News:

Subscribe to Fibsboard for just £5 a year - get a special badge! VIP Articles and Links and much more..ensure our future and earn real kudos with your friends and enemies alike! http://www.fibsboard.com/donate.php

Main Menu

Match 1, Game 3, Move 3, Forum double

Started by blitzxz, May 08, 2009, 01:24:41 PM

Previous topic - Next topic

blitzxz

Cube?

Score:
Green(forum) 2, White(factotum) 0, match to 5

Pips:
Green 159 White 182

GnuID:
4HnwAGjEc+QBIg:cAmgAAAAEAAA

diane

Spoiler
I voted no again, for similar reasons to the last time.  Either it means I am right this time, or I learned nothing  ;) I just cannot see him taking that cube, and so I really would rather see if my gammonish play last time takes me there, rather than settle for one point out of what was just a real stroke of luck for us..
[close]
Never give up on the things that make you smile

stog

Spoiler
double! or is already too good?
[close]

nice colour stones blitzxz!

ah_clem

Spoiler


Let's see: He's got two checkers on the roof.  We've got four blots hanging around. Neither of us have a board to speak of, but we're threatening a blitz.  We're ahead in the race.  Applying PRAT:

Position: more or less even - a bunch of candlesticks for both
Race: we're ahead by 23 pips
Threats: big blitz threat against factorum, a few scattered blots for us, so we're ahead in that column.

That's two out of three, so a double would be indicated.

How about market losers?  If we make a second point, white is a 55% chance to fail to enter both checkers. That is, he sits on the bar for another turn most of the time, giving us a chance to consolidate or apply more pressure.  So, many market losing sequences arguing for a double.

What about the match score?  We're ahead in the match, so we should be somewhat cautious about doubling.  If we were 2-away instead of 3-away we'd need to be even more cautious about doubling because a take would provoke an automatic redoubling.  But we're not at 2-away.  Eating a  recube would be painful - if it's a clear pass we're back to even in the match, if we take and lose it would still leave us only needing two wins in a row to win the match which is not exactly horrible.

So I say: double

Now I'm going to read other responses and see if I change my mind before voting.
[close]

socksey

Spoiler
I say cube.  We are ahead in the pip count.  We can only be hit once in the worst case scenerio if he rolls dbl 3.  Naaaaaaaaaa......he won't do that!   :laugh:  If he doesn't roll a 3, we're in great shape.  I think it's worth the risk.  If he doesn't take, we've won another sure thing.  If he does take, it could still be a gammon and game!
[close]

Oh, and thanks for the nice dice color change!  No not seeing green! ;)

socksey



"We call that person who has lost his father, an orphan; and a widower that man who has lost his wife. But that man who has known the immense unhappiness of losing a friend, by what name do we call him? Here every language is silent and holds its peace in impotence. - Muhammad Ali

ah_clem

Spoiler
Quote from: stog on May 08, 2009, 01:37:23 PM
double! or is already too good?
Too good?  Nah.  I'd say we win something like 70% here, meaning about 30% losses.  To play on for a gammon usually requires twice as many gammons as losses, and there is nothing approaching 60% gammons here.  This rule of thumb is distorted somewhat by the match score, but it's not too far off.

I'm guessing he'll pass, because of the high percentage of gammons (while it's not 60% it's probably 30 to 40%) - if he takes and we gammon the match is over.  Let's take the point.                             

[close]
Quote from: stog on May 08, 2009, 01:37:23 PM
nice colour stones blitzxz!
Yeah. Nice. Thanks.

NIHILIST

#6
Spoiler
Ship the cube.
[close]
Spoiler
[close]
stog spoiler edit! - select text to be spoiled then click radiation button, stand clear, then check if all is ok regards :)


Bob
Robert J Ebbeler

Zorba

Spoiler
I think it's a clear take for factotum, two men on the bar might be scary, but you have to look at what his chances are to enter. We haven't made any additional homeboard point yet. Most likely we will make the threepoint, but even then it's just a two point board. factotum figures to be able to enter his checkers in a couple of rolls, and it's not unlikely he will get to anchor at our 4 or 5pt. He would have a very playable game after that, since we don't have much else going for us other than the strong initiative here.

