BGonline.org Forums

MCG vs. Jeb Horton -- Rollouts and Evals

Posted By: Matt Cohn-Geier
Date: Thursday, 2 October 2008, at 4:51 a.m.

In Response To: MCG vs. Jeb Horton (Matt Cohn-Geier)

Position 1:

OTB I didn't like 8/4 6/5 as it strips the 8 and leaves me with a big stack on the mid. 13/8 looked quite wrong. 13/9 6/5 looked pretty natural to me: unstack the heavy mid and 6, White has no board and I have an anchor. I don't relish being hit but it doesn't hurt me too much. If he hits me with a 2 he can't do so and break the anchor at the same time. If I make the 9 point I'm happy as I can block 22 and put some pressure on the checkers on the 18 while bringing more checkers towards my home board. All in all, this play looked clear to me, GNU dinged me for .0446. The rollout shows 8/4 6/5 to be better but not by as much. Not sure what is going on here or if the swing on his 2s really is that big.

The score (after 0 games) is: Jeb Horton 0, Matt Cohn-Geier 0 (match to 9 points)

Move number 12: Matt Cohn-Geier to play 41

Jeb Horton	144
Matt Cohn-Geier	161
Position ID: 4Nsc4ADgc/ABBg Match ID: cAkmAQAAAAAA

• Matt Cohn-Geier moves 13/9 6/5

Alert: bad move ( -0.022)

#		Ply	Move	Equity
	1	R	8/4 6/5	-0.3478
			0.4003 0.0698 0.0025 - 0.5997 0.0751 0.0022 -0.2046 -0.3478 0.0011 0.0011 0.0003 - 0.0011 0.0011 0.0002 0.0027 0.0058
			Full cubeful rollout with var.redn.
			2381 games, Mersenne Twister dice gen. with seed 834890880 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
•	2	R	13/9 6/5	-0.3698 ( -0.0220)
			0.3967 0.0744 0.0028 - 0.6033 0.0885 0.0031 -0.2211 -0.3698 0.0013 0.0011 0.0003 - 0.0013 0.0013 0.0003 0.0032 0.0064
			Full cubeful rollout with var.redn.
			2379 games, Mersenne Twister dice gen. with seed 834890880 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	3	R	13/8	-0.3993 ( -0.0515)
			0.3898 0.0625 0.0021 - 0.6102 0.0800 0.0024 -0.2387 -0.3993 0.0016 0.0015 0.0004 - 0.0016 0.0018 0.0003 0.0039 0.0079
			Full cubeful rollout with var.redn.
			1026 games, Mersenne Twister dice gen. with seed 834890880 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Position 2:

This was interesting. Not sure whether or not I would have broken anchor here. Rollout is underway.

The score (after 0 games) is: Jeb Horton 0, Matt Cohn-Geier 0 (match to 9 points)

Move number 17: Jeb Horton to play 64

Jeb Horton	130
Matt Cohn-Geier	138
Position ID: aM+DAQawtznAAA Match ID: MAEzAQAAAAAA

• Jeb Horton moves 18/8

Alert: doubtful move ( -0.011)

#		Ply	Move	Equity
	1	3	7/3 7/1	+0.2015
			0.5679 0.0699 0.0014 - 0.4321 0.0734 0.0011
•	2	3	18/8	+0.1904 ( -0.0111)
			0.5604 0.0700 0.0015 - 0.4396 0.0821 0.0011

Position 3:

I tried to get cute here.

What I saw: if I play 18/13 my back checker will be severely disconnected and he's got 10 in the zone. If I stick around on the bar point his 9 point and midpoint are stripped, I can roll a 6 to make the bar point, and he won't really love to hit me with 2s or 6s. Even with his aces he still won't love hitting me. Also, if he uses the checkers on the 9 point to make inner board points (which he did as he rolled a 43 after my play), I have a direct shot at him.

What I didn't see: the race is close and I am ahead. Awkward stacks are not the way to win this game, racing leads are. True, my checker will have to weave an obstacle course, but he may do so successfully, and I'm more likely to survive that way.

What I still don't see: why 18/13 6/4* is coming out better than 18/13 10/8. It gains a tempo but everything else about it looks wrong. Still TCTC of course, but like, ugh.

