Any ideas about why these 2 move evaluate so differently in GNU?
mont...@lycos.com
Match ID : cAlnAAAACAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: opponent
| O | | O O X O | 0 points
| O | | O O O |
| | | O O O |
| | | O O |
| | | |
v| |BAR| | 3 point match (Cube:
1)
| | | |
| | | |
| X X | | |
| X X X | | X X X | Rolled 61
| O X X X | | X X O X | 1 point
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: me
23-17, 8-7 is +.657; 23-16 is +.620 (2 py).
TIA.
mont...@lycos.com
23-17, 8-7 is +.447 cubeless, +.681 cubeful.
My opponent gets a 15/36 shot at the blot on either roll, but 8-7 puts
the spare on the bar pt. so that I don't need to break it if I want to
hit - but is that worth .042 ? It seems like too much.TIA.
tallrock
Philippe Michel
If X stays on the 17pt, it is a little more difficult for O to safety his
blot on the 12pt.
You could edit the position and look at the equities not only after the
two "real" moves but after 23/17 and 23/16 8/7 as well. That should give
an indication on the relative importance of the place ot the spare vs.
that of the backman (assuming the 1 pip difference matters even less).
Grunty
Whether you place a spare on the 8 or 7-point is irrelevant here.
The difference lies on the number of extra safe rolls you allow him by
pointlessly advancing to the 16-point: 4-1 and 3-2 (4 rolls, a full
11% of all possible rolls).
mont...@lycos.com
certainly worth a bit, so if 23-16, O has to use the whole roll to hit.
My guess is that this 2/36 situation is much of the difference, after
thinking about it.
Grunty
Sorry, my mistake, you already pointed out both plays leave the same
number of shots. Not a matter of safe rolls - have to rethink... :-(
Grunty
O more vulnerable, because he's left with less timing (flexibility) to
play the other die, or the next roll.
For instance, from the 16-point he can play 6 safe, not so from the 17.
While small numbers play fairly safe from either point.
The key must be in that the more advanced the hit, the more forced big
numbers will play - the less ground to manoeuver to avoid leaving a
shot.
mont...@lycos.com
gammons and 8 more games (and loses 2 less gammons). Let's break down
the moves:
1,1 is very close - O needs to move up the backman and make his 2
point, no hit (says GNU).
2,1 is a bit better for O after 23-17
3,1 is much better for O after 23-16.
4,1 allows O to hit and move up the backman after 23-16 but not after
23-17, and is much better for O after 23-16.
5,1 is interesting. O would prefer to hit and move up even though he
leaves a blot on his 8 point and 21 points, rather than pass it and
move up, as he would after 23-16 (the difference is more than .300).
6,1 doesn't make a big difference.
2,2 is better for O after 23-16, obviously.
2,3 is better after 23-17.
2,4 is better after 23-16.
2,5 is very close; 23-16 is just a bit worse.
2,6 is the same.
3,4 is better after 23-16.
3,5 is better after 23-17.
3,6 is the same.
4,5 23-16 is about .140 better for O than 23-17.
4,6 is clearly better for O after 23-16.
5,6 is better after 23-17, but only very slightly.
3,3 is the same.
4,4 is a bit better for O after 23-16.
5,5 is much better for O after 23-17 (.335), since he makes his 2
point.
6,6 is very slightly better for O after 23-17.
Thus, it seems like the big difference is how much better 4,1 is for O
after 23-16. The other rolls seem to counterbalance each other about
equally.
Am I missing something?
Raccoon
On Nov 3, 10:22 am, monty1...@lycos.com wrote:
> GNU sees the difference as being that with 23-17, X wins 17 more
> gammons and 8 more games (and loses 2 less gammons). Let's break down
> the moves:
[snip]
> Thus, it seems like the big difference is how much better 4,1 is for O
> after 23-16. The other rolls seem to counterbalance each other about
> equally.
>
> Am I missing something?
You are correct that there is a big difference between 23/16 and 23/17
8/7 if Opponent rolls 4-1, but your list of moves had some obvious
errors. For instance, if Opponent rolls 61 X will very much wish he had
played 23/16, giving him a double direct shot with 6s and 4s, instead
of only 5s and 7s). If Opponent rolls 54 there's little difference
(after either play Opponent plays 13/X*/4, hitting).