The matchscore also matters a lot here, after doubling, we can only use 3 out of 4 points from a gammon, so they are not fully efficient. Most of all though, factotum will have a very strong recube to 4 here if he manages to turn it around, since we can only use 3 out of 4 points, and any residual gammons on a 4-cube would be worthless to us, whereas factotum would get to Crawford or even win the match with a gammon.

So, big take I think, but probably enough to double because of the strong initiative we have here and a nice lead in the race.
[close]
The fascist's feelings of insecurity run so deep that he desperately needs a classification of some things as successful or superior and other things as failed or inferior. This also underlies the fascist's embracement of concepts like mental illness and IQ tests.  - R.J.V.

Luck is my main skill

blitzxz

Spoiler
No double, in this score. Only one homepoint made as Zorba said with big chances to enter and monster recube opportunities for factotum. If I would be factotum I would be cheering after the misfortune to get this big cube ownership. Huge take, so Woolseys's law would say no double but I personally hate that law. More important question is that are we about to lose markets here? What will most likely happens with next exchange of rolls? Propably we will make only one point and factotum will enter with both checkers and it's still take. Even if factotum enters only with one it's very close and I'm not sure, but not much is lost. So clear no double to me. In even score this would be complitely different of course...
[close]

ah_clem

Spoiler
Quote from: blitzxz on May 09, 2009, 07:25:25 PM
Huge take, so Woolseys's law would say no double...

Well, that's not quite what Woolsey's law says.
Woolsey's Law: If you are not 100% sure that the position is a take, then it is always correct to double.

What you've stated is the converse, which does not hold - there are plenty of clear takes that are still a correct double.  IOW, it only works one way, not both.

[close]

A related question, blitzxz: Is it possible to modify the poll so as to allow people to change their votes? That way if someone gives a  particularly trenchant argument we'd have the option to change our mind.  Or are we supposed to vote without reading the other analysis first?


blitzxz

#10
Quote from: ah_clem on May 09, 2009, 07:45:20 PM
Well, that's not quite what Woolsey's law says.
Woolsey's Law: If you are not 100% sure that the position is a take, then it is always correct to double.

What you've stated is the converse, which does not hold - there are plenty of clear takes that are still a correct double.  IOW, it only works one way, not both.

A related question, blitzxz: Is it possible to modify the poll so as to allow people to change their votes? That way if someone gives a  particularly trenchant argument we'd have the option to change our mind.  Or are we supposed to vote without reading the other analysis first?

Yeah, I propably have some misunderstanding about the law because I just don't get it. Now it sound like it's even more useless then I thought...  :laugh: So when you're sure it's a take, you don't know should you double or not and same when you're sure it's pass but when you're unsure then you should double. However this weird sounding logic only works if your opponent is unsure also. It might as well be that he's sure when you're unsure so there no chance for error and you end up doubling complitely wrong time because you're unsure... so all in all you're back to square one and this rule doesn't seem to help me at all. But everybody uses the rules of thumb which they prefer.  :)

I can put the chance to change vote to next poll. I didn't notice that option before.






ah_clem

Quote from: blitzxz on May 10, 2009, 02:06:06 AM
Yeah, I propably have some misunderstanding about the law because I just don't get it. Now it sound like it's even more useless then I thought...  :laugh: So when you're sure it's a take, you don't know should you double or not and same when you're sure it's pass but when you're unsure then you should double. However this weird sounding logic only works if your opponent is unsure also. It might as well be that he's sure when you're unsure so there no chance for error and you end up doubling complitely wrong time because you're unsure... so all in all you're back to square one and this rule doesn't seem to help me at all. But everybody uses the rules of thumb which they prefer.  :)

I can put the chance to change vote to next poll. I didn't notice that option before.

Here's the way I like to think about it:  The best time to double is at the top of your market - that's when your opponent is right on the take point, when your equity is unchanged whether he takes or not.  At that point, nobody is sure whether it's a double or not - even the bots disagree. So Woolsey's law will tell you that yes this is a double.  IOW, if you can make your opponent squirm trying to determine whether it's a take or not, it is the perfect time to double.

Whenever you are in that window where it's a difficult take decision for your opponent, you must be near the top of the market, so it must be a good time to double.

The thing is, it's only a sufficient condition, not a necessary one. There are plenty of examples where it's a clear take or a clear pass when it's still a correct double.  Woolsey's law won't help you there - you've got to find another way to make the decision.  But when it does apply, it's a slam dunk.  Tough take decision?  Ship the cube.


