Game 2, move 3, Herd to play 4-1

[Zorba]

The Herd played the 3-1 8/7* 5/2*, dorbel rolled a 6-2 to enter one checker from the bar and hit our blot B/23*.
Now it's the Herd to play 4-1 from the bar:



XGID=a-a---EAB---dD---d-e----BA:0:0:1:41:0:4:1:5:10
GNUBg Id: 4PPgAUjga/AAWA:8ImwAEAAAAAA

[ah_clem]

Spoiler

bar/21 8/7  Mobility and making points in order.

We don't want a third checker on the ace point, and playing into our homeboard is foolhardy.  The only other play is bar/20, which is probably about the same equity as bar/21 8/7 .

Actually, bar/21 6/5 may be playable, it just seems too risky for my stomach.

[close]

[diane]

Spoiler

bar/21 8/7  Mobility and making points in order.

We don't want a third checker on the ace point, and playing into our homeboard is foolhardy.  The only other play is bar/20, which is probably about the same equity as bar/21 8/7 .

Actually, bar/21 6/5 may be playable, it just seems too risky for my stomach.

[close]

Its me and you kid...trying to convince the Herd...

[Zorba]

Spoiler

Is it wise to keep hitting here? We have only 8 checkers left "in the zone" for a blitz, and we haven't made any homeboard points yet. On the other hand, dorbel has that one checker on the bar and we could make it two, which is always worth something. Still, with just that one point board of us, dorbel has 25/36 (bit more than 2/3) chance to simply enter both checkers. And of course, he's also 12/36 (1/3) to hit our blot then. How bad is it if that blot gets hit? Well, pretty bad. We'd have four checkers back, of which three are stuck on the acepoint. Oh yes, and we have another blot dangling on our barpoint that's not guaranteed to find safety.

So it looks like the loose hit 6/2* rarely works, doesn't give a very big advantage when it works, and is rather bad the many times it doesn't work. No thanks.

Let's use this roll to develop our backcheckers; we now have three of them and with dorbel on the bar it's relatively safe to move one up. B/21 8/7 or B/20? There's things to say for both plays. I have a tiny preference for B/20. With dorbel's backman on our 2pt, it might be a tad better to keep our 8pt and block sixes for that checker. This gives dorbel bad sixes from the bar, as he can't use sixes to enter, and 10 out of 36 rolls enter and contain a six.

[close]

[Krazula]

I voted for b/24 6/2*, it makes it so dorbel can't accomplish much with his next roll. It doesn't matter much if we're hit since dorbel hasn't made any points in his homeboard yet. Also this play unstacks from our 6point. I think b/21 6/5 is also as good, it unstacks from the 6point and puts a checker where we want it. I'm pretty sure we don't want to play cowardly with b/21 8/7 or b/20. This is the opening game and we should play aggressively keeping dorbel offbalance and try to get an early edge. Its hard to get the early lead if you're not willing to take risks.

[ah_clem]

@ Zorba

No, it is not wise to keep hitting here.  I didn't even really consider the hit, and now the more I look at it, the worse it seems from both a strategic and tactical perspective.  Burying a checker on the two and making a stack on the 24 is just *bad* (strategy), and the myriad of return hits and lack of checkers in the zone means the hit  won't accomplish anything (tactics).

Now that you've pointed out that bar/20 gives Dorbel bad sixes, I'm inclined to favor your choice over mine, but my hunch is that they're pretty close.  I'd change my vote if I thought it would do any good.

[blitzxz]

This looks bad, very bad. When you make errors it tends to lead to more difficult positions and even more errors. I would hate to play this and propably that's why I forgot to vote. I would have hit though too.

We already started our crazy hitting plan so let's take it to the bitter end. We have double edged plan here. Either we get really lucky and dorpel dances at least couple rolls while we anchor and build our home board OR we try to pick our wounded from the suicide attack and make the final harakiri in massive back game (if we get lucky and have even anything else then the one point). The latter would actually be ideal in my point of view as there is endless amount of possibilities for errors in back games and I could endlessly argue how the 2-ply rollouts are once again wrong in their analyzes.  