To get a handle on the differences, try using GNUBG's temperature map:
1. Analyze both plays.
2. In the analysis window, select "Temp. Map".
3. In the temperature map window, select "2 ply" and "Show equities"
(and if you want to, "Show best move").
Here's what I get for X's equity, according to GNUBG 2-ply (your
equities will differ somewhat depending on which Match Equity Table you
use):
Opponent's Move 23/17 8/7 23/16 Advantage 23/17 8/7
41 0.649 0.218 0.431
22 0.693 0.345 0.348
42 0.836 0.613 0.223
64 0.782 0.607 0.175
31 0.829 0.655 0.174
44 0.774 0.654 0.120
43 0.794 0.690 0.104
11 0.657 0.643 0.014
54 0.626 0.613 0.013
63 0.802 0.792 0.010
33 0.788 0.788 0.000
62 0.759 0.763 (0.004)
66 0.786 0.793 (0.007)
65 0.696 0.720 (0.024)
52 0.687 0.720 (0.033)
21 0.808 0.863 (0.055)
53 0.683 0.782 (0.099)
32 0.649 0.801 (0.152)
55 0.243 0.600 (0.357)
51 0.221 0.600 (0.379)
61 0.022 0.442 (0.420)
23/16 allows Opponent to hit with a 4 and so you can see from the list
above that 23/16 8/7 is much better if Opponent rolls any 4, including
22, except 54.
23/17 8/7 allows Opponent to hit with a 5. But you can from the list
above that only the rolls 51 55 32 53 give a big advantage to 23/16.
After 23/17 8/7 ... 65: 13/8*/2, X has a direct shot. After 23/17 8/7
... 52: 13/8* 3/1 X has indirect shots. And 54 makes little difference,
as was noted already.
mont...@lycos.com
move the backman all the way up than to play for less gammon losses by
just moving it up one pip and playing 13-7 with the other blot.
It's not easy to see that one play is clearly better than the other
when you have little time and know that you may need some later on in
the match, though to me this is what makes the game more interesting,
practically-speaking, than chess.
Since playing under such circumstances means that you really need
general ideas about what to look for, the "take home message" here
seems to be that you should consider which is best if your opponent
rolls hitting numbers. However, my thought at the time was that if he
hit after 23-17, I would have more return shots, and only a fairly
thoughtful analysis would most "good" players see that there may be
more to it that this.
tallrock
insisted on it after a rollout. It seems that 23-16 is best because it
covers against the one joker O has, another roll of 6--1 to leap the
prime.
Raccoon
> Not surprisingly, actually, Snowie's analysis differs and even
> insisted on it after a rollout. It seems that 23-16 is best because it
> covers against the one joker O has, another roll of 6--1 to leap the
> prime.
GNUBG 2-ply rollout also prefers 23/16:
GNU Backgammon Position ID: zz0GAgjDtjsAEA
Match ID : cAlnAAAACAAA
+24-23-22-21-20-19------18-17-16-15-14-13-+ O:
| O X O O | | O | 0 points
| O O O | | O |
| O O O | | |
| O O | | |
| | | |
| |BAR| |v 3 point match (Cube:
1)
| | | |
| | | |
| | | X X |
| X X X | | X X X | Rolled 61
| X O X X | | X X X O | 1 point
+-1--2--3--4--5--6-------7--8--9-10-11-12-+ X:
1. Rollout 23/16 Eq.: +0.6894
0.6913 0.0907 0.0022 - 0.3087 0.0527 0.0003 CL +0.4500 CF
+0.6894
[0.0012 0.0040 0.0004 - 0.0012 0.0029 0.0001 CL 0.0036 CF
0.0068]
2. Rollout 23/17 8/7 Eq.: +0.6613 (
-0.0281)
0.6812 0.0967 0.0023 - 0.3188 0.0438 0.0002 CL +0.4754 CF
+0.6613
[0.0011 0.0030 0.0003 - 0.0011 0.0026 0.0001 CL 0.0037 CF
0.0069]
Full cubeful rollout with var.redn., 1296 games, Mersenne Twister dice
gen. with seed 1014515584 and quasi-random dice
Play: 2-ply cubeful prune, up to 5 more moves within equity 0.08, skip
1-ply pruning
Cube: 2-ply cubeful prune [world class]