Thanks for allowing vote changes.  I realize I'm kind of a PITA for the mods on this board, asking for color changes and functionality changes and the like.  I do appreciate the swift accommodation to my requests.




dickbalaska

My favorite time to double is when i'm behind in pip count, but i feel my board is shaping up better.  One or two turns later, and opp is cursing the take.  (Or he's pounded me and doubled back).

Oh, and this being fibs, i have to think "If i double, will he drop me?"   :laugh:

dik

dorbel

Woolsey's law. "If you are not 100% sure that it is a take, it must be a double". The reasoning is that if you have any doubt about whether you would take if you were in your opponent's shoes, it's got to be a pretty strong double. This is the most helpful guide to good cube action that I know.
QuoteHowever this weird sounding logic only works if your opponent is unsure also.
This just isn't true. What your opponent thinks is immaterial. Your cube action will still be correct whatever he thinks or does.
QuoteThe best time to double is at the top of your market - that's when your opponent is right on the take point, when your equity is unchanged whether he takes or not.  At that point, nobody is sure whether it's a double or not - even the bots disagree.
Can this be right? I would be very interested to see a position where one player has a marginal take/pass but the bots disagree on whether it is correct to double. There are certain rare positions at lop-sided match scores where this might apply but generally if your equity is pretty much the same after a take or a pass, why would you not double?

ah_clem

#14
Quote from: dorbel on May 10, 2009, 11:30:14 AM
Can this be right?

Feh.  Me and my fumble fingers. I wrote "double" when I meant "take".
Here's what I meant to say:


The best time to double is at the top of your market - that's when your opponent is right on the take point, when your equity is unchanged whether he takes or not.  At that point, nobody is sure whether it's a double  take or not - even the bots disagree. So Woolsey's law will tell you that yes this is a double.  IOW, if you can make your opponent squirm trying to determine whether it's a take or not, it is the perfect time to double.


ah_clem

Spoiler


Quote from: Zorba on May 09, 2009, 02:52:56 PM
The matchscore also matters a lot here,

This is probably the single most cogent in the entire analysis.  Match score is decisive here. A quick evaluation & rollout show that with us leading 2-0 doubling is a mistake. A fairly big one, actually.

However, at 0-0 with the same position, it's a double-take.
And at 1-1, it's a double pass.

Playing with it in gnubg, I see that by changing the match score, doubling can be correct or a huge blunder.    Likewise not doubling  can be correct or a huge blunder depending on the match score. Same for the take decision.  It depends heavily on the score.

Interesting.

My appologies to the gang for arguing for the wrong decision.  Keep that in mind next time you're tempted to listen to me.  (c:
[close]

blitzxz

Quote from: ah_clem on May 10, 2009, 04:06:11 AM
Whenever you are in that window where it's a difficult take decision for your opponent, you must be near the top of the market, so it must be a good time to double.

If this would be true then no one would make big double errors... Opponent might have hard doubling decision when it's not even near the take point or the markets are lost by mile already. And also I might think that the decision is close when it actually is not and then I might do way too early double. And if my evalutions are correct in first place I don't need the law. But like I said everybody uses rules of thumb which they prefer.

Zorba

Quote from: blitzxz on May 11, 2009, 06:15:12 PM
If this would be true then no one would make big double errors... Opponent might have hard doubling decision when it's not even near the take point or the markets are lost by mile already. And also I might think that the decision is close when it actually is not and then I might do way too early double. And if my evalutions are correct in first place I don't need the law. But like I said everybody uses rules of thumb which they prefer.

Obviously, this rule won't help a player who's grossly misevaluating a position much. But that's not the goal of this rule. Also, the examples you give are not in any way contradicted by the Woolsey Rule, they are just situations outside the scope of it.

I think you fail to see the power of Woolsey's doubling rule: When there's any doubt in your mind about whether opponent has a take, you should always double. I consider it the single most best advice on using the cube, and even good players sometimes forget to apply it.

As said, the rule doesn't mean that there aren't other situations with a clear take or clear pass, where you should NOT double. And these are usually harder decisions to make. You will have to use other "tools" for those situations.

The reasons why Kit's rule works so well in practice are partly mathematical, and partly "psychological". Suppose you have indeed some doubt about whether it's a take. What are the scenarios?

