precarious position
badgolferman
often. Here is a link to the image of the board.
http://img600.imageshack.us/img600/6344/clipboard02oo.jpg
What do you play?
julia.m...@googlemail.com
keeps alive gammon chances which are valuable at the score; also if
red rolls a 5 he's forced past your blot.
tc...@lsa.umich.edu
This is a tough choice. The first thing to notice is that gammon wins
are worth very little at this score. If you win a gammon then you take
away your opponent's post-Crawford free drop, which isn't worth much.
The second thing to note is that you have a substantial lead in the race
and so you would like to play safely and not get hit. Unfortunately you
can't avoid leaving a direct shot, but 11/6 7/5 leaves two fewer shots
than 11/5*, so that's a strong argument in favor of 11/6 7/5.
What can be said in favor of 11/5*? Well, although you have a race lead,
it's not gin, so when 11/5* works it improves your racing chances. If
Jellyfish enters with a 5 then that helps you bear in safely. However,
on average I think you incur more shot jeopardy later on by hitting, since
Jellyfish could stay out or enter with a 3, causing you future problems.
I'd probably play 11/6 7/5 but I'm not confident of my choice. If you
make the race closer then that will make hitting more attractive, and it's
quite tricky to judge the crossover point.
--
Tim Chow tchow-at-alum-dot-mit-dot-edu
The range of our projectiles---even ... the artillery---however great, will
never exceed four of those miles of which as many thousand separate us from
the center of the earth. ---Galileo, Dialogues Concerning Two New Sciences
Stick
There are many scores that should be looked at as 'dmp checker play
scores'. These include -1 -X Crawford where X is an odd number. So
-1 -3 C, -1 -5 C, etc... Here we have one such position and should
make the same play we would make at dmp. The reasoning as Tim states
is that by winning a gammon all we do is deprive our opponent of a
free drop and that isn't worth hardly anything at all.
As for keeping gammon chances alive in the position even if they were
worth something it should be clear that it would take a small miracle
for you to win a gammon here.
What makes this position tricky is that your racing lead is large but
not gin but it's not gin mainly because of your horrid distribution.
You have a decent amount of wastage already and unless you roll small
numbers (which are bad for the race) you are likely to have more since
all your men are hanging out on the 7pt. I still believe you have
enough of a racing lead to counterbalance the two extra shots you'd
have to leave by hitting. Also of note is your distribution is better
for the race after the non hitting play. Lastly, you leave less
recurring shots after the non hit (4 immediate numbers, 66 55 65) than
after the hit (9 immediate numbers, 66 55 44 65 64 54).
Stick
http://www.bgonline.org/forums/
On Jan 25, 9:50 am, "Julia.M.Hayw...@googlemail.com"
Turz
> This is a tough choice. The first thing to notice is that gammon wins
> are worth very little at this score. If you win a gammon then you take
> away your opponent's post-Crawford free drop, which isn't worth much.
I agree. I think in this game Player2 must only concentrate on
avoiding losing this match point and move on to the next game.
> The second thing to note is that you have a substantial lead in the race
> and so you would like to play safely and not get hit. Unfortunately you
> can't avoid leaving a direct shot, but 11/6 7/5 leaves two fewer shots
> than 11/5*, so that's a strong argument in favor of 11/6 7/5.
I agree that it would be better to play safely, but as you said you
can't avoid leaving a direct shot.
You didn't address one point: if you hit, the opponent has 25%
probability of dancing. That would enhance Player2's racing advantage.
Whether it is worth those 2 hits (11/36 instead of 13/36) it is hard
to say.
On the other side, 11/6 7/5 makes you bear in 1 checker more, which is
also a small race advantage, especially when the remaining checkers
are on the 7-point.
In addition, a blot on the 5-point is easier to save than a blot on
the 4-point.
So 11/6 7/5.
Turz
badgolferman
So far the consensus seems to be to play 11/6, 7/5. My head told me to
play the same thing, mostly due to the fact that there is only one
number that can get me -- 1. However I ended up playing 11/4* instead
just to see what would happen. Jellyfish failed to enter the board on
the next roll and I then got double twos to essentially close all the
gaps. From that point on it was off to the races. Jellyfish got
gammoned this game and forfeited the next game when I doubled to make
the score 6-5. I doubled the final game and ended up winning the whole
match when I beared off all mine with only one of his remaining!
Despite winning that game and match I still don't think I would play
the same way against a live opponent. I think 11/6, 7/5 is a better
play, but you never know what will happen when the element of luck is
introduced.
