obvious is not always right - Page 12 - Backgammon Forum - Discuss Backgammon Strategy, Clubs and Books

Thanks, Yogiman, for the nice variants. I would have had trouble discriminating between them. Here is another showing that location matters. When White has only 3 blots, but all are within 8 pips of Black's anchor, then it is time for the bananas to split. After a White dance, Black will have 21 rolls that hit one of the 3. That number could be higher, but I am assuming Black will remake his board when he rolls 52.

White - Pips 89

Black - Pips 135
Black to Play 3-2

Code:

XGID=-BaBBBC-------aaa--b-dCcbA:2:-1:1:32:0:0:0:0:10

X:Player 1   O:Player 2
Score is X:0 O:0. Unlimited Game
Pip count  X: 135  O: 89 X-O: 0-0
Cube: 4, O own cube
X to play 32

    1. Rollout¹    Bar/22 4/2*            eq:-0.123
      Player:   50.25% (G:24.68% B:3.75%)
      Opponent: 49.75% (G:27.35% B:0.63%)
      Confidence: ±0.008 (-0.131..-0.115) - [100.0%]

    2. Rollout¹    Bar/22 6/4             eq:-0.244 (-0.121)
      Player:   47.85% (G:11.16% B:0.46%)
      Opponent: 52.15% (G:14.21% B:0.18%)
      Confidence: ±0.009 (-0.253..-0.236) - [0.0%]

    3. Rollout¹    Bar/20                 eq:-0.250 (-0.126)
      Player:   48.79% (G:11.72% B:0.67%)
      Opponent: 51.21% (G:19.83% B:0.27%)
      Confidence: ±0.008 (-0.258..-0.241) - [0.0%]

¹  1296 Games rolled with Variance Reduction.
   Dice Seed: 63453586
   Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.10

Mike

Quote

02-26-2015 , 04:08 AM

#277

yogiman

adept

Join Date: Jun 2012 Posts: 784

Position ID: 3gYwAhfY7gZiAA Match ID: cIkKAAAAAAAE

White - Pips 153

Black - Pips 121
Black to Play 5-2
Created with www.BGdiagram.com

Quote

02-26-2015 , 06:01 AM

#278

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Regarding cube position, Yogiman set this up as a money game without the Jacoby Rule. Black did not miss a double last turn. He is too good to double.

You can tell this is a money game because there is no mention of match score in the caption of the board diagram. Once you know that, you can determine that there is no Jacoby Rule by the “E” at the end of the GnuBg ID. When Jacoby is in effect (and the score is 0-0), the final letter is always an “A”. Easier, of course, is to paste the GnuBg ID into your bot, and check the status of Jacoby there.

For future reference, most of the positions in this thread have been set up without Jacoby. That’s the way I like to play in practice myself. It gives me a chance to rehearse the “am-I-too-good-to-double” decisions that are suppressed when Jacoby prevents you from winning a gammon on an unturned cube.

I have written up my play, but I will wait to post it until later.

Mike

Last edited by Taper_Mike; 02-26-2015 at 06:07 AM.

Quote

02-26-2015 , 09:47 AM

#279

Robertie

old hand

Join Date: May 2004 Posts: 1,954

Quote:

Originally Posted by Taper_Mike

Easier, of course, is to paste the GnuBg ID into your bot, and check the status of Jacoby there.
Mike

Even easier and clearer is for the original poster to say below the diagram "This is a money game played without the Jacoby Rule."

Quote

02-26-2015 , 11:13 AM

#280

yogiman

adept

Join Date: Jun 2012 Posts: 784

I suppose that viewers don't start with the last problem, and if they have been following this thread from the beginning, they have been informed by post#42 that "this is a money game played without the Jacoby Rule."

To make it utterly simplistic (bordering to laziness) the following conditions hold for all coming problems by yogiman:

It concerns a money game played without the Jacoby Rule.

