|
BGonline.org Forums
Magriel VS. GNU (long)
Posted By: Randy
Date: Monday, 21 July 2008, at 1:14 a.m.
I like Paul Magriel's classic book "BACKGAMMON" a lot, but I was always concerned about how some of the examples might stack up against modern, neural-network based advice. So, I went through the book with GNU, set on 2-ply, to take a look at the positions (I didn't want to make a career of it and do rollouts of the positions).
I have compiled the results, and decided to post them for whatever use they may be to others. A few points:
1. This shouldn't necessarily be construed as a list of Magriel errata, just a list of where GNU 2-ply disagrees.
2. I put all the positions into a 7-away, 7-away match, and used cubeful equity evaluations.
3. I have only given results for positions where GNU's recommended move is not the same as Magriel's. In most cases, the equities are pretty close (see stats below).
4. I only examined positions in chapters 12 through 29, excluding chapter 15 (openings). I excluded the backgame section of chapter 29, as bots are still suspect at backgames in rollouts, much less in 2-ply examinations.
5. For positions listed, the GNU recommended move is followed by the ordered list of other possibilities, with their equity differences from the GNU recommended move. The last move in the list is Magriel's recommended move. Occasionally when the list was long and I got lazy, I inserted an ellipsis (...) to show there were other moves between GNU's recommendation and Magriel's recommendation.
Please point out errors if you find them – thanks.
Randy
Statistics as follows:Positions examined: 267
Positions where GNU's move agreed with Magriel's move: 175
Average equity loss of Magriel's move, all 267 positions: .026
Average equity loss of Magriel's move, only the 92 positions where GNU disagrees: .076
Median equity loss of Magriel's move, only the 92 positions where GNU disagrees: .046
Chapter 12 - Builders and Flexibility
Position 12.1
GNU prefers 1A or 1B to 1C
8/7 8/5
8/6(2) -.026
7/6 7/4 -.041Position 12.4
GNU prefers 4A to 4B
10/9 6/4
10/8 6/5 -.142Position 12.8
GNU prefers 8A to 8B
9/5
8/5 6/5 -.020Position 12.11B
8/5* 6/5
6/5*(2) 8/7 -.016Position 12.12
b/23 13/12*/11*
b/24 13/12*/11*/10 -.017
b/24 13/12*/11* 7/6 -.057Position 12.16B
24/23
10/9 -.069Position 12.18
8/5 7/5
7/4 6/4 -.048Position 12.19
10/7 9/7
13/10 24/22 -.046Position 12.20
10/2
10/8 10/4 -.004Chapter 13 - Duplication and Diversification
Position 13.7D
9/3
9/8 9/4 -.028Position 13.11
18/16 5/4
18/15 -.017
18/17 6/4 -.017Position 13.14
24/18 13/11
24/22 13/7 -.002
13/7 6/4 -.011
13/11 13/7 -.038Position 13.20
24/18 13/11
24/22 13/7 -.002
24/22 24/18 -.008
13/11 13/7 -.102Position 13.24
b/21* 5/4
b/21* 6/5 -.034
b/21* 8/7 -.046Position 13.25A
21/15 10/6
...
10/6 9/3 -.086Position 13.25B
20/14 10/6
...
9/5 10/4 -.020Chapter 14 - When You Are Forced to Leave Shots
Position 14.3
16/11 6/5
16/11 3/2 -.004
16/10 -.022Position 14.4
8/7 8/2
8/1 -.118
11/4 -.163Position 14.8
4/o 3/o
4/o 4/1* -.026Position 14.12
20/18
4/2 -.040Position 14.15A
8/4
7/3 -.033Position 14.15B
8/4 6/4
14/8 -.137
7/1* -.207Chapter 16 - Safe Play vs. Bold Play
Position 16.4
7/6 (what an ugly move)
6/5 -.001
9/8 -.088Position 16.6
GNU agrees with the main point, don't slot, but
24/23
11/10 -.016Position 16.7
24/23
11/10 -.016
6/5 -.019Position 16.9
13/10
8/5* -.129Position 16.10
13/11 6/5
13/10 -.006Position 16.14
13/7 6/5
24/18 6/5 -.009
13/6 -.016Chapter 17 - Slotting
Position 17.2
24/23
11/10 -.016
6/5 -.019Position 17.8
13/10
24/23 13/11 -.045
24/23 8/6 -.050
8/5 -.095
13/11 6/5 -.118
3/2* 3/1 -.122
13/11 8/7 -.135Position 17.12
GNU agrees that switching points 4/3*(2) is best but prefers to
