Open Side Menu Go to the Top
Register
Black to play 3-2 Black to play 3-2

12-27-2011 , 09:44 PM
http://img526.imageshack.us/img526/8...ghdecision.jpg

This is a single point game. I ended up playing 13-10, 6-4.

I was hoping to get some opinions on this position. It was a tough move to make, but I was confident my move was correct.

When looking at what the computer says it appears to be reasonably close between a few moves.

I hope to post what the computers (Gnu and XG) say about the position both before and after I rolled out the move I made.

Thanks
Black to play 3-2 Quote
12-28-2011 , 12:48 AM
Maybe I'm just awful, but 23/18 seems way the best as it sets you up to play the rest of the game cleanly.
Black to play 3-2 Quote
12-28-2011 , 01:40 AM
23/18 seems pretty clear to me here, gives you a nice outpost and blocking position, keeps you connected and brings a guy out. Seems too early to go for a backgame/powerslot thing.
Black to play 3-2 Quote
12-28-2011 , 03:07 PM
I'd go with 23/18 too.
Black to play 3-2 Quote
12-29-2011 , 05:31 PM
My play is 23/18, as well.

Another interesting problem arises when you reverse the turn. Suppose your opponent had to play this 32. How should he move?

Obviously, he would hit on the bar point with the 2. But what about the 3?

Spoiler:
I would hit twice. 9/7* 6/3*. As Red, my game plan would be to go all out to make the bar point, and then force Black to crack.
Black to play 3-2 Quote
12-29-2011 , 07:57 PM
Quote:
Originally Posted by Taper_Mike
My play is 23/18, as well.

Another interesting problem arises when you reverse the turn. Suppose your opponent had to play this 32. How should he move?

Obviously, he would hit on the bar point with the 2. But what about the 3?
I would have to agree, it looks like a double hit...

This was an interesting move to me for a number of reasons.
First of all I would like to state that I was playing Nackgammon, (making me more inclined to play a backgame) and that I did not have a pip count (and did not do a count on my own). These both could have led me to make my move over something like 23/18, but even after the fact I believe my move is the best.

The only negative to my move seems to be the computer making his 7 point and having a 5 point board which is a possibility, but even so it is not that bad of a situation... Thinking about it some might say being double hit is also a problem, but it really isn't in most scenarios.

23/18 just seems ugly to me. It puts us in a bad situation with our timing, allows him to stack on his 2 point if necessary, and when I ask myself what roll will be good for us next turn I am at a relative blank.

18/15, 6/4 is what XG considers the best and GNU considers second best but I think it is just okay. If it is possible I would like someone to explain to me why the computer likes this move so much?

I am going to add two screen shots, both of which are taken after rollouts of the position. On GNU the percentages didn't change much after the rollouts, and the XG percentage of 13/10, 6/4 went from -0.305 to -0.292.

http://img834.imageshack.us/img834/1...rolloutgnu.jpg

http://desmond.imageshack.us/Himg97/...jpg&res=medium

There are major differences in what the two programs thought of the situation. I am relatively new to playing backgammon against the computer and have never checked the percentages on two programs for the same position, so I don't know if this happens often, but considering that they are supposed to be two of the top in all of backgammon it came as a surprise. I would be inclined to believe XG since what I have read about the two has led me to believe that it is superior to GNU, and also because it agrees with me more than GNU does .
Black to play 3-2 Quote
12-29-2011 , 09:16 PM
Quote:
Originally Posted by nonotyet
I hope to post what the computers (Gnu and XG) say about the position both before and after I rolled out the move I made.
You shouldn't just roll out one play and then compare it to the evaluations of the other moves (or compare different degrees of evaluations).

Apples and oranges (and bananas).

So you need to roll out all reasonable candidates.
Black to play 3-2 Quote
12-29-2011 , 10:26 PM
Quote:
Originally Posted by mute
You shouldn't just roll out one play and then compare it to the evaluations of the other moves (or compare different degrees of evaluations).

Apples and oranges (and bananas).

So you need to roll out all reasonable candidates.
Ya your right. I should done it differently.

