Game 2, move 9, roadkillbooks doubles, Forum?

[ah_clem]

Yes, but if he rolls even dice, he prolly won't hit us.  That's what I mean.     I stand by my original vote.   

22 42 44 62 66 all point on our head.  So does 32 and 43. 


What are these even dice of which you speak that prevent him from hitting us?

[Zorba]

Just for clarification: It's great to discuss the moves here and get rollouts etc., that's the learning experience for me. Great if someone can persuade you that an other move or cube decision would've been better. Being able to change your vote afterwards does not really add much to that, in my experience. I think it might sometimes make people's initial votes (and thought processes) lazier. In normal backgammon, you also get only once chance to decide, and also, you have to determine how much time (and energy) to spend on that decision (apart for clocks or impatient opponents  ). If you know you can always change your vote later, it might not benefit your original vote decision, and could therefore be less of a learning experience. Also, it can be interesting for certain positions to see what people's original choices were, it teaches you something about which kind of decisions are the hardest in backgammon.

Just my 2 cents... I don't really mind how people vote, it's up to each to decide for themselves.

And now, I'm gonna NEXT this particular decision...   

[diane]

In normal backgammon, you also get only once chance to decide

And that sums it up, this is NOT normal backgammon, here we have the chance to do something different, to think differently - to hear the voice of others and respond if we wish.  That is why this is so much more valuable than playing a 100 games where our own decisions are final.

[playBunny]

 
Rollout 3888  trials, 2-ply throughout

Spoiler

1 .. Double, take ..... 0.992
2 .. Double, pass ..... 1.000 .. (+0.008)
3 .. No double ........ 0.703 .. (-0.289)
 
Proper cube action: Double, take
But, given the size of the error standard deviation (0.006), take and pass are pretty much the same. 
 
Centered 1-cube:
   68.4%  14.5%   0.6% -  31.6%   9.1%   0.2% CL  +0.462 CF  +0.703
 [  0.1%   0.1%   0.1% -   0.1%   0.1%   0.0% CL   0.002 CF   0.004]
Forum owns 2-cube:
   69.2%  16.2%   1.8% -  30.8%   9.5%   0.3% CL  +1.206 CF  +0.992
 [  0.1%   0.1%   0.1% -   0.1%   0.1%   0.0% CL   0.005 CF   0.006]
 
Full cubeful rollout with var.redn.
3888 games, Mersenne Twister dice gen. with seed 863121964 and quasi-random dice
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]

[close]

[diane]

Spoiler

Proper cube action: Double, take
But, given the size of the error standard deviation (0.006), take and pass are pretty much the same.

[close]

Well, that is well reflected in the closeness of the vote 

But once again  I was right    I need better words to try and persuade you lot 

[blitzxz]

So it's another borderline. Very efficient doubling by roadkillbooks! I would have made major error by doubling too early. It's also surprise to me that we should able to win over 30 % games. I just can't see how that's happening.

[Zorba]

And another rollout to chime in, nearly identical settings:

Spoiler

A huge double, and take/drop is too close to call, a true borderline decision according to GnuBG:

Rollout details:
Centered 1-cube (No double):
   68,38  14,59   0,62 -  31,62   9,15   0,22 CL  +0,4634 CF  +0,7043
[  0,06   0,10   0,05 -   0,06   0,09   0,01 CL   0,0018 CF   0,0034]
Player FIBSBoard forum owns 2-cube (Double, take):
   69,29  16,15   1,69 -  30,71   9,66   0,30 CL  +1,2046 CF  +0,9947
[  0,07   0,14   0,15 -   0,07   0,08   0,04 CL   0,0051 CF   0,0061]

Full cubeful rollout with var.redn.
3888 games, Mersenne Twister dice gen. with seed 863038590 and quasi-random dice
Play:  2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 16 more moves within equity 0,2
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

[close]

[socksey]

But once again    I was right    I need better words to try and persuade you lot   

Well, maybe me, too!   

Cheers, diane!

socksey



"The consequences of our crimes long survive their commission, and, like the ghosts of the murdered, forever haunt the steps of the malefactor" - Sir Walter Scott

[stog]

good voting by all

[adrian]

Hi all,

I was the moron who voted in 10th of july at about 18PM for take, when the score was 6-5 in favour of a pass. All was because I am not very used to play having homeboard at my left side. I always play with the homeboards on the right side of the board.
Sorry guys, even for me (as dumb lucky as I am) it is a pass, so I tried today to withdraw my initial vote for the take. It didn't worked 

[diane]

Sorry guys, even for me (as dumb lucky as I am) it is a pass, so I tried today to withdraw my initial vote for the take.