Walt
Here, you are at 5-away crawford. Gammons are not particularly
valuable, since the difference between 3- and 4- away post-crawford is
negligible (you need to win the next two games or a gammon either way).
And of course gammon losses are no worse than single losses, so your
checker play would be basically the same as DMP.
At DMP, objective 1 is to minimize shots. So 11/6 7/5. This leaves
your opp only 11/36 to hit, and if not hit you'll safety the blot next
turn 32/36.
//Walt
Walt
> In article <xn0h9j93...@reader.albasani.net>,
> badgolferman <REMOVETHISb...@gmail.com> wrote:
>> Please forgive the non-gnubg board, but I play Jellyfish Light more
>> often. Here is a link to the image of the board.
>>
>> http://img600.imageshack.us/img600/6344/clipboard02oo.jpg
>>
>> What do you play?
>
> than 11/5*, so that's a strong argument in favor of 11/6 7/5.
>
> What can be said in favor of 11/5*? Well, although you have a race lead,
I think you have to factor in your ability to safety the blot when
you're not hit. If you leave the blot on the 5, only 65 55 and 66
prevent you from safetying it. With the blot on the 4, add in 64 54 and
44. So it's more than just the two additional immediate shots, the
extra non-safetying rolls next time contribute too.
//Walt
tc...@lsa.umich.edu
Turz <bit...@gmail.com> wrote:
>You didn't address one point: if you hit, the opponent has 25%
>probability of dancing. That would enhance Player2's racing advantage.
Actually, I did say that "although you have a racing lead, it's not gin,
so when 11/5* works it improves your racing chances."
And yes, of course, 11/5* should have read 11/4*.
tc...@lsa.umich.edu
badgolferman <REMOVETHISb...@gmail.com> wrote:
>Jellyfish got gammoned this game and forfeited the next game when
>I doubled to make the score 6-5.
You must have waited far too late to double if Jellyfish dropped a cube
at 1-away/3-away post-Crawford.
julia.m...@googlemail.com
free drop or not" (or perhaps "don't pontificate on positions in a
spare minute or so while software builds"). I looked at it again this
morning and I could have sworn I saw the cube turned yesterday -
obviously not, though, as it's Crawford. D'oh! :(
Walt
> In article <xn0h9jgv...@reader.albasani.net>,
> badgolferman <REMOVETHISb...@gmail.com> wrote:
>> Jellyfish got gammoned this game and forfeited the next game when
>> I doubled to make the score 6-5.
>
> You must have waited far too late to double if Jellyfish dropped a cube
> at 1-away/3-away post-Crawford.
Yep. Trailing post Crawford you want to double at the first opportunity.
The only reason not to is to attempt "the trick"
see: http://www.bkgm.com/faq/Matches.html#what_is_the_trick_
The trick works against humans, but Jellyfish won't fall for it.
//Walt
tc...@lsa.umich.edu
Walt <walt_...@SHOESyahoo.com> wrote:
>Yep. Trailing post Crawford you want to double at the first opportunity.
>The only reason not to is to attempt "the trick"
Right. Note, though, that what you say here applies to odd-away
post-Crawford scores. At a post-Crawford score where your opponent has a
free drop, there's another reason that you might not double at your first
chance---you might be too good. For example, say you win the opening
roll with a 3-1 and your opponent rolls a 6-2. If you're trailing (say)
2-away/1-away post-Crawford, you're probably now too good to double!
This might seem impossible but the point is that you'll almost certainly
still be able to cash next roll, so you lose very little by playing on
and hoping for a great roll that will put you on the road to Gammon City.
Walt
> In article <9qW%o.646653$iV7.3...@en-nntp-15.dc1.easynews.com>,
> Walt <walt_...@SHOESyahoo.com> wrote:
>> Yep. Trailing post Crawford you want to double at the first opportunity.
>> The only reason not to is to attempt "the trick"
>
> Right. Note, though, that what you say here applies to odd-away
> post-Crawford scores. At a post-Crawford score where your opponent has a
> free drop, there's another reason that you might not double at your first
> chance---you might be too good. For example, say you win the opening
> roll with a 3-1 and your opponent rolls a 6-2. If you're trailing (say)
> 2-away/1-away post-Crawford, you're probably now too good to double!
> This might seem impossible but the point is that you'll almost certainly
> still be able to cash next roll, so you lose very little by playing on
> and hoping for a great roll that will put you on the road to Gammon City.
Hmmmm.... trying to get my head around this...