That being said if I remember correctly the changes in GNU were minimal from 3ply to the rollout and the differences in percents were pretty big from lets say 23/18 to 13/10, 6/4 so I didn't think it was too necessary as I sort of knew what the result would be (if that makes sense?)

XG took something crazy like 2.5 hours for my move so I called it quits... I will try and set up the position again and do it like I should have though, thanks for the comment.
Black to play 3-2 Quote
12-30-2011 , 06:46 AM
Quote:
Originally Posted by nonotyet
Ya your right. I should done it differently.

That being said if I remember correctly the changes in GNU were minimal from 3ply to the rollout and the differences in percents were pretty big from lets say 23/18 to 13/10, 6/4 so I didn't think it was too necessary as I sort of knew what the result would be (if that makes sense?)

XG took something crazy like 2.5 hours for my move so I called it quits... I will try and set up the position again and do it like I should have though, thanks for the comment.
I've started a rolloput for you on my comp. Should be faster. GNU is a lot slower than XG, so I am wondering how many trials you have done for that rollout?

So far 23/18 is slightly better than 18/15 6/4, but a lot of moves are reasonable.
Black to play 3-2 Quote
12-30-2011 , 10:08 AM
Code:
XGID=-----BCbB--baC--cbAcbBAA--:0:0:1:23:0:0:0:1:10

    1. Rollout¹    23/18                        eq:-0.264
      Player:   36.82% (G:6.57% B:0.58%)
      Opponent: 63.18% (G:27.31% B:1.84%)
      Confidence: ±0.006 (-0.270..-0.257) - [95.7%]
      Duration: 36 minutes 50 seconds

    2. Rollout¹    18/15 6/4                    eq:-0.272 (-0.008)
      Player:   36.41% (G:6.48% B:0.86%)
      Opponent: 63.59% (G:29.39% B:3.52%)
      Confidence: ±0.007 (-0.278..-0.265) - [4.2%]
      Duration: 41 minutes 07 seconds

    3. Rollout¹    18/13                        eq:-0.278 (-0.014)
      Player:   36.10% (G:7.01% B:0.60%)
      Opponent: 63.90% (G:25.69% B:2.24%)
      Confidence: ±0.006 (-0.284..-0.272) - [0.1%]
      Duration: 36 minutes 46 seconds

    4. Rollout¹    21/18 6/4                    eq:-0.283 (-0.019)
      Player:   35.87% (G:6.42% B:0.39%)
      Opponent: 64.13% (G:32.44% B:3.48%)
      Confidence: ±0.007 (-0.289..-0.276) - [0.0%]
      Duration: 31 minutes 31 seconds

    5. Rollout¹    13/10 6/4                    eq:-0.293 (-0.029)
      Player:   35.35% (G:6.46% B:0.81%)
      Opponent: 64.65% (G:33.57% B:4.46%)
      Confidence: ±0.008 (-0.301..-0.285) - [0.0%]
      Duration: 45 minutes 04 seconds

    6. Rollout¹    13/8                         eq:-0.300 (-0.036)
      Player:   35.01% (G:6.28% B:0.62%)
      Opponent: 64.99% (G:30.61% B:2.90%)
      Confidence: ±0.007 (-0.307..-0.293) - [0.0%]
      Duration: 34 minutes 11 seconds

    7. Rollout¹    23/21 18/15                  eq:-0.301 (-0.037)
      Player:   34.97% (G:6.07% B:0.77%)
      Opponent: 65.03% (G:29.92% B:2.26%)
      Confidence: ±0.006 (-0.307..-0.294) - [0.0%]
      Duration: 40 minutes 42 seconds


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


eXtreme Gammon Version: 2.01, MET: Rockwell-Kazaross
Black to play 3-2 Quote
12-30-2011 , 12:50 PM
I noticed that this was a problem that came up at DMP, might be useful to RO as a money game problem and then see what's top pick if white owns the cube or black owns the cube.

Not sure this will change anything though because 23/18 accomplishes so much...
Black to play 3-2 Quote
01-01-2012 , 02:52 AM
FYI: You can post an image by clicking the yellow icon that shows a mountain and little round sun. Then, just enter the URL for imageshack, and click OK.

Here is the original position:



Next come your two screen shots:



Black to play 3-2 Quote

      
m