Are you less awful than me? - Backgammon Forum - Discuss Backgammon Strategy, Clubs and Books

Two Plus Two Forums Sports and Games Chess and Other Board Games Backgammon Forum hosted by Bill Robertie.

Are you less awful than me?

Post Reply Subscribe

...

06-26-2009 , 02:13 AM

ChrisV

Carpal \'Tunnel

Join Date: Jul 2004 Posts: 40,336

I just played a few cash games with a friend and GNU rated me Awful! on both cubing and checker play, which is not a common occurrence. I thought it would be fun to embarrass myself by posting some of the moves and cube decisions I got badly wrong and see if everyone else can do any better. (Note: As I said, these are bad moves I made, so they shouldn't be too tricky to work out for decent players. It's always easier when you know it's a "problem" rather than a standard move of course...)

1. Cash game, Red owns 2-cube, Black to play 4-4

2. Cash game, center cube, Black on roll, cube action?

3. Cash game, center cube, Black to play 1-1

4. Cash game, center cube, Black to play 3-1

Quote

06-26-2009 , 03:33 AM

dsaxton

adept

Join Date: Oct 2005 Posts: 709

1.) I'd play Bar/4 13/5 4/2*, trying to blitz red's last back man and duplicating 1's and 2's.

2.) I'd say double / take. You're favored to enter and hit, have vastly superior structure and some sequences will be devastating for red. Regardless, red obviously still has takeable winning chances.

3.) I'd play 8/7(2) and either 5/4*(2) or 6/4*. Switching is safer but breaks your prime and doesn't start a new point. Keeping the 5 point and hitting loose risks a hit from the bar, but better maintains your priming potential and has the possibility of leading to a very strong position when missed. Overall, I think hitting loose is slightly better.

4.) I'd make the 5 point with 8/5 6/5.

Last edited by dsaxton; 06-26-2009 at 03:38 AM.

Quote

06-26-2009 , 03:50 AM

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

1. i like bar/21 6/2* 13/5

i mean i think it has to be that or bar/21 6/2* 21/13 but if u leave the checker on the 21 pt u have 2 more red blots to hit. plus i guess u could potentially anchor on the 21 pt if red enters and hits next turn

2. i'm pretty torn but im gonna go with double/take because black has the better board plus the threat of hitting the blot on the 22 pt and then if red dances, there are 2 more blots on the 17 and 15 pts.

3. i think id always do 6/4* and 8/7(2), altho that seems like the obvious choice

4. maybe 8/7* 24/21 that might leave too many return shots, but the safe option of 8/7* 7/4 seems kinda weak/passive.

Quote

06-26-2009 , 09:29 PM

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

if anyone cares, here are 628 game cubeless 2ply rollouts for each spot

Spoiler:

Quote

06-26-2009 , 10:51 PM

dsaxton

adept

Join Date: Oct 2005 Posts: 709

Really? I have a hard time believing that the difference between switching and hitting loose would be so enormous in situation #3, or that making the 5 point would somehow be a big blunder. How reliable are these estimates?

Quote

06-26-2009 , 11:41 PM

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

these are the rollout settings that robertie recommended, so as far as i know they should be fairly accurate, but i could be wrong.

Quote

06-27-2009 , 09:34 PM

ChrisV

Carpal \'Tunnel

Join Date: Jul 2004 Posts: 40,336

Here are my results, based on cubeful 2-ply GNU rollouts:

Spoiler:

1)
Correct answer: bar/21 13/5 6/2*. However, bar/21 6/2*(3) is very close to as good.
What I actually did: bar/13 6/2* (-0.140)
Diagnosis: This was a case of moving without thinking. I was so happy to get my back checker the hell out of dodge that I simply moved it out of harm's way then figured out what to do with the rest of the 4s, whereupon hitting loose was pretty obvious.

NOTE: The above is what I wrote originally. I think it's pretty weird that GNU and Snowie rate these moves so differently, expecially since they're in complete agreement over all the other problems. I wonder if one of us did something wrong. Also bar/21 13/5 6/2* looks clearly correct to me in hindsight.

