
Micke Nilsson
the Snowie interface on screen. There the only answer is the first one,
Redouble, pass as proper action. The other one only appears when you export
it to a file. So if you didn't make this file export and only did the live
cube rollout in snowie you would only see Redouble, pass. Confusing enough?
Perhaps someone with more understanding of this than me can come with a
better answer?
/Micke
"Grunty" <grunti...@yahoo.com> skrev i meddelandet
news:1126651923.1...@z14g2000cwz.googlegroups.com...
Komodo
position. Gnu 2-ply 2-ply rollout has it as a double and a trivial take
and snowie live cube (Live cube is, I believe a rollout with the cube
in play, which is a must here) as a borderline take.
I trust gnu in this case.
Cat_in_awe
> Actually I'm not sure. I just saw the difference too. It looks
> different in the Snowie interface on screen. There the only answer is
> the first one, Redouble, pass as proper action. The other one only
> appears when you export it to a file. So if you didn't make this
> file export and only did the live cube rollout in snowie you would
> only see Redouble, pass. Confusing enough?
>
> Perhaps someone with more understanding of this than me can come with
> a better answer?
>
In the Snowie screen there are two panels with two different tabs when you
do a live cube rollout. You have to click the 2nd tab to see the live cube
result.
--
GPC
Jim Segrave
ANd here's a gnubg rollout, stopped at 2 J.S.D. It has significantly
differing cubeful and cubeless equities, but the same conclusion re
redouble/take.
Cube analysis
Rollout cubeless equity +0.599
Cubeful equities:
1. Double, take +0.729
2. Double, pass +1.000 ( +0.271)
3. No double +0.694 ( -0.036)
Proper cube action: Redouble, take
Rollout details:
Player jim_segrave owns 2-cube:
0.563 0.393 0.079 - 0.437 0.000 0.000 CL +0.599 CF +0.694
[0.002 0.002 0.002 - 0.002 0.000 0.000 CL 0.005 CF 0.008]
Player gnubg owns 4-cube:
0.557 0.392 0.084 - 0.443 0.000 0.000 CL +1.181 CF +0.729
[0.002 0.002 0.002 - 0.002 0.000 0.000 CL 0.012 CF 0.016]
Full cubeful rollout with var.redn.
1318 games, Mersenne Twister dice gen. with seed 145821309 and
quasi-random dice
Stop when best play is enough JSDs ahead: limit 2 (min. 648 games)
Play: supremo 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 16 more moves within equity
0.32
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]
--
Jim Segrave (j...@jes-2.demon.nl)
Zorba
> settlement 0,550 at 4 pts,
I think this may cause problems. The cube is already on 2 or even 4
after double/take. A true cubeful ("live cube") rollout should at least
allow for 8-cubes here IMO and only settle if the cube would reach 16.
With a skyhigh gammon rate, all one-sided, settling a game at the VALUE
of 0.550 also seems wrong in itself. A value of 0.65 or even higher is
probably closer to the truth, because there's relatively a lot of
recube vigorish in this position.
Example:
A gammonless race with 78.5% winning chances is usually a borderline
drop/take. Cubeless equity: 0.785-0.215= 0.57. Pretty close to the
settlement value of 0.55.
Now look at this position and let's use these figures: 60% wins, 40%
gammon wins, 10% backgammons, 40% losses. Cubeless equity:
0.6-0.4+0.4+0.1=0.70. A lot above the settlement value of 0.55, yet it
seems clear that this is not a big drop, and probably a rather clear
take.
In general, you should raise the settlement value for gammonish
positions. But I think it's best just not to use it at all. It causes
more problems than it solves IMO.
--
_
/
_ orba
Micke Nilsson
only taken the Snowie defaults.
"Zorba" <zo...@chello.nl> skrev i meddelandet
news:1126731761.3...@o13g2000cwo.googlegroups.com...
Micke Nilsson
"Cat_in_awe" <rl31...@excite.com> skrev i meddelandet
news:7mVVe.483$nF3...@newssvr24.news.prodigy.net...
Micke Nilsson
and top in my first "answer" post) is not a rollout but only a fast 3-ply
instant evaluation, like Snowie had played it if you played a match against
"it".
3-Ply Money equity: 0,645
9,0% 39,0% 58,3% 41,7% 0,0% 0,0%
1. Redouble, take 0,986
2. No redouble 0,801 (-0,185)
3. Redouble, pass 1,000 (+0,014)
Proper cube action: Redouble, take
My rollout with live cube showed a pretty big pass, se below. At least this
is how I read it.
Rollout Money equity: 0,725
9,5% 42,1% 60,5% 39,5% 0,0% 0,0%
95% confidence interval:
- money cubeless eq.: 0,725 ą0,036,
- live cube no redouble: 0,858 ą0,044,
- live cube redouble take: 1,089 ą0,091.
Rollout settings:
Full rollout,
324 games (equiv. 4741 games),
played 2-ply (fast), cube 2-ply,
settlement 0,550 at 16 pts,
seed 1, without race database.
Evaluations
1. Redouble, pass 1,000
2. No redouble 0,879 (-0,121)
3. Redouble, take 1,161 (+0,161)
Proper cube action: Redouble, pass
Live cube
1. Redouble, pass 1,000
2. No redouble 0,858 (-0,142)
3. Redouble, take 1,089 (+0,089)
Proper cube action: Redouble, pass
"Komodo" <kom...@swipnet.se> skrev i meddelandet
news:1126691477....@g44g2000cwa.googlegroups.com...
Micke Nilsson
borderline take here with my 3-ply rollout. But I also saw, as Zorba noticed
that the settlement settings are different in the two rollouts also. I don't
know how much that affects the result, but I guess they must be the same for
comparison.
"Micke Nilsson" <na...@home.se> skrev i meddelandet
news:dgb8co$n0s$1...@green.tninet.se...
Micke Nilsson
O on roll, cube action
+-1--2--3--4--5--6--------7--8--9-10-11-12-+
| X O O X X X | | X |
| X O O | | |
| O O | | | S
| O O | | | n
| O | | | o
| |BAR| | w
| | | | i
| | | | e
| | | |
| X X X X | | |
| X X X X O | | X |
+24-23-22-21-20-19-------18-17-16-15-14-13-+
Pipcount X: 163 O: 41 X-O: 0-0/Money (1)
Men Off X: 0 O: 5
CubeValue: 2, O owns Cube
Rollout Money equity: 0,682
9,1% 39,9% 59,6% 40,4% 0,0% 0,0%
95% confidence interval:
- money cubeless eq.: 0,682 ą0,031,
- live cube no redouble: 0,804 ą0,043,
- live cube redouble take: 0,974 ą0,077.
Rollout settings:
Full rollout,
324 games (equiv. 6192 games),
played 3-ply (standard), cube 3-ply,
settlement 0,550 at 16 pts,
seed 1, without race database.
Evaluations
1. Redouble, pass 1,000
2. No redouble 0,837 (-0,163)
3. Redouble, take 1,068 (+0,068)
Proper cube action: Redouble, pass
Live cube
1. Redouble, take 0,974
2. No redouble 0,804 (-0,171)
3. Redouble, pass 1,000 (+0,026)
Proper cube action: Redouble, take
"Zorba" <zo...@chello.nl> skrev i meddelandet
news:1126731761.3...@o13g2000cwo.googlegroups.com...
Komodo
56.11% 39.25% 8.26% - 43.89% 0.00% 0.00% +1.1948 +0.7618
The main difference seems to be that O wins 59,6% of the games in the
snowie rollout and only 56,38% in the gnu rollout, so it is probably
safe to say that the bots play the containment game differently.