Truncated cubeful rollout (depth 10) with var.redn.
147 games, Mersenne Twister dice gen. with seed 768213351 and quasi-random dice
Stop when std.errs. are small enough: ratio 0,1 (min. 144 games)
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]

Quote

02-26-2015 , 01:54 PM

#281

uberkuber

veteran

Join Date: Jul 2009 Posts: 3,043

14/9* 17/15.

Trying to extend the prime and bringing a builder in direct range.

Quote

02-26-2015 , 02:16 PM

#282

Robertie

old hand

Join Date: May 2004 Posts: 1,954

What's the point of posting money game problems without the Jacoby Rule when the Jacoby Rule is universal in money games?

Quote

02-26-2015 , 03:24 PM

#283

yogiman

adept

Join Date: Jun 2012 Posts: 784

Dear moderator,

Sorry that you tempt me to defend myself. I don't need to tell you that it is useless in anyway to present money game problems when no money is at stake.

Every linguist can tell you what you already know, namely that language is a natural phenomenon, and whether we are talking about inner nature or outer nature, it has generally a bent and as a consequence it doesn't always follow the ways of logic.

Last edited by yogiman; 02-26-2015 at 03:47 PM.

Quote

02-26-2015 , 04:22 PM

#284

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Quote:

Originally Posted by Robertie

What's the point of posting money game problems without the Jacoby Rule when the Jacoby Rule is universal in money games?

Think of it as practice for match play. An unlimited (a.k.a., money) game with neither Jacoby nor beavers is similar to a match where the score is tied many points away from victory. Mathematically, it is identical to match play where both players are an infinite number of points away from winning.

This is a good way to practice because it lets you experience the "too-good" decisions that you will face in match play, and, at the same time, eliminate any influence of score. The same is true for other checker plays and cube decisions. No Jacoby/no beavers lets you practice match play without any bias due to score.

I know this is not news for you, Bill. I make the explanation just in case there are a few other readers who are not yet aware of this.

Mike

Quote

02-26-2015 , 06:07 PM

#285

Robertie

old hand

Join Date: May 2004 Posts: 1,954

Actually, it was news to me.

Why not just practice for matches by playing matches? You'll get plenty of bias-free double/too-good decisions early in matches.

Quote

02-26-2015 , 06:55 PM

#286

uberkuber

veteran

Join Date: Jul 2009 Posts: 3,043

Quote:

Originally Posted by Robertie

Actually, it was news to me.

Why not just practice for matches by playing matches? You'll get plenty of bias-free double/too-good decisions early in matches.

The way I see it, with the money game, no Jacoby rule, I approach it as the first game of an 11-pt match for instance.

I think Taper_Mike might have hinted that setting to yogiman since it might have been easier to set up (compared to entering a match length) in GnuBG or XG.

Quote

02-26-2015 , 09:03 PM

#287

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Quote:

Originally Posted by Robertie

Why not just practice for matches by playing matches? You'll get plenty of bias-free double/too-good decisions early in matches.

If the intention is to practice many-away vs. many-away, you cannot do better than an unlimited game with neither Jacoby nor beavers. Those settings completely eliminate any influence of score.

The shorter 7- and 9-point matches that we all now so often play are probably not long enough to do the same thing. Depending on the position, a match tied at 11-away may not do the job either. A gammon on a recube is an 8-point swing, so whatever part of your sample space includes those is going be heavily impacted by the MET.

Anything that is not an exact tie can also cause problems. I have been surprised at the frequency with which I find that a checker play or cube decision at 10-away, 11-away is not the same as it is when tied at 11-away. I don't have any links for you, but I have seen many of these in posts at BgOnline.

I turned off Jacoby (and beavers) for my unlimited game practice sessions and rollouts several years ago. It has worked out very well for me.

