Game 6, move 4 : Forum 5-1

[diane]

Forum, blue to move 5-1

This is more tricky I think - reflected by a lot of options, so hopefully we will get some good discussion on what on earth to do with this, and why seemingly very similar moves can be right or wrong.

mGfGESDgc/AANA:8ImmACAAIAAA

[ah_clem]

Spoiler

 

Where to play the one?

23/23 is a terrible way to play the one - making the two point is pretty much useless.  Sloting the three is worthwhile.  The other way to play the one is to slot our own 5 point - slotting the bar point is not nearly as goos as it strips the eight.

Where to play the 5?

We can either hit on our one point or play it safe with 13/8.

This leaves four combinations, but we can throw 6/5 6/1* out as too aggressive.

With us being 30 pips behind and gammons not counting,  I think an action play is called for.  13/8 23/22 is too passive for the situation.

That leaves 6/1* 23/22 and 13/8 6/5.  I like the latter  If we can blockade th runner we might have a game. 

[close]

[diane]

Spoiler

  23/23 is a terrible way to play the one

[close]

er, what?   

Spoiler

I like 23/22, I do hope you dont mean that....

[close]

[ah_clem]

Spoiler

> 23/23 is a terrible way to play the one

er, what?  
I like 23/22, I do hope you dont mean that....

well, yes, 23/23 is a terrible move.  In the same category as "king's bishop to Albuquerque"

I meant 24/23, making the two point. Making the two point is weak.

23/22 is much better - at least it slots a point we might actually want.

[close]

[stog]

Spoiler

mmmm i might be able to persuade the idiot who wanted to make the 2 point to change his vote to the more useful 3 point, if diane could add "remove vote" or is it preferable to only allow such change as the vote nears end and and a split decision?

[close]

[diane]

Spoiler

if diane could add "remove vote"

[close]

sowwreeee

Change away 

[diane]

Spoiler

well, yes, 23/23 is a terrible move.  In the same category as "king's bishop to Albuquerque"

[close]

[stog]

Spoiler

np thx! but i wonder whether we should only add the remove vote option towards the end of voting anyway? (ps i spoke to my inner idiot!)

[close]

[dorbel]

Spoiler

Ah Clem doesn't say why 24/23 is so bad. It is an anchor, always handy when behind and it's a better anchor than the 24pt. However, I agree that we can find a better 1 somewhere else so early in the game. I don't mind 13/8, 6/5, but it leaves the three checkers at the back somewhat isolated. We might try 24/18, bidding to make it next turn or at least exchange hits there, but I would try 23/18, 6/5! It keeps the anchor so we can't be blown away and although one of these blots is going to get hit, we get return shots and may even be able to steal one of the points with a good number from the bar. We can't be doubled out and don't much care if we get stuck with a gammonish game, so now is an opportunity, perhaps the last that we will get, to make a big leap forward.

[close]

[stiefnu]

Spoiler

but I would try 23/18, 6/5!

An interesting suggestion but, before I'm persuaded to change, I'd have to see there'd been least one vote for this move! 

[close]

[ah_clem]

Spoiler

Ah Clem doesn't say why 24/23 is so bad. It is an anchor, always handy when behind and it's a better anchor than the 24pt. However, I agree that we can find a better 1 somewhere else so early in the game. I don't mind 13/8, 6/5, but it leaves the three checkers at the back somewhat isolated. We might try 24/18, bidding to make it next turn or at least exchange hits there, but I would try 23/18, 6/5! It keeps the anchor so we can't be blown away and although one of these blots is going to get hit, we get return shots and may even be able to steal one of the points with a good number from the bar. We can't be doubled out and don't much care if we get stuck with a gammonish game, so now is an opportunity, perhaps the last that we will get, to make a big leap forward.

I trying to remember where I read it (but I can't at the moment), so here's what I recall about the deuce point: it's a deep anchor that doesn't provide much outfield coverage, can easily be primed, and usually already faces an opponent's point 6 points away on the eight.  This adds up to the two point being not much better than the ace point. Unless you're just looking for a place to waste a 1-roll, switching points from the ace to the deuce doesn't accomplish much.