If the double/take equity of a position is indeed close to 1, it's nearly always correct to double (certain lopsided matchscores are an exception). Also, it's often a pretty big to very big error NOT to double in these cases. Occasionally, even with the double/take equity close to 1, volatility may be so low that it's not a double yet. However, you will lose very little when you double those. And you may actually gain if opponent errs on the take! So, the rule works well here, preventing you from big errors at an extremely small cost.

If the double/take equity is more like 0.8 and thus a big take, it's often still correct to double these, sometimes by a fair amount, and if it's not, then you still won't lose much usually. BUT!!! Your opponent may well see the position like you did! And as such, have doubts about the take. If he drops, you gain a lot, whether it was a correct double or not. So the Woolsey Rule overall still works very well here.

If the double/take equity is more like 1.2 and thus a big drop, once more it's often still correct to double, and sometimes by a fair amount. It could be too good to double, but usually not by much, so you won't lose much on these. And once again, your opponent may have same problems evaluating as you did, and take your cube. Bingo! Once more, applying the Woolsey Rule stands to gain a lot at very little risk.

Of course, it's possible to misevaluate a position by more than this. Maybe it's a 0.5 take when you think it's close, or a 1.5 drop. Obviously, in those cases, you may lose quite a bit on a misguided double. But once more, it's not totally surprising to see your opponent misevaluate it as well, and the benefits from that only increase! Anyhow, even if you lose on those, you'd probably make errors from gross misevaluations anyway, so the Woolsey Rule is unlikely to make things any worse for you.

So overall, the Woolsey Rule is not a rule to help you evaluate winning chances or equity; it's a general rule on situations where you should double. It reminds you of the small risks in doubling compared to the big gains from either correct cube action by both, or an error from your opponent. It also reminds you of the fact that the question whether to double or not, should start with looking at opponent's take decision.
The fascist's feelings of insecurity run so deep that he desperately needs a classification of some things as successful or superior and other things as failed or inferior. This also underlies the fascist's embracement of concepts like mental illness and IQ tests.  - R.J.V.

Luck is my main skill

socksey

I want to change my vote to no double.  :)  Can that reflect in the poll, because I already voted to double.   :cry:

socksey



"There are two kinds of people, those who do the work and those who take the credit. Try to be in the first group; there is less competition there." - Indira Gandhi





ah_clem

Quote from: socksey on May 13, 2009, 10:26:32 PM
I want to change my vote to no double.  :)  Can that reflect in the poll, because I already voted to double.   


Well, the voting is closed, we've tallied the results, made the decision to pass the cube, had factorum accept the cube and rolled the dice for the next move. I think it's a little late to change your vote. 

If they changed it at this point, then no-double would be the winner and people who look at the thread later will wonder why we doubled when the vote said not to.

socksey

Sorry, was that rather stupid of me?   :laugh:

socksey



"The power of accurate observation is commonly called cynicism by those who have not got it." - George Bernard Shaw

Zorba

GNUBG 2-ply ROLLOUT:

Spoiler
Oh boy! Is it really this bad? I have a little doubt about the size of the error the rollout result gives (evaluations have it much closer), but there is no doubt that it's a big whopper with extras to cube this. The matchscore plays a big role in this. Only needing three points and ahead in the match, we should've been much more careful doubling in a (mutually) gammonish position.

Size of the doubling error in bold and underlined.

Rollout details:
Centered 1-cube:
   63,49  30,25   5,45 -  36,51   8,02   0,74 CL  +0,4638 CF  +0,5537
[  0,25   0,43   0,40 -   0,25   0,15   0,15 CL   0,0080 CF   0,0183]
Player factotum owns 2-cube:
   63,25  33,14   9,37 -  36,75   8,05   1,08 CL  +0,9276 CF  +0,3037
[  0,29   0,54   0,57 -   0,29   0,17   0,20 CL   0,0168 CF   0,0238]
Full cubeful rollout with var.redn.
1296 games, Mersenne Twister dice gen. with seed 857820087 and quasi-random dice
Play:  2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 6 more moves within equity 0,09
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

No double: +0.5537
Double, take: + 0.3037 (-0.2501)
Double, pass: +1.000 (+0.4463)

No double, take (35.9%)