In your example (trailing 2-away/1-away post-Crawford) if I cash my MWC
is 50%. Surely if my gammon chances are greater than 50% it would be
correct to play on, but if my gammon chances are less than 50% it would
seem correct to cash. I suppose if the cash is gin this changes things,
but I'm not sure that it is.
//Walt
tc...@lsa.umich.edu
Walt <walt_...@SHOESyahoo.com> wrote:
>In your example (trailing 2-away/1-away post-Crawford) if I cash my MWC
>is 50%. Surely if my gammon chances are greater than 50% it would be
>correct to play on, but if my gammon chances are less than 50% it would
>seem correct to cash. I suppose if the cash is gin this changes things,
>but I'm not sure that it is.
The cash is not gin, but your comment about gammon chances being greater
than 50% is not correct.
I think Trice is the one who put things this way: When you decide to play
on, you're not making an irrevocable vow to play the game out to completion
without ever touching the cube. You're just deciding to play on for *one
more roll*. Thus what matters is not your chance of ultimately winning a
gammon. What matters is how much you gain between this roll and the next
by holding on to the cube.
You're already familiar with the market-loser concept. In a "normal"
position, when deciding whether to double, you assess your market losers
and what I call your market freezers, by which I mean the sequences that
lead you to wish that you *hadn't* doubled. You double if your market
losers outweigh your market freezers.
Similarly when your opponent has a drop and you're deciding whether you're
too good, you should assess your market makers (those sequences that lead
you to wish you had cashed) and your market freezers (those sequences that
make you glad you played on). If your market freezers outweigh your market
makers then you play on.
In the example at hand, there are very few market makers and some market
freezers so it's a play-on.
Richard Henry
> On 1/25/2011 4:05 PM, tc...@lsa.umich.edu wrote:
>
> > badgolferman <REMOVETHISbadgolfer...@gmail.com> wrote:
> >> Jellyfish got gammoned this game and forfeited the next game when
> >> I doubled to make the score 6-5.
>
> > You must have waited far too late to double if Jellyfish dropped a cube
> > at 1-away/3-away post-Crawford.
>
> Yep. Trailing post Crawford you want to double at the first opportunity.
> The only reason not to is to attempt "the trick"
>
> see:http://www.bkgm.com/faq/Matches.html#what_is_the_trick_
>
> The trick works against humans, but Jellyfish won't fall for it.
>
> //Walt
??? What would Jellyfish do, then?
Richard Henry
wrote:
I have read the analysis and I understand it is wrong, but I would
still would hit with 11/4.
tc...@lsa.umich.edu
Do you understand what "the trick" is? If you do then you should be able
to answer your own question.
Jellyfish might fall for the trick but it's not too likely, and you risk
making a mistake yourself by waiting too long and losing your market.
Richard Henry
> In article <e9252fd7-4e48-4dca-8e1a-bbbb10711...@j32g2000prh.googlegroups.com>,
> Richard Henry <pomer...@hotmail.com> wrote:
>
> >On Jan 26, 6:43 am, Walt <walt_ask...@SHOESyahoo.com> wrote:
> >> Yep. Trailing post Crawford you want to double at the first opportunity.
> >> The only reason not to is to attempt "the trick"
>
> >> see:http://www.bkgm.com/faq/Matches.html#what_is_the_trick_
>
> >> The trick works against humans, but Jellyfish won't fall for it.
>
> >??? What would Jellyfish do, then?
>
> Do you understand what "the trick" is? If you do then you should be able
> to answer your own question.
>
tc...@lsa.umich.edu
The straight answer is, Jellyfish would correctly take or drop.
I didn't say this before because it probably doesn't make sense to you,
because you probably haven't taken the time to understand what "the trick" is.
Did you even bother clicking on the link and trying to understand what it
said?
Andrew B.
> In article <XnX%o.546290$Ua4.448...@en-nntp-09.dc1.easynews.com>,
>
> Walt <walt_ask...@SHOESyahoo.com> wrote:
> >In your example (trailing 2-away/1-away post-Crawford) if I cash my MWC
> >is 50%. Surely if my gammon chances are greater than 50% it would be
> >correct to play on, but if my gammon chances are less than 50% it would
> >seem correct to cash. I suppose if the cash is gin this changes things,
> >but I'm not sure that it is.
>
> The cash is not gin, but your comment about gammon chances being greater
> than 50% is not correct.
>
> I think Trice is the one who put things this way: When you decide to play
> on, you're not making an irrevocable vow to play the game out to completion
> without ever touching the cube. You're just deciding to play on for *one
> more roll*. Thus what matters is not your chance of ultimately winning a
> gammon. What matters is how much you gain between this roll and the next
> by holding on to the cube.