By contrast, the 3 point observes the 9 point giving better outfield control and is harder to prime.  It's a much better point to hold than the deuce.

If we didn't have the ace point already, making the deuce-point would make sense.  In this position and at this match score making the deuce seems wrong - it's too passive. A deuce point game is a recipe for losing a single point (i.e. it prevents you from losing a gammon, but provides little winning advantage).  That's not what we want to do here.

Gammons don't count.  We're 30 pips behind.  We need to make something happen, not play conservatively and lose a single point. Hit his blot, slot a useful point or two. Dorbel's suggestions all accomplish this although I'm leery of moving to the opponent's bar point - too many checkers bearing down on it with the option to hit loose.  I'm not sure which action play is best, but something has to be better than the passive play.

[close]

[ah_clem]

i wonder whether we should only add the remove vote option towards the end of voting anyway?

In the spirit of the game of backgammon where you are allowed to change your mind until you pick up the dice, we should allow people to change their minds anytime before the vote closes.

The hard part is to remember to check the box everytime you put up a poll.  Is there any way to configure the server to have it checked it by default?

[diane]

The hard part is to remember to check the box everytime you put up a poll.  Is there any way to configure the server to have it checked it by default?

OOooh, nice suggestion...is there master stog? is there?  *holds breath*

[stog]

u mean a master checkbox that auto ensures all polls don't need checkmarking........


no



but then again polls where u don't want folks changing their minds are default!

[dorbel]

Spoiler

Good post Clem. One advantage that the 23pt has which you didn't mention, is that it freezes the 8pt, so that the opponent can't use a checker from there to make a point, unless he has a spare of course. I'm not arguing for it though, but I do think it to be as least as good as slotting the 22pt. I doubt if we are going to see big differences between any of these plays, so if several plays appeal and all appear to be equally strong, I like to take the most complex play. I can't see 23/22, 13/8 giving rkb many problem rolls!

[close]

[ah_clem]

gnubg rollout

Spoiler

Lot's of possibilities here, so I rolled out a bunch of moves.  

gnubg prefers an action play  - slot a valuable point or hit the blot.  The top three moves are a statistical tie - the "best" move slots the 5, the second "best" slots the opponent's bar point, and the third "best" slots both.  Immediately behind (.019) these is the hitting play 24/23 6/1*.  Even slotting our own bar point is not too bad at -.024.  OTOH, the hyper aggressive  6/5 6/1*  ranks way down in 13th place.

Bottom line - make an agressive play, but not too aggressive.

The passive choice of the Forum, 13/8, 23/22 rolls in in tenth place as a  .042 error.  It's close sister and second place vote getter, 13/8 24/23, does much better at fifth place and .024 error

My apologies if my ranting about the evils of making the two point swayed the group in the wrong direction.  24/23 is stronger in every case than comparable moves with 23/22.  My mistake - I probably should adjust my attitude towards the two point.

And while I didn't post about it, I see that I undervalued the utility of slotting the opponent's bar point - in retrospect I see that if we can make that point we go into a holding game making our race deficit much less of a liability.

Surprisingly, 24/23 8/3 is only a .024 error - that's a bot move if ever there was one.

Note that with a 0-ply rollout, none of this is dispositive, and even if you believe the numbers observe that there are a lot of moves that are statistical ties.

I'd be interested in hearing some other's takes on these results, or seeing a more precise rollout.


