Game 3, move 22: Forum 5-4

[diane]

Forum (blue) to move 5-4 [sorry no 4-4  ]

c93AAAPbzQUYAA:QYmyACAAAAAA

[ah_clem]

Spoiler

Pay now or Pay Later? 

15/6 is the "pay now" option.  We start to clear our back point by moving one checker te safety.  We leave 2 blots, but we duplicate 2s, reducing the number of shots to 16.  (did I get it right this time?)  rkb has a four point board with a blot, which will soon become a 5 point board - if we're going to be hit, now is better than a move or two from now.

There are two "pay later" options: 7/3, 6-1 and 7/2, 6/2.  I don't see much difference between the two, so I'll consider them together as the "safe play".  The safe play puts off leaving a shot for another move.  Unfortunately, there are few rolls that force rkb to leave a shot.  It looks pretty likely that safe play will just cause us to leave a shot (or 2!)  next roll.

Still, two blots are harder to clean up than one, and two checkers on the bar is very gammonish.  That steers me to the safe play, even though we'll probably be paying soon against a 5 pt board.

So, which safe play?  Spares on the 3 and 1 look slightly better than two spares on the 2, so 7/4 6/1

[close]

[diane]

Spoiler

There are two "pay later" options: 7/3, 6-1 and 7/2, 6/2.  
So, which safe play?  Spares on the 3 and 1 look slightly better than two spares on the 2, so 7/4 6/1

[close]

Spoiler

er, that option isn't there, but I think I get the gist 
agree about pay now or later though - and I am going for later too  , albeit a different way!

[close]

[ah_clem]

Spoiler

er, that option isn't there, but I think I get the gist 
agree about pay now or later though - and I am going for later too  , albeit a different way!

Yeah, I meant 7/3, 6-1. Sloppy typing.

Anybody care to offer an opinion on why 7/2, 6-2 is better?  I doubt there's much difference, but I'm curious what would point most votes in that direction.

[close]

[diane]

Spoiler

Working as I do with every move, when there is something I dont like, I eliminate the worst cases until we have a least unpleasant option.  The 6-1 is the one I dont like, because putting checkers on the ace point, particularly when it is stacking and not making a point, whilst we still have men out there to bring home, feels like a bad idea.  I did angst over this one for ages and then go with my gut feel - it would be nice to see if there is an answer to whether there is any real advantage to 2 on the 2 point over 1 on the ace and 1 on the 3?
It is one of the things I considered when I was making up the poll, it would be nice to see this discussed and reasoned out.

[close]

[ah_clem]

Spoiler

The 6-1 is the one I dont like, because putting checkers on the ace point, particularly when it is stacking and not making a point, whilst we still have men out there to bring home, feels like a bad idea. ... it would be nice to see if there is an answer to whether there is any real advantage to 2 on the 2 point over 1 on the ace and 1 on the 3?

Agree that burying a checker by placing a spare on the 1 point is not very useful in a contact position.  But spares on the 2pt are just as "burried" when the one point is already made - there's nowhere for them to go other than  off or the one point.  My take is that a spare on the 3 allows us to "waste" a 2 roll at some point which might be useful if we get an awkward roll. Stacking the 2pt only allows us to waste a 1 roll.  The slightly more flexible position steered me towards the spares-on-3-and-1 play.

As I said, I doubt there's much difference between the plays, but seeing how lopsided the vote is I feel I must be missing something...

[close]

[Whaat]

Spoiler

Ah_clem - I think you're right. Leaving the spares on the 3 point allows us to play either four '1s' or two '2s'. In contrast, the alternative move allows  four '1s' but only one '2'. 

As someone who played 6/2, my reasoning at the time was: we want flexibility --> spares generally equal flexibility --> a checker on the ace by definition is not a spare --> pile them on the 2 point.

I suspect this train of thought was fairly common...

[close]

[diane]

Spoiler

Ah_clem - I think you're right. Leaving the spares on the 3 point allows us to play either four '1s' or two '2s'. In contrast, the alternative move allows  four '1s' but only one '2'. 

As someone who played 6/2, my reasoning at the time was: we want flexibility --> spares generally equal flexibility --> a checker on the ace by definition is not a spare --> pile them on the 2 point.

[close]

yes...maybe I am persuaded 

[diane]

We have a clear leading move, and voting closes today.  I will leave it a short while longer, to see if anyone wants to change their vote, based on ah_clem and whaats comments.  I am persuaded, but just me changing my vote won't alter the outcome.

[Zorba]

Thanks diane for continuing the match smoothly while I was on vacation, good job! 

Shall I take over again from here, or are you having so much fun that you want to continue?

Oh and to stay on topic, here's my move:

Spoiler

Paying now has something going for it with the duplication of twos and a tough position to bring home if we don't pay. Still, I don't like the two blots it leaves. We can survive getting hit ( as already happened before), but if both checkers get sent back we're in serious trouble.

rkb has to move too, and his next move might detoriate his position, making a future hit less damaging.