Mike

Quote

02-27-2015 , 03:55 PM

#288

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Position ID: 3gYwAhfY7gZiAA Match ID: cIkKAAAAAAAE

White - Pips 153

Black - Pips 121
Black to Play 5-2

Although White might fall into a backgame if he makes a second anchor, he is not playing one yet. Not only is he short one anchor, he is also short on timing. Even if Black hits on the 9pt, White will trail by only 48 pips after this turn. Moreover, it is very hard to play a backgame after you bust your board. If White is lucky enough to get a late hit, it won’t help him if the checkers he needs to build a block are buried on the low points of his board. For all these reasons, Black does not need to worry that he helps White by hitting him.

I like hitting on the 9pt, and then leaving a blot there. The 2 can be played 7/5. Leaving a blot on the 9pt accomplishes two things. First, Black might make that point, completing his prime. Second, if White jumps out, the delay caused by being hit will help Black maintain his timing. Whether or not Black can make his 9pt, if White does not get another checker into the outfield soon, he will end up burying the two checkers now on his midpoint (if they don’t get hit first).

Mike

Quote

02-27-2015 , 04:45 PM

#289

yogiman

adept

Join Date: Jun 2012 Posts: 784

Code:

    1. 14/9* 7/5                    Eq.:  +1,345 
       0,878 0,480 0,025 - 0,122 0,012 0,000 CL  +1,248 CF  +1,345 
     
    2. 17/15 14/9*                  Eq.:  +1,331 ( -0,014) 
       0,875 0,475 0,027 - 0,125 0,016 0,000 CL  +1,236 CF  +1,331 
      
    3. 7/2* 6/4                     Eq.:  +1,245 ( -0,100) 
       0,854 0,406 0,018 - 0,146 0,022 0,000 CL  +1,110 CF  +1,245

Maybe some participants were fooled by the amount of text following the problem post. But again, obvious is just obvious here, and please blame the dice. Alter white's pipcount, and the timing makes it proper to play 7/2*.

White - Pips 191

Black - Pips 121
Black to Play 5-2
Created with www.BGdiagram.com

Code:

    1. 7/2* 6/4                     Eq.:  +1,006 
       0,689 0,384 0,018 - 0,311 0,051 0,002 CL  +0,726 CF  +1,006 
      
    2. 14/9* 7/5                    Eq.:  +0,964 ( -0,042) 
       0,682 0,440 0,024 - 0,318 0,042 0,001 CL  +0,786 CF  +0,964

Moving the black and white checker 1 pip shows that with the stronger homeboard hitting becomes profitable for white, sothat the equity difference gets reversed:

White - Pips 190

Black - Pips 122
Black to Play 5-2
Created with www.BGdiagram.com

Code:

    1. 17/15 7/2*                   Eq.:  +1,074 
       0,719 0,404 0,018 - 0,281 0,043 0,001 CL  +0,816 CF  +1,074 
     
    2. 15/10* 7/5                   Eq.:  +0,975 ( -0,099) 
       0,685 0,457 0,027 - 0,315 0,040 0,001 CL  +0,811 CF  +0,975

Quote

02-28-2015 , 06:59 AM

#290

yogiman

adept

Join Date: Jun 2012 Posts: 784

Position ID: 2nyQATCwbysAAw Match ID: cIkGAAAAAAAE

White - Pips 131

Black - Pips 129
Black to Play 5-1
Created with www.BGdiagram.com

Quote

02-28-2015 , 09:53 AM

#291

Fllecha

journeyman

Join Date: Mar 2010 Posts: 321

Just hit! Cant see alternatives...

Quote

02-28-2015 , 04:16 PM

#292

Robertie

old hand

Join Date: May 2004 Posts: 1,954

This is the sort of position where Black doesn't want to get in a blot-hitting contest. Hitting leaves Black with a weaker board plus 4 blots. Sitting on the position with something like 10/5 6/5 or 10/5 9/8 leaves Black ahead in the race with no blots and plenty of ways to make 5 in a row or 5 out of 6 next turn.

Now for an even better question: Why is the cube in the middle here? Before rolling, Black has a monster double -- ahead in the race with many rolls that make the 4-point or 3-point. Meanwhile White has no counterplay whatever. Double and clear pass, I would think.