Even if it were "double now, or vow never to touch the cube", wouldn't
it be correct to play on whenever the chance of a gammon is greater
than the chance of losing?
Walt
> In article <XnX%o.546290$Ua4.4...@en-nntp-09.dc1.easynews.com>,
> Walt <walt_...@SHOESyahoo.com> wrote:
>> In your example (trailing 2-away/1-away post-Crawford) if I cash my MWC
>> is 50%. Surely if my gammon chances are greater than 50% it would be
>> correct to play on, but if my gammon chances are less than 50% it would
>> seem correct to cash. I suppose if the cash is gin this changes things,
>> but I'm not sure that it is.
>
> In the example at hand, there are very few market makers and some market
> freezers so it's a play-on.
Hmmmm... gnu 0-ply begs to differ. Playing on is not a big error, but
turning the cube is best. And the cash is far from gin, in fact it's
not even a cash yet.
GNU Backgammon Position ID: 4HPkQSCwZ/ABMA
Match ID : cAmgAEAAEAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: O
| X O O | | O X | 4 points
| X O | | O X |
| X O | | O |
| X | | O |
| X | | O |
v| |BAR| | 5 point match (Cube: 1)
| | | |
| O | | X |
| O | | X |
| O X | | X X | On roll
| O X O | | X X O | 2 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: You
Rollout details:
Centered 1-cube:
0.572 0.195 0.020 - 0.428 0.162 0.022 CL +0.490 CF +0.566
[0.002 0.002 0.002 - 0.002 0.005 0.003 CL 0.008 CF 0.007]
Player docgege owns 2-cube:
0.570 0.201 0.024 - 0.430 0.166 0.027 CL +0.590 CF +0.590
[0.002 0.002 0.003 - 0.002 0.005 0.003 CL 0.008 CF 0.008]
Full cubeful rollout with var.redn.
1296 games, Mersenne Twister dice gen. with seed 873625579 and
quasi-random dice
Play: 0-ply cubeful [expert]
Cube: 0-ply cubeful [expert]
Cubeful equities:
1. Double, take +0.590
2. Double, pass +1.000 ( +0.410)
3. No double +0.566 ( -0.024)
Proper cube action: Double, take
Walt
> On 1/26/2011 11:20 AM, tc...@lsa.umich.edu wrote:
>> In article <XnX%o.546290$Ua4.4...@en-nntp-09.dc1.easynews.com>,
>> Walt <walt_...@SHOESyahoo.com> wrote:
>>> In your example (trailing 2-away/1-away post-Crawford) if I cash my MWC
>>> is 50%. Surely if my gammon chances are greater than 50% it would be
>>> correct to play on, but if my gammon chances are less than 50% it would
>>> seem correct to cash. I suppose if the cash is gin this changes things,
>>> but I'm not sure that it is.
>
>>
>> In the example at hand, there are very few market makers and some market
>> freezers so it's a play-on.
>
>
> Hmmmm... gnu 0-ply begs to differ.
Oops. I rolled this out at the wrong match score. Nevermind
Walt
Rollout (this time at the right match score)
This confirms what you were saying - I'll bet you're not surprised.
Even with a modest 20% gammon chances it's correct to play on.
GNU Backgammon Position ID: 4HPkQSCwZ/ABMA
Match ID : cAmgAEAAGAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: O
| X O O | | O X | 4 points
| X O | | O X |
| X O | | O |
| X | | O |
| X | | O |
v| |BAR| | 5 point match (Cube: 1)
| | | |
| O | | X |
| O | | X |
| O X | | X X | On roll
| O X O | | X X O | 3 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: You
Rollout details:
Centered 1-cube:
0.565 0.202 0.013 - 0.435 0.123 0.012 CL +0.461 CF +1.056
[0.001 0.002 0.002 - 0.001 0.003 0.002 CL 0.004 CF 0.005]
Player O owns 2-cube:
0.569 0.248 0.040 - 0.431 0.137 0.022 CL +1.274 CF +1.274
[0.003 0.006 0.004 - 0.003 0.005 0.003 CL 0.010 CF 0.010]
Full cubeful rollout with var.redn.
