Game 2 Move 17 - fourm re-cube?

[ah_clem]

I don't have access to the software needed to post a diagram at the moment, but here's the poll. I'll put up a picture in an hour or two...meanwhile, check the last position and use your imagination.

UPDATE:  Her's the diagram.  Enjoy.  Also, I've attached the latest .sgf file for those interested in doing some rollouts.  I'll probably post a couple of interesting ones when I get the chance.  But let's try not to look at any that are too recent - we don't want to pollute our cube decision. 

BTW, the snow is nice, too bad it's not accumulating in any significant quantity.

[ah_clem]

Spoiler

This is not an easy cube decision.  

In favor of cubing:  We are ahead in the pipcount, ahead by one crossover and on roll.  Our opponent has a stripped 5 point while we have a near perfect distribution. We're looking good here.

Against:  Our pip lead is only four pips or about 6%.  The race lead could evaporate in a single roll of the dice.  Moreover, at this match score a double will be followed by a recube to eight and suddenly we're playing for the match. (I'm pretty sure this is a take - If I thought he would pass I'd turn the cube so fast the 64 would go flying across the room)  Am I willing to stake the match on sheer luck?  Were I playing him myself, sure - my best chance to beat a much better player is to reduce the match to a coin flip.  But here we have enough expertise that I don't want to stake the match on pure dumb luck.

So I say no double.  Yes this is timidity and caution speaking because we have a clear measurable advantage.  For money, I'd double.  If we were behind in the match, I'd double.  At this match score, I'd wait.

No double

[close]

[stog]

[dorbel]

Spoiler

Our doubling window opens at around 63% holding the cube at this score. As Red can pass and retain 25% ME, it closes at 75%. If anybody wants to know how one derives these figures over the board I am happy to expand on this in another post.
So what have we got here? On roll in an even race of this length, with equal distribution we would be 61-62%, a very useful figure to remember. Clearly we are better than that here, leading in the race, without a gap and for what it is worth one crossover ahead. A good sequence can lose the market. No thought required, ship it in. 

[close]

[diane]

The snow is  a lot less questionable now, even on safari    Just as well, or we would have a questionable snowman!!   

[socksey]

Spoiler

We have the pips by a small margin and we have the position.  I say recube before I change my mind!

[close]

socksey



Reminds me of my safari in Africa.  Somebody forgot the corkscrew, and for several days we had to live on nothing but food and water. - W.C. Fields

[NIHILIST]

Spoiler

Knowing that he'll rewhip the cube at this point and put the match on the line for both of us, I'd roll. I don't think I mind terribly losing my market here.

[close]

Bob

[stog]

Spoiler

yes it's an easy take for him....

[close]

[ah_clem]

Spoiler

Our doubling window opens at around 63% holding the cube at this score. As Red can pass and retain 25% ME, it closes at 75%. If anybody wants to know how one derives these figures over the board I am happy to expand on this in another post.

I understand the 25% ME / 75% take point.  I'm don't understand where the doubling window opening at 63% comes from.  I would think that at this match score we'd want a higher than normal GWC in order to double.  Euclidation please?

So what have we got here? On roll in an even race of this length, with equal distribution we would be 61-62%, a very useful figure to remember.

Is this a reference position?  Should I remember: even pipcount + even distribution + about to begin bearoff + on roll -> 62%?  These sorts of mental shortcuts are very useful, but only if you remember them correctly.

Clearly we are better than that here, leading in the race, without a gap and for what it is worth one crossover ahead. A good sequence can lose the market. No thought required, ship it in.

My take is that a market losing sequence is just as likely as a game-losing sequence.  Most sequences will preserve the status quo (or as the race goes on accrue advantage to us, all other things being equal.) so we'll probably still have a cube next roll.  This tells me that rolling is probably not a very big cube error even if it's technically correct to cube now.