   1. Rollout          13/8 6/5                     Eq.:  -0.338
      0.351 0.090 0.012 - 0.649 0.277 0.011 CL  -0.338 CF  -0.338
     [0.003 0.004 0.002 - 0.003 0.003 0.002 CL   0.006 CF   0.006]
       
   2. Rollout          24/18                        Eq.:  -0.341 ( -0.003)
      0.349 0.096 0.013 - 0.651 0.273 0.011 CL  -0.341 CF  -0.341
     [0.002 0.004 0.002 - 0.002 0.003 0.001 CL   0.005 CF   0.005]
     
   3. Rollout          23/18 6/5                    Eq.:  -0.345 ( -0.007)
      0.351 0.092 0.012 - 0.649 0.301 0.013 CL  -0.345 CF  -0.345
     [0.003 0.004 0.002 - 0.003 0.003 0.002 CL   0.006 CF   0.006]
       
   4. Rollout          24/23 6/1*                   Eq.:  -0.357 ( -0.019)
      0.339 0.090 0.009 - 0.661 0.256 0.010 CL  -0.357 CF  -0.357
     [0.003 0.004 0.002 - 0.003 0.003 0.001 CL   0.006 CF   0.006]

   5. Rollout          24/23 13/8                   Eq.:  -0.361 ( -0.023)
      0.339 0.097 0.010 - 0.661 0.241 0.011 CL  -0.361 CF  -0.361
     [0.002 0.004 0.002 - 0.002 0.003 0.001 CL   0.006 CF   0.006]
       
   6. Rollout          24/23 8/3                    Eq.:  -0.361 ( -0.023)
      0.343 0.089 0.010 - 0.657 0.290 0.014 CL  -0.361 CF  -0.361
     [0.003 0.004 0.002 - 0.003 0.003 0.001 CL   0.006 CF   0.006]
       
   7. Rollout          13/7                         Eq.:  -0.362 ( -0.024)
      0.338 0.086 0.007 - 0.662 0.293 0.010 CL  -0.362 CF  -0.362
     [0.002 0.004 0.003 - 0.002 0.003 0.002 CL   0.006 CF   0.006]

   8. Rollout          23/18 8/7                    Eq.:  -0.371 ( -0.033)
      0.337 0.088 0.009 - 0.663 0.307 0.013 CL  -0.371 CF  -0.371
     [0.003 0.004 0.001 - 0.003 0.004 0.002 CL   0.007 CF   0.007]
 
   9. Rollout          8/3 6/5                      Eq.:  -0.374 ( -0.036)
      0.345 0.085 0.012 - 0.655 0.344 0.020 CL  -0.374 CF  -0.374
     [0.003 0.003 0.002 - 0.003 0.004 0.002 CL   0.008 CF   0.008]
       
  10. Rollout          23/22 13/8                   Eq.:  -0.380 ( -0.042)
      0.328 0.086 0.010 - 0.672 0.247 0.010 CL  -0.380 CF  -0.380
     [0.002 0.003 0.002 - 0.002 0.003 0.001 CL   0.005 CF   0.005]
       
  11. Rollout          23/22 6/1*                   Eq.:  -0.380 ( -0.042)
      0.332 0.084 0.010 - 0.668 0.269 0.014 CL  -0.380 CF  -0.380
     [0.003 0.004 0.002 - 0.003 0.003 0.002 CL   0.007 CF   0.007]
       
  12. Rollout          23/22 8/3                    Eq.:  -0.383 ( -0.045)
      0.332 0.087 0.013 - 0.668 0.308 0.013 CL  -0.383 CF  -0.383
     [0.003 0.004 0.002 - 0.003 0.003 0.002 CL   0.006 CF   0.006]
       
  13. Rollout          6/5 6/1*                     Eq.:  -0.385 ( -0.047)
      0.336 0.080 0.009 - 0.664 0.333 0.018 CL  -0.385 CF  -0.385
     [0.003 0.004 0.002 - 0.003 0.004 0.002 CL   0.008 CF   0.008]
     
  14. Rollout          8/7 6/1*                     Eq.:  -0.395 ( -0.057)
      0.329 0.084 0.010 - 0.671 0.296 0.016 CL  -0.395 CF  -0.395
     [0.003 0.004 0.002 - 0.003 0.004 0.002 CL   0.008 CF   0.008]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 872235963 and quasi-random dice
       Play: 0-ply cubeful prune [expert]
       Cube: 0-ply cubeful prune [expert]

[close]