Little offtopic, but another thing is that I thinks that a player should always have an overall plan and usually also clear back-up plan when playing backgammon. Just looking into the technical aspects (like unstacking or hitting numbers and so on) you get easily lost. Even a bad plan is better then no plan at all.

What would be our plan after splitting the backmen? Propably getting to some kind of bad holding game if we are luck and winning with some kind of lucky late shot. Yeah, that might be correct but I like more about the crazy plan. Also gammons don't count so there is absolutely no reason to be afraid of too many backmen.

[ah_clem]

I'm not particularly afraid of too many back men. But I am wary of stacking three men on the ace and then moving a checker too far forward.  It's not so much that the checker on the two is likely to get hit, it's that even when it doesn't get hit it's in the wrong place.  From a connectivity perspective, the hitting play just looks awful - checkers strewn at both ends of the board without any communication.

The upside is that it puts him off-balance; the downside is that it also puts us off balance with our off-balanced-ness being more persistent.

[blitzxz]

checkers strewn at both ends of the board without any communication.

The communication is through the bar if needed. 

[ah_clem]

Voting is closed, and the vote is pretty lopsided, so here's the rollout:

Spoiler

Surprise!  The top two plays according to 0-ply rollout did not get any votes at all.   The hitting play is fourth and not that big of an error.  Zorba's choice (bar/20) is within the margin of error of the top plays.  We probably need a better rollout to distinguish among them.

Why is bar/21 6/5 rated so highly?  From a "pure play" perspective, it's the hands down winner.  It slots two points that we want to make (the 5 and 7), and the checker on the 21 gives great mobility.  The other top choice spreads the runners around, also giving great mobility.  I'd also surmise that spreading the runners around facilitates making good backgame points.

Bar/21 7/6 is surprisingly not bad.  It looks awful.



    1. Rollout          bar/21 6/5                   Eq.:  -0.063
       0.462 0.113 0.004 - 0.538 0.300 0.080 CL  -0.063 CF  -0.063
      [0.005 0.003 0.001 - 0.005 0.007 0.006 CL   0.009 CF   0.009]

    2. Rollout          bar/21 24/23                 Eq.:  -0.066 ( -0.003)
       0.461 0.114 0.004 - 0.539 0.260 0.041 CL  -0.066 CF  -0.066
      [0.003 0.002 0.001 - 0.003 0.007 0.004 CL   0.006 CF   0.006]

    3. Rollout          bar/20                       Eq.:  -0.066 ( -0.003)
       0.461 0.116 0.004 - 0.539 0.264 0.051 CL  -0.066 CF  -0.066
      [0.003 0.002 0.001 - 0.003 0.007 0.004 CL   0.006 CF   0.006]

    4. Rollout          bar/24 6/2*                  Eq.:  -0.079 ( -0.016)
       0.454 0.128 0.005 - 0.546 0.268 0.074 CL  -0.079 CF  -0.079
      [0.004 0.002 0.001 - 0.004 0.007 0.005 CL   0.008 CF   0.008]

    5. Rollout          bar/21 7/6                   Eq.:  -0.081 ( -0.018)
       0.454 0.118 0.003 - 0.546 0.229 0.045 CL  -0.081 CF  -0.081
      [0.003 0.002 0.002 - 0.003 0.006 0.004 CL   0.006 CF   0.006]

    6. Rollout          bar/21 8/7                   Eq.:  -0.094 ( -0.031)
       0.448 0.111 0.003 - 0.552 0.259 0.058 CL  -0.094 CF  -0.094
      [0.003 0.003 0.001 - 0.003 0.006 0.005 CL   0.006 CF   0.006]
 
    7. Rollout          bar/24 13/9                  Eq.:  -0.099 ( -0.036)
       0.445 0.119 0.004 - 0.555 0.286 0.075 CL  -0.099 CF  -0.099
      [0.003 0.002 0.001 - 0.003 0.006 0.005 CL   0.007 CF   0.007]
        Full cubeful rollout with var.redn.
        1296 games, Mersenne Twister dice gen. with seed 890851118 and quasi-random dice
        Play: 0-ply cubeful prune [expert]
        Cube: 0-ply cubeful prune [expert]

[close]

[blitzxz]

Why is bar/21 6/5 rated so highly?  From a "pure play" perspective, it's the hands down winner.