Quote

02-28-2015 , 05:21 PM

#293

yogiman

adept

Join Date: Jun 2012 Posts: 784

Code:

Cube analysis
Rollout cubeless equity  +0,719

Cubeful equities:
1. Double, pass         +1,000
2. Double, take         +1,312  ( +0,312)
3. No double            +0,958  ( -0,042)
Proper cube action: Double, pass

This has possibly been my fault, instead of being a cube error. The tutor signalled an error, and I was forgetting about the cube. Observing the equity difference, by hitting black would in most cases spoil his finishing cube opportunity next turn.

Code:

    1. 10/5 9/8                     Eq.:  +0,941
       0,733 0,187 0,010 - 0,267 0,052 0,001 CL  +0,609 CF  +0,941
      
    2. 10/5 6/5                     Eq.:  +0,896 ( -0,045)
       0,739 0,191 0,010 - 0,261 0,052 0,001 CL  +0,626 CF  +0,896
      
    3. 20/14*                       Eq.:  +0,696 ( -0,244)
       0,696 0,239 0,024 - 0,304 0,109 0,010 CL  +0,536 CF  +0,696

The result when white has the cube:

Code:

    1. 10/5 9/8                     Eq.:  +0,487
       0,727 0,182 0,009 - 0,273 0,056 0,002 CL  +0,587 CF  +0,487
      
    2. 20/14*                       Eq.:  +0,429 ( -0,058)
       0,697 0,237 0,024 - 0,303 0,108 0,010 CL  +0,537 CF  +0,429

Kudos are not going only to Robertie. Fllecha didn't know about my error, which proofs that truth and beauty can go hand in hand.

Quote

03-02-2015 , 03:59 AM

#294

yogiman

adept

Join Date: Jun 2012 Posts: 784

Position ID: 3Z4DBgDbdgMAAw
Match ID: cAkKAAAAAAAE

White - Pips 73

Black - Pips 87
Black to Play 4-2
Created with www.BGdiagram.com

Quote

03-02-2015 , 08:10 AM

#295

uberkuber

veteran

Join Date: Jul 2009 Posts: 3,043

5/1 4/2.

Quote

03-02-2015 , 08:22 AM

#296

bleep69

grinder

Join Date: Sep 2008 Posts: 483

The lesson drummed in to me countless times by XG is never be the first one to run if you can avoid it. I'm not sure if the exposed blot one white's 1 point changes things much here; I can't see how it does - white will cover with 5, 3 or 1, and if we run we haven't duplicated any of those. We are way behind as well.

6/4 6/2, despite it being obvious.

Quote

03-02-2015 , 12:20 PM

#297

Fllecha

journeyman

Join Date: Mar 2010 Posts: 321

The race is relatively close in this holding game and white is bout to break the midpoint so i dont run Now. Simply i dump 6/4 6/2 and i wait for a likely (deadly) shot

Quote

03-02-2015 , 01:42 PM

#298

Robertie

old hand

Join Date: May 2004 Posts: 1,954

Staying back here isn't all that appealing.

> The race is close (Black only trails by 8 after the roll) so staying on the 20-point and breaking the board sacrifices a lot of race equity in all the variations where no one gets hit (which is most variations).

> White will have to leave a shot on all his sixes except 6-6. If you stay back and break your board, you'll have an 8-number shot with only a 5-point board. If you come out and White rolls a bad 6, you'll have a direct plus an indirect and a closed board.

Those two considerations make coming out look pretty appealing. Now there's another choice: 20/12 or 20/16 5/3?
To choose between them, let's look at both the downside risk and the upside gains.

The downside risk is the chance that White hits and covers. If we play 20/14, I count 10 hit-and-cover numbers: 2-1, 2-3, 2-5, 1-4, 1-1, and 4-4. If we play 20/16 5/3, I count 7 hit-and-covers: 4-1, 4-3, 4-5, and 4-4. The downside risk favors stopping on the 16-point.