2)
Correct answer: Double, take.
What I actually did: No double (-0.157)
Diagnosis: I never even thought about doubling. I have a checker on the bar, one back on the 24 point, only a three point board. and I am a little behind in the race. My position doesn't seem that amazing. However, looking at things from Red's point of view, I realise I would be hating life. He has a lot of problems. The 6 point is badly overstacked, he has no anchor, and he is pretty loose. The blot in his home board is reasonably likely to be hit this roll and if it is he could quickly get into a lot of trouble.

3)
Correct answer: 8/7(2) 5/4*(2)
What I actually did: 8/7(2) 6/4* (-0.270)
Diagnosis: Never break a made 5 point, right? Golden point, amiright? It seemed obvious to me that I should make the bar point, and after that, I never even considered moving the 5. With one of his checkers on the bar and one on the 1 point, and the race about even, I was thinking about creating a prime to trap his back checkers. A little risk seemed worth achieving this aim. However this is not the dominant theme of the position. His board is quite strong and I should be thinking about not getting hit and springing my remaining back checker. Think about how awesome Black's position would look if that back checker were on the midpoint and it should be obvious that the right idea is to stay safe and buy some time to extract the back checker. This theme is so dominant that all the moves with 5/4*(2) are better than my move. 6/4*/3 24/23 is also better, as it starts trying to spring the back checker while buying some time.

4)
Correct answer: 24/21 8/7*
What I actually did: 8/5 6/5 (-0.077)
Diagnosis: I don't think GNU understands this game. The 5 point is the Golden Point. I read it in a book.
The problem with making the 5 here is that Red is threatening to make my bar point. My position is going to be very stacked after 8/5 6/5 and having him own my bar will be a disaster. With him having even less of a board than I do, I should be entering into a blot-hitting war.

Last edited by ChrisV; 06-27-2009 at 09:40 PM.

Quote

06-28-2009 , 12:27 AM

dsaxton

adept

Join Date: Oct 2005 Posts: 709

These are interesting problems.

In position #3, my first impression was to switch, and then I changed my mind after looking at it for a while. That's one of the difficult things about backgammon; sometimes it seems you can trust your instincts, and sometimes you need to ignore them and try to think through the problem more thoroughly. I believe why the computer likes switching is because hitting loose not only risks losing ground in the race, but because it risks losing a tempo in a highly volatile position, which can be quite costly. Also, the position after 6/4* seems a bit more brittle than the alternative.

I think what's going on in position #4 is that red has excellent diversification and is preparing to make multiple improvements in his position, but putting him on the bar prevents him from doing this. If we make the 5-point, we allow red to do as he pleases. And as ChrisV said, inducing action when you have a stronger board is always good.

Quote

06-28-2009 , 12:57 AM

ChrisV

Carpal \'Tunnel

Join Date: Jul 2004 Posts: 40,336

I have to say that it still surprises me how awful the loose hit is compared to switching in #3. -0.270 is a lot.

In #4, saying that red has a lot of good rolls is more accurate than what I said (that he is threatening to make the bar). I do think him making the bar is a pretty serious threat though. I think his loose checker in the outfield is also important, since in a blot-hitting war that is liable to find itself a target pretty quickly.

Quote

06-28-2009 , 01:25 AM

#10

WesleyC

enthusiast

Join Date: Feb 2006 Posts: 72

Maybe slightly less awful....

1. Bar/21, 13/5, 6/2*
Trying to close out red. The risk of leaving a man on the 21 isn't so big because red needs 1's and 2's to roll on. He also offers the chance to clean up all the red blots which improves your chance of gammon. How did red get into such a horrible position in the first place? He must truly be awful!

2. Double.

Although i tend to miss doubles from the bar at the table. The 1/9 chance of not rolling on out is countered by hits on any 3, 52, 22, 12 and 11 (almost 50%). Add to that a better board and the slew of red blots ready to be cleaned up (who is this chump playing red?) and it seems a pretty clear double.