I'm still on the side of waiting.  For one thing, he's not going to squirm a bit when we ship the cube - isn't that half the fun of cubing, watching your opponent agonize over whether to take it or not?  (c:

[close]

[dorbel]

Spoiler

Deciding whether you gain by doubling requires what pro-betters call a Risk/Gain analysis. For this you need to divide what you risk by your risk plus what you stand to gain. Eh? What does that mean. The situation here is that we lead 4-away, 5-away and we hold the 2 cube. We can see what we stand to gain and to lose as follows.
(a)No double/win, we lead 3-0, MWC (match winning chances)75%
(b)Double/win we win the match, MWC 100%
(c)No double/lose, we trail 1-2, MWC 43%
(d)Double/lose, we lose the match, MWC 0%
Our risk is the difference between c & d, 43% is what we stand to lose.
Our gain is the difference between a and b, 25% is what we stand to gain.
Risk/Risk + Gain = 43/68 or 63%. This is what we need as a minimum in order to turn the cube.
You get very slightly different results with different ME tables, but the principle remains the same. There is also some degree of simplification here. In line a for example, I have ignored the fact that we may double later.

We could just use Woolsey's rule of course. Are you absolutely sure that this is a take? Some of us are, but I'm not! I think it probably is, but I wouldn't be surprised to be wrong. Flattish full board bearoffs, which this is just about to become, favour the leader by a bit more than the pipcount, having some characteristics of pip races and some of roll races. Pip/roll hybrids as they are known are notoriously tough for the trailer when he can't use the cube.

My take is that a market losing sequence is just as likely as a game-losing sequence.  Most sequences will preserve the status quo (or as the race goes on accrue advantage to us, all other things being equal.) so we'll probably still have a cube next roll.  This tells me that rolling is probably not a very big cube error even if it's technically correct to cube now.

There are no game losing sequences, as we can't be doubled out. Certainly we will usually have a double next turn, but often we won't get a take. If we were only just in the window, we wouldn't lose much by waiting, but here we are a lot better than our minimum.
Look at it another way. In money play, redoubling means that you lose the ability to cash (or make an efficient redouble) later, while your opponent gains the advantage that you have just given away. Here we lose later use of the cube, but the opponent doesn't gain anything like as much as usual. His auto recube only gains him one extra point rather than the normal 4.

[close]

[roadkillbooks]

CUBE CUBE CUBE.

I say this immediately as i was online when schig got his 2-1 and and quickly sent the message asking for a cube poll.  This is very close to a drop.  With the match information shared by dorbel we can see that this is a VERY efficient double.  The only thing possibly holding this back is that if we considered ourselves a collective "much better player" than Schig.  I think this double is so clear that only if we felt we were 200 points better than him we wouldnt have a double.

Each pip here is worth a little less than 2.5%.

If we take the 62% from being even and on roll add say 9.5% more for the 4 pips..plus 1.25% for the cross over.. plus say 2.25% for the 5 point gap. That is 75%....

exactly his take point.  It is as an efficient redouble as you can get.  You must be able to shove the cube when needed. If we roll 8 pips and he rolls 6 hellhave a drop.  So its really a question of if schig wants to take or not.  Since he is the guest player i think it is quite kind to hand him a very eficient double so he can decide if he wants it all to end now or if ed enjoy to play a couple more games.


The cube doesnt even seem to be close.. for me the take is whats questionable. I think he will be squirming here.  Its all in the numbers.  We are 75% to win here.. whether he takes and whips it back or if he drops.  Why not do the same thing in less time? Its not like we have to stop playing forever.

[ah_clem]

Spoiler

Thanks for the refresher on the risk/gain analysis.  I need to internalize this a bit better.

My quibble is with using the minimum doubling point (MDP) as a indication that it's correct to double.  Obviously any GWC below the min double point will make it a mistake to double, but the converse doesn't follow.  For instance, at 0-0 the minimum doubling point is 50%, however 51% is not usually good enough to double - you need more than just GWC > MDP.

I'm still not convinced it's a double.  I'm also not sure that it isn't.

[close]

[ah_clem]

This is a close vote so I'm going to let it run for the full 24 hours. 