The upside risk is the chance that we hit after White throws a bad 6. Suppose his next roll is 6-5. If we played 20/14, we now have 11 twos plus 1-1, 5-3 and 4-4 to hit, a total of 15 numbers. If we played 20/16 5/3, we have all fours plus 6-2, 5-3, 1-1, 2-2, and 3-1, a total of 19 hits. So the upside gain also favors stopping on the 16-point.

My play: 20/16 5/3. Plus, it wows the spectators!

Quote

03-02-2015 , 02:40 PM

#299

Taper_Mike

adept

Join Date: Sep 2009 Posts: 1,030

Grunch. I have not read the answers of others.

In a holding game, he who breaks anchor first generally loses. Despite the blot in White’s board, that is probably true here. Although is may be obvious, then, I will break the board rather than the anchor.

What is not obvious is whether to leave a blot on the 6pt or not.

Except for 54, White can safely play any roll not containing a 6. Most 6s, allow him to play one checker from the midpoint to a safe landing spot in his board. 62 is his horror roll. That leaves a direct shot. The shots Black gets on 65, 64, 63, and 61 are all fly shots. If Black misses, however, White has repeaters. There are many rolls that will not clear a blot on White’s midpoint, 19 to be exact.

If Black hits a fly shot next turn, he will wish he did not have a blot in his board. In the much more likely circumstance that he does not hit, he will be glad he can play a 5 without being forced off his anchor. Playing 6/4 5/1 now, means Black can handle 53 and smaller without cracking further.

I’ll try that.

Mike

Quote

03-03-2015 , 01:24 PM

#300

yogiman

adept

Join Date: Jun 2012 Posts: 784

Thanks to my forum mates for blundering, because that gives me the feeling I am not the only one who shows at times an undesirable lack of infallibility. However, the heroism of Taper_Mike and Uberkuber is suspicious, as a problem with the same theme can be found in the -problem of the week- sequence. I am glad that Robertie took the trouble to give a detailed analysis, because that saves me a lot of tedious puzzling.

Code:

    1. 20/16 5/3                    Eq.:  -0,326 
       0,386 0,053 0,002 - 0,614 0,056 0,001 CL  -0,230 CF  -0,326 
     
    2. 20/14                        Eq.:  -0,379 ( -0,054) 
       0,380 0,055 0,002 - 0,620 0,046 0,001 CL  -0,229 CF  -0,379 
   
    3. 6/4 5/1                      Eq.:  -0,575 ( -0,250) 
       0,321 0,014 0,000 - 0,679 0,024 0,000 CL  -0,367 CF  -0,575 
     
    4. 20/18 5/1                    Eq.:  -0,655 ( -0,329) 
       0,319 0,040 0,001 - 0,681 0,064 0,001 CL  -0,386 CF  -0,655

Some might think that I try to trick you with white's blot in order to tempt you into bold play, but that's not the case. Though without that blot bold play would have been a big blunder:

Code:

    1. 6/4 6/2                      Eq.:  -0,624
       0,301 0,011 0,000 - 0,699 0,024 0,000 CL  -0,411 CF  -0,624
     
    2. 20/14                        Eq.:  -0,839 ( -0,215)
       0,293 0,038 0,002 - 0,707 0,079 0,002 CL  -0,455 CF  -0,839
      
    3. 20/16 5/3                    Eq.:  -0,954 ( -0,330)
       0,270 0,041 0,002 - 0,730 0,095 0,002 CL  -0,514 CF  -0,954

I could add that in the case of 43 it is better to do 20/13 instead of 20/16 because of the duplication of 1's.

Code:

    1. 20/13                        Eq.:  -0,289 
       0,401 0,035 0,001 - 0,599 0,030 0,000 CL  -0,192 CF  -0,289 
      
    2. 20/16 4/1                    Eq.:  -0,325 ( -0,036)

Quote

Page 12 of 13

First

2 7 8 9 10 11 12 13

Last

Post Reply Subscribe

...

Page 12 of 13

First

2 7 8 9 10 11 12 13

Last