3. 8/7 (2), 5/4*(2)

The problem here is you've got too many great 1's. 24/23 is tempting, but your back man isn't in such a hurry to escape that you would give up the 2 other great options.

4. 8/7*, 24/21

2nd choice is 8/7*, 13/10. Not so much a fan of building the 5 because you dont have any builders to follow up, it leaves too many hits and isn't going to be much use once reds back men have already escaped.

ps. please don't post any of my awful moves :P

Quote

06-28-2009 , 01:48 AM

#11

WesleyC

enthusiast

Join Date: Feb 2006 Posts: 72

I just read the analysis - its certainly unusual that it agrees with my choices of moves :P

I think one theme that comes up in P3 and P4 is that axioms like 'always build the 5' are often taken too far.

In problem 3, I'm surprised that people are even considering 8/7(2) 6/4* - i considered and rejected it almost right away. In this situation the 5 point isn't significantly better than the 4 and hanging onto it gives up so much. Loose hitting leaves red 16/36 numbers that retaliate immediately (plus 62 which hits your only builder) and even if they miss, you've only got one builder to cover it on your next roll.

Quote

06-28-2009 , 01:54 AM

#12

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

i'm not sure either about the discrepancy between snowie and gnu, but i redid the rollouts at 3ply for 628 games (cubeless still) and got the same order

Spoiler:

Quote

06-28-2009 , 01:54 AM

#13

gigabet77

centurion

Join Date: Jun 2008 Posts: 192

hai dere

Quote

06-28-2009 , 02:33 AM

#14

ChrisV

Carpal \'Tunnel

Join Date: Jul 2004 Posts: 40,336

djk123: I'm not familiar with the details of how rollouts work, but it must be because mine is cubeful and yours is cubeless, no? bar/21 13/5 6/2* is a somewhat riskier move, so maybe it's Red being able to redouble at some point that makes the difference?

Quote

06-28-2009 , 03:45 AM

#15

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

Quote:

Originally Posted by ChrisV

well, let me first say that i am in no way an expert on rollouts either. i've only even been playing backgammon for a few months. but i know that robertie mentioned how snowie cubeless rollouts have a variance-reducing element while cubeful rollouts do not, so 628 cubeful rollouts might not be a big enough sample. for example in the thread about whether or not to split with 63 after your opponent gets an early 55, my 324 game 3 ply cubefull rollouts produced results contrary to contemporary theory/logic/reasoning. but when i did 3888 cubeful 3ply rollouts, they produced results consistent with the current theory. so i guess the 324 sample was just a product of variance?

but anyways im curious so im going to do some cubeful 3 ply rollouts, which will take a while

edit: from the snowie help file:

If you performed a cubeless rollout then a cubeful equity is estimated from the cubeless rollout result and is displayed in the corresponding panels (either Cube or Moves).

so cubeless rollouts don't completely ignore the cube. i guess it's just a question of how accurate the estimate is..

Last edited by djk123; 06-28-2009 at 03:50 AM.

Quote

06-28-2009 , 06:31 PM

#16

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

3888 3ply (both checker decisions and cube decisions)cubeful rollouts. i did them in separate snowie instantiations to save time

Spoiler:

Quote

06-28-2009 , 07:55 PM

#17

dsaxton

adept

Join Date: Oct 2005 Posts: 709

I can see why that might be true. With the somewhat safer play, black should be able to cash the game more frequently, whereas if the game is played to the end, moving the builder to the 5 point probably has better long-term prospects. I didn't even consider the cube at first, but it's probably a main theme of the position.

Quote

06-28-2009 , 08:39 PM

#18

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

Spoiler:

Quote

06-28-2009 , 11:03 PM

#19

djk123

Pooh-Bah

Join Date: Jun 2005 Posts: 5,069

Spoiler:

Quote

Post Reply Subscribe

...