1296 games, Mersenne Twister dice gen. with seed 873625579 and
quasi-random dice
Play: 0-ply cubeful [expert]
Cube: 0-ply cubeful [expert]
Cubeful equities:
1. No double +1.056
2. Double, take +1.274 ( +0.219)
3. Double, pass +1.000 ( -0.056)
Cubeful equities:
1. No double 51.39%
2. Double, take 56.86% ( 5.47%)
3. Double, pass 50% ( -1.39%)
tc...@lsa.umich.edu
Andrew B. <bul...@gmail.com> wrote:
>Even if it were "double now, or vow never to touch the cube", wouldn't
>it be correct to play on whenever the chance of a gammon is greater
>than the chance of losing?
That's an interesting question and I'd have to think about it. We'd also
have to decide on what other rules are governing the cube; for example,
does the opponent get to recube later? Do we get to re-recube?
In any case, in the situation under discussion, the chance of a gammon is
*far less* than the chance of losing, so you'd certainly not play on if
that were your guiding principle.
tc...@lsa.umich.edu
<> In article <XnX%o.546290$Ua4.448...@en-nntp-09.dc1.easynews.com>,
<>
<> Walt <walt_ask...@SHOESyahoo.com> wrote:
<> >In your example (trailing 2-away/1-away post-Crawford) if I cash my MWC
<> >is 50%. Surely if my gammon chances are greater than 50% it would be
<> >correct to play on, but if my gammon chances are less than 50% it would
<> >seem correct to cash. I suppose if the cash is gin this changes things,
<> >but I'm not sure that it is.
<>
<> The cash is not gin, but your comment about gammon chances being greater
<> than 50% is not correct.
<Even if it were "double now, or vow never to touch the cube", wouldn't
<it be correct to play on whenever the chance of a gammon is greater
<than the chance of losing?
Re-reading this, I see that maybe your point is that I should have said,
"your comment about gammon chances being *less* than 50% is not correct"?
You're right that that would have been a more accurate comment.
Richard Henry
> In article <ae344a31-68c4-44f3-ae13-e55dcd504...@c13g2000prc.googlegroups.com>,
> Richard Henry <pomer...@hotmail.com> wrote:
>
> >On Jan 26, 10:03 am, tc...@lsa.umich.edu wrote:
> >> Do you understand what "the trick" is? If you do then you should be able
> >> to answer your own question.
>
> >Why is it so difficult to get a straight answer?
>
> The straight answer is, Jellyfish would correctly take or drop.
>
Well, that's illuminating.
tc...@lsa.umich.edu
Glad to be of help. I didn't think the straight answer would make sense
to you, but evidently I underestimated you.
Richard Henry
> In article <398161b7-f14c-4f4f-aa59-5b44bfcfd...@r4g2000prm.googlegroups.com>,
> Richard Henry <pomer...@hotmail.com> wrote:
> <On Jan 26, 1:50 pm, tc...@lsa.umich.edu wrote:
> <> In article <ae344a31-68c4-44f3-ae13-e55dcd504...@c13g2000prc.googlegroups.com>,
> <> Richard Henry <pomer...@hotmail.com> wrote:
> <>
> <> >On Jan 26, 10:03 am, tc...@lsa.umich.edu wrote:
> <> >> Do you understand what "the trick" is? If you do then you should be able
> <> >> to answer your own question.
> <>
> <> >Why is it so difficult to get a straight answer?
> <>
> <> The straight answer is, Jellyfish would correctly take or drop.
> <
> <Well, that's illuminating.
>
> Glad to be of help. I didn't think the straight answer would make sense
> to you, but evidently I underestimated you.
The so-called "trick" is rather obvious once the math is laid out, but
my question bears to your opinion on jellyfish. I can see that there
are at least three ourfor you to have formed opinion of its response
to the situation.
1. You are familiar with its programming.
2. You are experienced with its response from having played it (or
observed it playing) in a similar situation.
3. Blind faith.
Is it one of those? Is it something else?
tc...@lsa.umich.edu
Richard Henry <pome...@hotmail.com> wrote:
>1. You are familiar with its programming.
>2. You are experienced with its response from having played it (or
>observed it playing) in a similar situation.
>3. Blind faith.
>
>Is it one of those? Is it something else?
#1 and #2 are pretty close. I don't know all the details of Jellyfish's
code because it isn't open source, but I'm familiar with the basic
concepts of neural-network programming that underlie all the strong
bots. I've studied many examples of Jellyfish rollouts and moreover,
its performance has been thoroughly documented by people running Dueller
and/or bot-comparison projects.
And as I said before, Jellyfish will sometimes fall for the trick, but
in practice, most humans are more likely to screw up trying to trick
Jellyfish than Jellyfish is to fall for the trick. This is even more
true for bots such as GNU and XG whose match-play algorithms are better
than Jellyfish's.