============================================================
Cube analysis
3-ply cubeless equity  +0,6204 (Money:  +0,5497)
   66,13  28,35   1,96 -  33,87   7,39   0,20
Cubeful equities:
1. No double            +0,6434
2. Double, pass         +1,0000  ( +0,3566)
3. Double, take         +0,5685  ( -0,0748)
Proper cube action: No double, take (17,3%)

Cube analysis
4-ply cubeless equity  +0,5542 (Money:  +0,4847)
   63,71  27,53   1,88 -  36,29   8,08   0,28
Cubeful equities:
1. No double            +0,5581
2. Double, pass         +1,0000  ( +0,4419)
3. Double, take         +0,3991  ( -0,1590)
Proper cube action: No double, take (26,5%)


[close]
The fascist's feelings of insecurity run so deep that he desperately needs a classification of some things as successful or superior and other things as failed or inferior. This also underlies the fascist's embracement of concepts like mental illness and IQ tests.  - R.J.V.

Luck is my main skill

blitzxz

#22
Quote from: Zorba on May 12, 2009, 05:09:42 PM
Obviously, this rule won't help a player who's grossly misevaluating a position much. But that's not the goal of this rule. Also, the examples you give are not in any way contradicted by the Woolsey Rule, they are just situations outside the scope of it.

I think you fail to see the power of Woolsey's doubling rule: When there's any doubt in your mind about whether opponent has a take, you should always double. I consider it the single most best advice on using the cube, and even good players sometimes forget to apply it.

As said, the rule doesn't mean that there aren't other situations with a clear take or clear pass, where you should NOT double. And these are usually harder decisions to make. You will have to use other "tools" for those situations.

The reasons why Kit's rule works so well in practice are partly mathematical, and partly "psychological". Suppose you have indeed some doubt about whether it's a take. What are the scenarios?

If the double/take equity of a position is indeed close to 1, it's nearly always correct to double (certain lopsided matchscores are an exception). Also, it's often a pretty big to very big error NOT to double in these cases. Occasionally, even with the double/take equity close to 1, volatility may be so low that it's not a double yet. However, you will lose very little when you double those. And you may actually gain if opponent errs on the take! So, the rule works well here, preventing you from big errors at an extremely small cost.

If the double/take equity is more like 0.8 and thus a big take, it's often still correct to double these, sometimes by a fair amount, and if it's not, then you still won't lose much usually. BUT!!! Your opponent may well see the position like you did! And as such, have doubts about the take. If he drops, you gain a lot, whether it was a correct double or not. So the Woolsey Rule overall still works very well here.

If the double/take equity is more like 1.2 and thus a big drop, once more it's often still correct to double, and sometimes by a fair amount. It could be too good to double, but usually not by much, so you won't lose much on these. And once again, your opponent may have same problems evaluating as you did, and take your cube. Bingo! Once more, applying the Woolsey Rule stands to gain a lot at very little risk.

Of course, it's possible to misevaluate a position by more than this. Maybe it's a 0.5 take when you think it's close, or a 1.5 drop. Obviously, in those cases, you may lose quite a bit on a misguided double. But once more, it's not totally surprising to see your opponent misevaluate it as well, and the benefits from that only increase! Anyhow, even if you lose on those, you'd probably make errors from gross misevaluations anyway, so the Woolsey Rule is unlikely to make things any worse for you.

So overall, the Woolsey Rule is not a rule to help you evaluate winning chances or equity; it's a general rule on situations where you should double. It reminds you of the small risks in doubling compared to the big gains from either correct cube action by both, or an error from your opponent. It also reminds you of the fact that the question whether to double or not, should start with looking at opponent's take decision.

I'm so stubborn person that you can't change my mind so easily. :D And also I'm raised by bots. I only read about the law when I had already played over year analyzing every game with gnu. Of cource you want to be as close as possible to take point but to know when you're there you have to understand volatility which is the single most important thing when doubling. When vola is high you have to double far from take point because there is still big chance to lose markets and vica versa. And also you want to make hard decisions to opponent but there is no reason to believe that your opponent thinks same as you. To me Woolsey's law is all about this psychology. It tells you nothing about equities, volatility or nothing what's important to me when doubling. Only thing that it tells is that it makes opponents decision hard if he thinks same as me. And if you trust it as general guideline you end up doubling way too early in confusing positions which you don't understand or you easily wait too long for Woolsey positions when you actually should double right away because big volatility. But I'm not trying to sunk the whoie law because the general idea is pretty solid but it just doesn't suite my taste.