| |
BGonline.org Forums
Snowie Rollout and Analysis
Posted By: Paul Weaver In Response To: Where the Hell is Stick? (Chuck Bower)
Date: Monday, 24 November 2008, at 3:23 p.m.
According to my Snowie rollout (data posted below), the surprise winner is 24/22 6/5. Very few players (certainly not I) would have made this play over the board. Making the 5pt (7/5 6/5), according to the rollout, is a very close second.
Here are some of the advantages of 24/22 6/5 over 7/5 6/5:
Splitting with 24/22 6/5 duplicates White’s 5s (13/8 and 8/3*) and 1s (21/20* and 6/5). White has to forgo making her 4pt to hit with 11 and forgo making an outfield point to hit with 41 61. She has to give up an advanced anchor to hit with her aces (other than 11), after which Blue has four punishing doublet returns from the bar. After White hits with 21 41 61, Blue also gets a lot of additional return shots.
By making the 5pt (7/5 6/5), Blue is trying to prime an opponent who already has an advanced anchor. He should, instead, make sure that he is not the one who gets primed on the 24pt. Splitting makes it tougher for White to develop a prime and gives Blue more shots after White’s lousy rolls (54 62 63 65). There is always a risk involved in splitting, but it is minimal when White has only two builders. Splitting is the best way for Blue to exploit White’s garbage distribution.
By slotting the 5pt (6/5) instead of making it (7/5 6/5), Blue keeps alive the possibility of covering it next roll with an ace, thereby making the most efficient possible use of his 6pt stack.
Magriel’s well-known advice (from 1976, many backgammon centuries ago) not to split and slot simultaneously is often correct. However, there are many exceptions, and this position is one of them.
The gammon save play is 24/21, not risking getting a third checker sent back. At DMP and gammon go, 24/22 6/5 is best by a larger margin than for money.
(Can someone please tell me a better way to post Snowie rollout results?)
1. R 24/22 6/5 Eq.: -0.241
0.6% 10.7% 45.6% 54.4% 19.1% 1.0%
Live cube rollout: -0.253
95% confidence interval:
- money cubeless eq.: -0.177 ±0.005,
- live cube: -0.253 ±0.008.
Rollout settings:
Truncated rollout, depth 7, 648 games,
played 3-ply (precise), cube 3-ply,
settlement 0.550 at 8 pts,
random seed, without race database.
2. R 7/5 6/5 Eq.: -0.250 (-0.009)
0.5% 10.6% 44.4% 55.6% 17.2% 0.9%
Live cube rollout: -0.271
95% confidence interval:
- money cubeless eq.: -0.183 ±0.004,
- live cube: -0.271 ±0.008.
Rollout settings:
Truncated rollout, depth 7, 648 games,
played 3-ply (precise), cube 3-ply,
settlement 0.550 at 8 pts,
random seed, without race database.
3. R 24/23 13/11 Eq.: -0.263 (-0.022)
0.5% 9.4% 44.4% 55.6% 17.1% 0.8%
Live cube rollout: -0.281
95% confidence interval:
- money cubeless eq.: -0.193 ±0.005,
- live cube: -0.281 ±0.008.
Rollout settings:
Truncated rollout, depth 7, 648 games,
played 3-ply (precise), cube 3-ply,
settlement 0.550 at 8 pts,
random seed, without race database.
4. R 24/22 24/23 Eq.: -0.265 (-0.024)
0.4% 8.9% 44.4% 55.6% 16.8% 0.7%
Live cube rollout: -0.281
95% confidence interval:
- money cubeless eq.: -0.193 ±0.004,
- live cube: -0.281 ±0.007.
Rollout settings:
Truncated rollout, depth 7, 648 games,
played 3-ply (precise), cube 3-ply,
settlement 0.550 at 8 pts,
random seed, without race database.
5. R 24/21 Eq.: -0.267 (-0.026)
0.5% 9.1% 44.2% 55.8% 16.3% 0.7%
Live cube rollout: -0.284
95% confidence interval:
- money cubeless eq.: -0.191 ±0.005,
- live cube: -0.284 ±0.008.
Rollout settings:
Truncated rollout, depth 7,
648 games (equiv. 277546 games),
played 3-ply (precise), cube 3-ply,
settlement 0.550 at 8 pts,
random seed, without race database.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.