From the safe moves, it looks like a wash between 7/2 6/2 and 7/3 6/1. The latter gives us an extra two to play, but do we need more than one? Stacking two on the twopoint is likely a bit better in a race; checkers on the acepoint are the worst for that, and generally stacking on even points is more efficient for the bear off than stacking on odd points. If we get hit again and escape at some point, the race might become close and then it could matter. But that's quite a parlay and the difference will be very small anyhow.

So I think it's a cointoss; mine landed on 7/2 6/2.

[close]

[diane]

Thanks Zorba - I can continue for now, but am going away at some point soon [dates not yet finalised], so you will have to pick it up then, and maybe start my one for me - so a fair swap 

[ah_clem]

gnubg results:

Spoiler

Zorba is correct that paying now has merits.  It's gnubg's choice on both analysis and rollout. The safe plays are pretty close with the forum's choice ( stacking on the 2)  a close third.  Not a blunder though.

Why pay now? rkb has very few rolls that make him leave a shot, so we're probably going to be faced with clearing our trailing point next move.  Now we face a four point board with a blot and next move it'll probably be a 5 point board.  So now is our best chance to clear since it'll be worse later. The duplication of 2's is icing on the cake.




Analysis:

    1. Cubeful 2-ply    15/6                         Eq.:  +0.279
       0.642 0.079 0.001 - 0.358 0.050 0.001
       
    2. Cubeful 2-ply    7/3 6/1                      Eq.:  +0.257 ( -0.022)
       0.645 0.053 0.000 - 0.355 0.047 0.001
       
    3. Cubeful 2-ply    7/2 6/2                      Eq.:  +0.256 ( -0.023)
       0.643 0.055 0.000 - 0.357 0.051 0.001
        2-ply cubeful prune [world class]

Rollout

    1. Rollout          15/6                         Eq.:  +0.250
       0.630 0.078 0.001 - 0.370 0.065 0.001 CL  +0.341 CF  +0.250
      [0.001 0.002 0.000 - 0.001 0.003 0.000 CL   0.004 CF   0.005]
       
    2. Rollout          7/3 6/1                      Eq.:  +0.230 ( -0.020)
       0.629 0.055 0.001 - 0.371 0.062 0.003 CL  +0.293 CF  +0.230
      [0.001 0.001 0.000 - 0.001 0.002 0.001 CL   0.004 CF   0.005]
       
    3. Rollout          7/2 6/2                      Eq.:  +0.222 ( -0.028)
       0.626 0.055 0.001 - 0.374 0.065 0.003 CL  +0.282 CF  +0.222
      [0.002 0.001 0.000 - 0.002 0.002 0.001 CL   0.004 CF   0.005]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 867797248 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[Whaat]

Just a thought - does the safe play make things more complex for rbk?

And if so, does any expected error from RKB compensate for our technical error?

I'd be interested in any comments...

[Zorba]

and a 2-ply rollout:

Spoiler

Paying later looks like an error. Paying now wins a few more games, but most of all it wins 2.5% extra gammons, probably because of the possible exchange of hits.

    1. Rollout          15/6                         Eq.:  +0,2272
        61,78   8,19   0,16 -  38,22   6,60   0,33 CL  +0,3172 CF  +0,2272
      [  0,08   0,09   0,02 -   0,08   0,23   0,08 CL   0,0020 CF   0,0025]
    2. Rollout          7/3 6/1                      Eq.:  +0,1949 ( -0,0323)
        61,54   5,57   0,09 -  38,46   6,24   0,36 CL  +0,2643 CF  +0,1949
      [  0,08   0,11   0,01 -   0,08   0,16   0,08 CL   0,0020 CF   0,0024]
    3. Rollout          7/2 6/2                      Eq.:  +0,1946 ( -0,0327)
        61,48   5,66   0,10 -  38,52   6,13   0,42 CL  +0,2629 CF  +0,1946
      [  0,08   0,12   0,01 -   0,08   0,14   0,09 CL   0,0020 CF   0,0023]
        Full cubeful rollout with var.redn.
        2592 games, Mersenne Twister dice gen. with seed 866958214 and quasi-random dice
        Play:  2-ply cubeful prune [world class]
        keep the first 0 0-ply moves and up to 12 more moves within equity 0,16
        Skip pruning for 1-ply moves.
        Cube: 2-ply cubeful prune [world class]

[close]

[Zorba]

Just a thought - does the safe play make things more complex for rbk?

And if so, does any expected error from RKB compensate for our technical error?

I'd be interested in any comments...

A lot of rolls seem straightforward to play even after our safe play, but a few of them present interesting problems I guess: 3-3, perhaps 4-4, 5-1, maybe 5-3, 5-5, 6-2, 6-5, 6-6.

That's 12 out of 36. Since the safe play looks like a ~0.03 error, rkb would have to make a 0.09 error on average on all of these "problem rolls" for us to break even. That seems highly unlikely: rkb probably won't err on all 12 of these difficult rolls, and even when he does, it's probably not a 0.09 error or worse he's gonna be making.

So it looks like the best play is simply the best play here.