This move shares almost the same plan that I stated. However spliting increases anchor chances for our back game (and might leave open the holding game also by the way) and instead of crazy blitzing it goes for priming. Extremely long and high ply rollout is needed to say anything because this is the committing move of which plan we take early in the game.

[ah_clem]

This move shares almost the same plan that I stated. However spliting increases anchor chances for our back game (and might leave open the holding game also by the way) and instead of crazy blitzing it goes for priming. Extremely long and high ply rollout is needed to say anything because this is the committing move of which plan we take early in the game.

Agree that blitzing is a mug's game and that the best bet is a priming strategy.

FWIW, I've started a 2-ply rollout which gnubg is estimating will be done sometime Wednesday.

[Zorba]

I'll run XG, also at higher settings, as time allows.

BTW, the plan with B/21 or B/20 is of course not to play a bad holding game. With the 21 or 20pt anchor, it's nearly always a good holding game to have, especially since dorbel hasn't got any anchor yet. The idea of a high holding game is not to get a late shot, it's to make opponent's life difficult in his outfield, hit an early shot perhaps or simply run later when one is ahead in the race. Also, the high anchor prevents the backmen from getting primed or attacked, for a long time to come. This can actually lead to much better blitzes later on, as you can take a lot more risk, compared to a situation where you have three men stacked on your 24pt.

B/21 6/5 could be dubbed a bot play, it's very hard for humans to spot plays like that. Often it's wrong to both split and slot, but the bots show us exceptions every now and then. It's very pure, and that usually works out well at quasi-DMP. But has anyone counted how many shot and double-shot numbers it leaves?

[ah_clem]

gnubg 2-ply rollout

Spoiler

The "bot" play is still on top and the next three remain in the same order as before, but there is  more separation  between the top play and the also-rans.

The passive-looking  bar/21 24/23 does surprisingly well again.  



   1. Rollout          bar/21 6/5       Eq.:  -0.025
      0.481 0.119 0.004 - 0.519 0.278 0.089 CL  -0.025 CF  -0.025
     [0.004 0.004 0.007 - 0.004 0.007 0.006 CL   0.009 CF   0.009]

   2. Rollout          bar/21 24/23    Eq.:  -0.053 ( -0.028)
      0.467 0.121 0.003 - 0.533 0.259 0.059 CL  -0.053 CF  -0.053
     [0.003 0.003 0.001 - 0.003 0.007 0.006 CL   0.007 CF   0.007]

   3. Rollout          bar/20            Eq.:  -0.058 ( -0.034)
      0.464 0.112 0.005 - 0.536 0.264 0.059 CL  -0.058 CF  -0.058
     [0.004 0.003 0.002 - 0.004 0.008 0.006 CL   0.007 CF   0.007]

   4. Rollout          bar/24 6/2*     Eq.:  -0.063 ( -0.039)
      0.460 0.135 0.005 - 0.540 0.271 0.084 CL  -0.063 CF  -0.063
     [0.005 0.003 0.001 - 0.005 0.009 0.008 CL   0.011 CF   0.011]

   5. Rollout          bar/24 13/9     Eq.:  -0.079 ( -0.055)
      0.452 0.127 0.006 - 0.548 0.277 0.090 CL  -0.079 CF  -0.079
     [0.004 0.003 0.001 - 0.004 0.007 0.006 CL   0.007 CF   0.007]

   6. Rollout          bar/21 7/6       Eq.:  -0.092 ( -0.068)
      0.448 0.116 0.002 - 0.552 0.252 0.044 CL  -0.092 CF  -0.092
     [0.003 0.003 0.001 - 0.003 0.007 0.004 CL   0.006 CF   0.006]