No, it was a take - the bot showed that.  Whatever happened with your vote, it didn't affect anything, it stayed 6-5 to pass - and so we passed 

[diane]

Well, maybe me, too!   

Yes, us girls knew which was the right way - we need to get them boys paying more attention   

[playBunny]

No, it was a take - the bot showed that. 

Strictly speaking what the bot showed is that it's about 90% likely that the equity for taking is better for us than dropping and 10% that it's not. (That's going by my rollout. Zorba's indicates that it's more even). This is given by the standard deviation of error that the bot reports (the 0.006 that I highlighted in the rollout) along with the equity (0.992). It's a measure of how accurate the equity is. The lower the standard deviation of error the more likely the true equity is to be close by. A "standard error" of 0.006 isn't that low but it's not too bad in this situation given that we are comparing one estimated equity with a defined one (dropping = -1.000). When comparing two close moves, each with an estimated equity, the standard error would preferably be smaller still to give greater confidence in the gap reported between the moves.

That's the stats of the matter but the facts of the matter are that a difference of 0.008 is dwarfed by other, incalculable, factors such as how well we'd expect ourselves and our opponent to handle the game scenario that follows a take.

[diane]

That's the stats of the matter but the facts of the matter are that a difference of 0.008 is dwarfed by other, incalculable, factors such as how well we'd expect ourselves and our opponent to handle the game scenario that follows a take.

Or - a borderline take, in a nutshell. 

And why all this ongoing negativity around our performance - results so far indicate the forum more often than not finds the right moves.  And it would happen even more, if you only payed more attention to me 

[playBunny]

Or - a borderline take, in a nutshell.

What do you want nutshells for when you can waffle on for paragraph after paragraph? (You should see the version of the post that I whittled down to leave that wee morsel! ;-) One day I shall do a web page about GnuBg's standard deviations of error and j.s.ds. Just as soon as I fully understand them.  )

And why all this ongoing negativity around our performance

Ah, that's interesting. Did you read my neutral statement and see a negative judgment about ourselves and perhaps also a positive one about our opponent? Hmmmm..... Lie on this couch here and tell me about your childhood.

results so far indicate the forum more often than not finds the right moves.

But you're right. If we take away our bloops and blunders then we're putting on a World Class performance! 

And it would happen even more, if you only payed more attention to me 

Well, you've certainly got me rapt around your little finger. 

[diane]

 I really only have one over riding sentiment with regard to that post............   



Ok, it seems I read more into what you were saying about our performance than you actually said, can I just put it down to a very early start at work this morning 

I do feel there are enough contributers and enough discussion that the forum is getting to the right moves in these matches, and that like any good world class player, we should be playing to the numbers not superstition about potential errors 

[playBunny]

I really only have one over riding sentiment with regard to that post............   

Well, I was rather more interested in exploring your mind and thinking processes but if you want me to check your tongue .... say Ahhhhh.

I do feel there are enough contributers and enough discussion that the forum is getting to the right moves in these matches, and that like any good world class player, we should be playing to the numbers not superstition about potential errors 

Superstition?! Would you care to place a small wager on us and roadblockskill playing GnuBg-perfectly for the rest of the match?

Actually, over on bgonline.org, you will hear world class players talking about taking their opponent's skill level into account. 

[diane]

I think we better finish this conversation over a match sometime 

[roadkillbooks]

Actually, over on bgonline.org, you will hear world class players talking about taking their opponent's skill level into account. 

I was hoping for the drop.  While I obviously had good chances my situation had begun to go down hill and I could see a fast redouble under a number of scenarios.  Snowie on mini-rollout says this is a take and the drop was a -.11 error.  I think I didnt really have any gammon potential.  Forum probally had about as high gammon possibilities as I did so the take comes down to raw winning chances. 
I don't think one has to be world class to take skill level into account.  Many people clearly do different cube actions based on playing different people.  This makes complete sense and is correct.  Cube action is based on match equities so while say 3away, 3away is 50% winning chances....well if you are playing a very weak opponent,  Your winning chances at 3away, 3away are higher than 50%.maybe let's say 60%.  This totally changes what is a take/drop at prior match scores.  The same is true the other way around. I think some sort player based strategy should always be used.  One can always revert back to "normal" if one considers the opponent an equal.
Anyway compared to what cube errors can look like.. this was small.