play the last 1 as 24/23 rather than 8/7 equity difference -.034.Position 17.13
6/5
8/7 -.072Position 17.16A
GNU agrees that slotting 8/7 isn't good but has a slight preference
(.002 equity) for playing 11/10 rather than 6/5.Position 17.16B
b/22 6/5
b/22 8/7 -.044Chapter 18 - Action Play
Position 18.1
9/5 6/5
24/20 6/5 -.021Position 18.2
24/22 13/7
9/1 -.003
24/18 9/7 -.013Position 18.3
11/6
23/18 -.170Position 18.4
8/4 6/5
8/7 8/4 -.007
13/9 6/5 -.024
24/20 6/5 -.037Position 18.5
10/5 6/5
23/22 10/5 -.096
23/18 6/5 -.149Chapter 19 - One Man Back
Position 19.2
11/5 6/5
13/7 8/7 -.098Position 19.4
13/8 6/5
8/3 6/5 -.011
22/21 6/1* -.012
13/7 -.014
8/7 6/1* -.023Position 19.5
15/11 13/8
13/9 6/1* -.004
15/11 6/1* -.013
15/10 13/9 -.017
11/7 6/1* -.025Position 19.10
24/21 13/9
24/20 13/10 -.008Chapter 20 - Golden Point
Position 20.1
24/22(2) 6/4(2)
13/11(2) 6/4(2) -.007
24/22 13/11 6/4(2) -.027
24/20(2) -.029Position 20.5
20/14*
8/2 -.136Position 20.7
b/16*
b/20 24/20 -.066Position 20.8
24/20*
8/5 6/5 -.074Position 20.11
No Double, Take
Double, Take -.052Chapter 21 - Splitting
Position 21.1A
13/10 11/10
24/23. 11/8 -.017
13/10 6/5 -.024Position 21.1B
13/10 11/10
13/10 6/5 -.024Position 21.4
8/4 6/5
8/7 8/4 -.007
13/9 6/5 -.024
24/20 6/5 -.037Position 21.12
13/11
24/22 -.060Chapter 22 - Doubling Theory
Position 22.4
GNU believes X is too good to double by a hair.
No Double
Double, pass -.003
(Double, take +.449)Position 22.5
Double, take
Double, pass +.258
(that is O should take and loses .258 equity by passing)Chapter 23 - No Possible Contact
Only positions 8A and 8B were checked - both OK
Chapter 24 - Avoiding Contact
Position 24.11
10/5
8/3 -.149Position 24.15
6/o 2/o
6/o 3/1 -.077Position 24.24
6/4 6/2*
...
6/2* 4/2 -.017Position 24.40
8/6 5/1
9/5 8/6 -.211
9/3 -.216Position 24.42
12/6 11/6
12/1 -.000
12/7 11/5 -.111Position 24.43
16/10 11/10
16/15 11/5 -.076
16/10 6/5 -.083
16/9 -.087Position 24.44
18/17*/15
18/17* 6/4 -.002
...
18/16 6/5 -.259Position 24.46
18/16*/13
...
18/13 -.085Position 24.47
3/o
6/5 6/4 -.045Position 24.49
10/4 8/4
7/1* 5/1 -.033Position 24.52
18/16* 13/8
...
18/11 -.192Position 24.58
6/3
9/6 -.030Chapter 25 - After Contact
Position 25.1
16/13
20/17 -.005
17/14 -.012Position 25.2
11/9
6/4 -.057Position 25.5
18/14
16/12 -.046Position 25.6
18/13
9/4 -.131Position 25.7
24/22 13/7
13/5 -.069
13/11 13/7 -.118Position 25.16
20/16 13/7
20/10 -.006
20/14 13/9 -.006
13/7 5/1 -.023Position 25.17
20/16 19/13
20/10 -.003
20/14 19/15 -.004
19/9 -.006
19/13 5/1 -.022Position 25.22
13/10 6/1*
19/16 6/1* -.010
13/5 -.013Chapter 26 - Priming and Blocking
Position 26.10
24/22 13/9
24/20 11/9 -.013
13/9 11/9 -.033Position 26.11A
9/4 8/4
13/9 13/8 -.033Position 26.12A
24/22(2) 13/11(2)
...
13/9(2) -.126Position 26.14
14/9 13/8 6/1*(2)
8/3(2) 6/1*(2) -.028
13/8(2) 6/1*(2) -.031
14/9 13/3 8/3 -.073Position 26.15
6/4 6/3
13/8 -.016
8/3 -.031
13/11 6/3 -.087Position 26.17
24/18
24/23 11/6 -.048
13/7 -.068
13/8 11/10 -.070Position 26.18
13/10 5/3*
24/21 5/3* -.029
11/8 5/3* -.046Chapter 27 - Control of the Outside
Position 27.8
8/5
6/3 -.082Position 27.14
11/7 10/7
24/20 13/10 -.039Chapter 28 - Timing
Position 28.2
18/13
7/2* -.151Position 28.4
24/13
24/18 7/2* -.151Position 28.7
13/8
6/1* -.030Position 28.9A
6/2 4/1
10/7 10/6 -.117Position 28.11A
8/6 8/3
...
20/13 -.259Position 28.14
10/6 4/2
9/5 4/2 -.053
10/6 9/7 -.058Position 28.15
24/23*
10/9 -.469Position 28.19
8/6 8/2*
...
8/2* 3/1* -.371Chapter 29 - Holding Game and Backgame
(only positions 1 - 9 examined - no backgames)Position 29.2
24/16(2)
24/20 13/9(2) 8/4 -.026
24/20(2) 13/9(2) -.030
24/20 13/5 13/9 -.033Position 29.8
13/11 13/7
7/1 3/1 -.042