   7. Rollout          bar/21 8/7       Eq.:  -0.095 ( -0.070)
      0.445 0.115 0.005 - 0.555 0.270 0.072 CL  -0.095 CF  -0.095
     [0.004 0.003 0.001 - 0.004 0.008 0.007 CL   0.008 CF   0.008]
       Full cubeful rollout with var.redn.
       1296 games, Mersenne Twister dice gen. with seed 890946208 and quasi-random dice
       Play: world class 2-ply cubeful prune [world class]
       keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
       Skip pruning for 1-ply moves.
       Cube: 2-ply cubeful prune [world class]

[close]

[stiefnu]

Surprise!  The top two plays according to 0-ply rollout did not get any votes at all.

Ahem!  I think you'll find that Bar/21, 6/5 did get one vote, mine. 
Just wish I could find such brilliant moves more often - and then find the words to say why I thought it was best.  "it seems like a good idea" would not have come up to the standard of erudition continually demonstrated by the Herd's top brains.  Thanks guys, this is a great way to learn how to think about the game.

The "bot" play is still on top

Oy, watch who you're calling a bot!

[Zorba]

Spoiler

XG's rollout has the same order of plays, but the top 4 are all very close.

    1. Rollout¹    Bar/21 6/5                   eq:-0,052
      Player  : 46,61% (G:11,86% B:0,58%)
      Opponent: 53,39% (G:24,86% B:5,61%)
      Confidence: ± 0,007 (-0,059<E<-0,045)
      Duration: 4 hours 04 minutes

    2. Rollout¹    Bar/21 24/23                 eq:-0,055 (-0,004)
      Player  : 46,46% (G:12,63% B:0,48%)
      Opponent: 53,54% (G:21,69% B:3,19%)
      Confidence: ± 0,006 (-0,061<E<-0,049)
      Duration: 3 hours 31 minutes

    3. Rollout¹    Bar/20                       eq:-0,058 (-0,006)
      Player  : 46,37% (G:12,13% B:0,46%)
      Opponent: 53,63% (G:21,32% B:3,75%)
      Confidence: ± 0,007 (-0,065<E<-0,051)
      Duration: 3 hours 37 minutes

    4. Rollout¹    Bar/24 6/2*                  eq:-0,064 (-0,012)
      Player  : 45,98% (G:13,96% B:0,48%)
      Opponent: 54,02% (G:23,02% B:4,63%)
      Confidence: ± 0,007 (-0,071<E<-0,057)
      Duration: 3 hours 31 minutes

    5. Rollout²    Bar/21 8/7                   eq:-0,083 (-0,031)
      Player  : 45,20% (G:11,88% B:0,28%)
      Opponent: 54,80% (G:21,90% B:3,40%)
      Confidence: ± 0,023 (-0,106<E<-0,060)
      Duration: 16 minutes 28 seconds

    6. Rollout²    Bar/21 7/6                   eq:-0,105 (-0,054)
      Player  : 43,90% (G:12,23% B:0,62%)
      Opponent: 56,10% (G:17,93% B:3,13%)
      Confidence: ± 0,017 (-0,122<E<-0,088)
      Duration: 14 minutes 45 seconds

    7. Rollout³    Bar/24 13/9                  eq:-0,125 (-0,073)
      Player  : 42,94% (G:11,95% B:0,62%)
      Opponent: 57,06% (G:22,36% B:4,15%)
      Confidence: ± 0,025 (-0,150<E<-0,100)
      Duration: 8 minutes 10 seconds

¹ 2592 Games rolled with Variance Reduction.
  Cubeless
  Moves: 3 ply Large

² 180 Games rolled with Variance Reduction.
  Cubeless
  Moves: 3 ply Large

³ 108 Games rolled with Variance Reduction.
  Cubeless
  Moves: 3 ply Large

eXtreme Gammon Version: 1.20, MET: Rockwell-Kazaross

[close]