And since I have an engagement this evening I won't get to it until about 3 hours after the deadline.  Consider this "extra time"  (or injury time if you prefer) to present your arguments and sway others to your side.

If it's tied when I get back tonight then... then... well, I'll jump off that bridge when I get to it. (c:

[blitzxz]

Spoiler

I agree with dorbel and rkb, but generally I would be quite unhappy to be in this position. We had very short first game and this second game has presented very little difficulties to our opponent. Now we are in position that will most likely decide the whole match in bear-off rolling. In other words we have had very little chance to outplay our opponent. I was looking back and the one thing I disliked before and still dislike is our take in the beginning of this game. Passing was clearly wrong but taking turned the game to highly volatile battle with very little skill. I think passing then or waiting now might be the best strategy even with quite small skill differences. Also in this score/redouble I have found out that lots of people will take clear passes.

[close]

[NIHILIST]

Spoiler

Why would he pass ? For the challenge of playing from 4 - 0 Crawford ? Would you pass ? Would dorbel pass ?

It seems to me the only question is do we want to put the entire match on the line at this point. I prefer to take a shake.

[close]

Bob

[roadkillbooks]

Spoiler

Why would he pass ? For the challenge of playing from 4 - 0 Crawford ? Would you pass ? Would dorbel pass ?

It seems to me the only question is do we want to put the entire match on the line at this point. I prefer to take a shake.

[close]

It will be 3-0.. not 4-0.. i wouldnt even propose the double then cause his take point would be 15% instead of the 25% it is now.

Why wouldnt someone make a perfectly efficient redouble.  blytz has a decent point if you know for sure that the opponent would take after an exchange of say 54 us and then 42... id say lets wait but im afraid that is droparama territory.  I agree though bob if we were up 2-0 with a 2 cube a would not double.

[stog]

Spoiler

we need him to pass on this, if we think he will take, then we must wait until he can't take; - i think we lose equity here, to chance. the skill is to avoid a situation where we have to vote in a throw or 2, to take/pass his rewhip, and be 4-1 down..
now, if we were already 12 ahead in pips, then things may well be different. i vote we wait a throw....

[close]

[diane]

Spoiler

we need him to pass on this, if we think he will take, then we must wait until he can't take

[close]

Spoiler

Erm, that might not happen...and that is why we cube now, whilst the probabilities are still on our side.  The point is good though - sooooo many take these when they shouldn't, and I seem to lose more than the probability suggests should happen

[close]

[ah_clem]

The vote is 6-5.  It's close, but we have a decision for better or worse.  We're shipping the cube.

One positive is that after Schig's take decision we can rollout every previous move without giving anything away.

[dorbel]

Spoiler

My quibble is with using the minimum doubling point (MDP) as a indication that it's correct to double.  Obviously any GWC below the min double point will make it a mistake to double, but the converse doesn't follow.  For instance, at 0-0 the minimhttp://www.fibsboard.com/seasons/snowflake_6.gifum doubling point is 50%, however 51% is not usually good enough to double - you need more than just GWC > MDP.

Worth chasing this one down as part of the learning process. A "normal" doubling window, at the start of a 13pt match let's say, runs from 50% to 78%. You are right of course that 51% wouldn't be enough to double if for example you were ahead after the first roll of the match. This is partly because it would be almost impossible to lose your market, i.e suddenly jump above 78% on the next sequence and partly because giving your opponent the exclusive use of the cube would be enough to make you an underdog and thus drop out of the window on the wrong side! However, there are situations where being just in the window is enough, at the start of a 2 point match for example, where the cube is dead after the turn and where the upper limit of the window is lower than normal, allowing for a few market losing sequences.
Here, two things are different. The window is abnormally small, between 63 and 75%, so even if we were only just in it, market loss would be very easy. Here, rolling 8 pips to Red's 5 would be a market loser. Moreover, we are not just in the bottom part of the window, we are near the top! I think we might be 73% here, rkb thinks we might be 75% All clear?

[close]