The score (after 1 game) is: Jeb Horton 0, Matt Cohn-Geier 2 (match to 9 points)

Move number 5: Matt Cohn-Geier to play 52

Jeb Horton	159
Matt Cohn-Geier	158
Position ID: 4HPDgAXgc+JBIA Match ID: cIkqAQAAEAAA

• Matt Cohn-Geier moves 13/8 10/8

Alert: bad move ( -0.080)

#		Ply	Move	Equity
	1	R	18/13 6/4*	-0.2540
			0.4634 0.0913 0.0074 - 0.5366 0.1523 0.0067 -0.1541 -0.2540 0.0022 0.0019 0.0016 - 0.0022 0.0022 0.0007 0.0058 0.0139
			Full cubeful rollout with var.redn.
			995 games, Mersenne Twister dice gen. with seed 835177876 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	2	R	18/13 10/8	-0.2741 ( -0.0200)
			0.4594 0.0705 0.0037 - 0.5406 0.1456 0.0025 -0.1705 -0.2741 0.0020 0.0017 0.0014 - 0.0020 0.0022 0.0006 0.0052 0.0116
			Full cubeful rollout with var.redn.
			995 games, Mersenne Twister dice gen. with seed 835177876 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
•	3	R	13/8 10/8	-0.3336 ( -0.0795)
			0.4417 0.0821 0.0021 - 0.5583 0.1625 0.0038 -0.2122 -0.3336 0.0019 0.0017 0.0007 - 0.0019 0.0023 0.0007 0.0051 0.0109
			Full cubeful rollout with var.redn.
			995 games, Mersenne Twister dice gen. with seed 835177876 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Position 4:

I saw that I was coming up on a cube here and wanted to do it efficiently rather than by way of overkill. The 4 prime with builders looks quite strong, however if I hit White and he enters I don't have much of anything.

In retrospect: my play doesn't actually give me a cube. If I hit White and he dances then I have a very powerful cube. I am still not sure on this one since the 4 prime looks quite strong but I will reverse my choice and go with 15/5 7/2*(2). Rollout underway.

The score (after 2 games) is: Jeb Horton 2, Matt Cohn-Geier 2 (match to 9 points)

Move number 8: Matt Cohn-Geier to play 55

Jeb Horton	137
Matt Cohn-Geier	147
Position ID: ZpvCASiwe+AJIA Match ID: cIk2ASAAEAAA

• Matt Cohn-Geier moves 15/5 13/8(2)

Alert: doubtful move ( -0.018)

#		Ply	Move	Equity
	1	3	15/5 7/2*(2)	+0.3797
			0.5744 0.2270 0.0123 - 0.4256 0.1433 0.0040
•	2	3	15/5 13/8(2)	+0.3618 ( -0.0178)
			0.5920 0.1738 0.0081 - 0.4080 0.1292 0.0024

Position 5:

I didn't put too much thought into doubling here. I am behind -5, -3, I can roll a 3 from the bar, 44, 11, and 33 are crushing, 26 and 46 are good numbers, hitting two is also very good, plus I have a ton of stuff to hit inside my board and it all adds up to big big market losers. White isn't going to turn the cube at me for awhile even if I leave it in the center, by doubling him I pretty much take away his recube for the rest of the game. I was surprised when GNU flagged this as an error but a rollout shows that GNU is way off here.

The score (after 6 games) is: Jeb Horton 6, Matt Cohn-Geier 4 (match to 9 points)

Move number 10: Matt Cohn-Geier on roll, cube decision?

Jeb Horton	135
Matt Cohn-Geier	129
Position ID: ZLfEAwIbPokBUA Match ID: cAkgAWAAIAAA

• Matt Cohn-Geier doubles

Cube decision
Rollout cubeless equity		+0.0383	(Money: +0.0408)
Cubeful equities:
1.	Double, take	+0.2145
2.	Double, pass	+1.0000	+0.7855
3.	No double	+0.0982	-0.1163
Proper cube action:		Double, take
Rollout details
Win W g W bg   Lose L g L bg Cubeless Cubeful Centered 1-cube 0.4689 0.2852 0.0187 - 0.5311 0.1830 0.0180 +0.0383 +0.0982 Standard error 0.0024 0.0027 0.0018 - 0.0024 0.0036 0.0035 0.0086 0.0149 Player Jeb Horton owns 2-cube 0.4758 0.2967 0.0202 - 0.5242 0.1840 0.0407 +0.3706 +0.2145 Standard error 0.0032 0.0042 0.0019 - 0.0032 0.0055 0.0061 0.0202 0.0230
Full cubeful rollout with var.redn.
395 games, Mersenne Twister dice gen. with seed 835398816 and quasi-random dice
Play: world class 2-ply cubeful prune [world class]
keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
Skip pruning for 1-ply moves.
Cube: 2-ply cubeful prune [world class]

Position 6:

GNU flagged this as a big error. I had made a similar mistake the night before in not running vs. Munitz when I should have, this time I decided to jst run and cut down on his shots. I like 22/17 6/3 a bit better now, but it isn't the .07 error GNU says it is.

The score (after 6 games) is: Jeb Horton 6, Matt Cohn-Geier 4 (match to 9 points)

Move number 22: Matt Cohn-Geier to play 53

Jeb Horton	169
Matt Cohn-Geier	113
Position ID: sBvxgEHb+IABFA Match ID: QYkuAWAAIAAA

• Matt Cohn-Geier moves 22/14

Alert: bad move ( -0.012)

#		Ply	Move	Equity
	1	R	22/17 6/3	+0.2128
			0.5118 0.2604 0.0026 - 0.4882 0.1016 0.0086 +0.3252 +0.2128 0.0012 0.0013 0.0004 - 0.0012 0.0020 0.0010 0.0034 0.0039
			Full cubeful rollout with var.redn.
			3560 games, Mersenne Twister dice gen. with seed 835398816 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
•	2	R	22/14	+0.2009 ( -0.0120)
			0.5212 0.2359 0.0018 - 0.4788 0.0924 0.0078 +0.3162 +0.2009 0.0012 0.0015 0.0002 - 0.0012 0.0019 0.0009 0.0032 0.0039
			Full cubeful rollout with var.redn.
			3559 games, Mersenne Twister dice gen. with seed 835398816 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Position 7:

Same theme, I ran with 23/14 and GNU flagged it again in favor of 23/17 14/11. This time I think GNU is off, I don't need the fourth checker and making the 14 point is fine. I will probably leave a shot next roll but I might not and I have some spares on the 6 point to play with. White's prime doesn't look like much but if I get hit and he starts bringing checkers off the mid it could be imposing soon. Granted, going after the 4th checker will jack up my gammons, but probably not by as much as GNU thinks. Will roll it out.

The score (after 6 games) is: Jeb Horton 6, Matt Cohn-Geier 4 (match to 9 points)

Move number 24: Matt Cohn-Geier to play 63

Jeb Horton	169
Matt Cohn-Geier	105
Position ID: sBvxgEHb+IAFEA Match ID: QQkvAWAAIAAA

• Matt Cohn-Geier moves 23/14

Alert: doubtful move ( -0.011)

#		Ply	Move	Equity
	1	3	23/17 14/11	+0.5263
			0.5935 0.3383 0.0060 - 0.4065 0.0811 0.0016
•	2	3	23/14	+0.5154 ( -0.0109)
			0.6208 0.2842 0.0039 - 0.3792 0.0718 0.0013

Position 8:

I made the 12 point here. I think I looked at 12/8 briefly but didn't see how it could be right. I don't really feel like giving him immediate 11s. TCTC, GNU prefers 12/8.

The score (after 7 games) is: Jeb Horton 6, Matt Cohn-Geier 8 (match to 9 points, Crawford game)

Move number 16: Matt Cohn-Geier to play 31

Jeb Horton	168
Matt Cohn-Geier	134
Position ID: Ym8BYCzgu6EjAA Match ID: 8IklAWAAQAAA

• Matt Cohn-Geier moves 16/12

Alert: doubtful move ( -0.004)

#		Ply	Move	Equity
	1	R	12/8	+0.2901
			0.6479 0.2594 0.0452 - 0.3521 0.0619 0.0014 +0.2901 +0.2901 0.0022 0.0081 0.0041 - 0.0022 0.0021 0.0004 0.0044 0.0044
			Full cubeful rollout with var.redn.
			1323 games, Mersenne Twister dice gen. with seed 834790729 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
•	2	R	16/12	+0.2864 ( -0.0037)
			0.6452 0.2425 0.0343 - 0.3548 0.0643 0.0005 +0.2864 +0.2864 0.0020 0.0076 0.0035 - 0.0020 0.0019 0.0075 0.0039 0.0039
			Full cubeful rollout with var.redn.
			1322 games, Mersenne Twister dice gen. with seed 834790729 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Position 9:

I played 13/9 here. My thought was if White hits me, at least he'll have to break the 5 point, and I also want to reduce shots. I did see 13/12 6/3 but rejected it. It does good things of course, I would like inner board points much more than the 9 point, and the best way to make inner board points is to slot them. I thought reducing shots here was paramount, but GNU says it's worth the extra risk. 12/8, leaving two blots, didn't even really occur to me and I still hate it, but I guess it's better than my play.

The score (after 7 games) is: Jeb Horton 6, Matt Cohn-Geier 8 (match to 9 points, Crawford game)

Move number 20: Matt Cohn-Geier to play 31

Jeb Horton	161
Matt Cohn-Geier	119
Position ID: stsAYCzge4cFAA Match ID: 8IklAWAAQAAA

• Matt Cohn-Geier moves 13/9

Alert: bad move ( -0.053)

#		Ply	Move	Equity
	1	R	13/12 6/3	+0.1455
			0.5769 0.2163 0.0291 - 0.4231 0.0880 0.0019 +0.1455 +0.1455 0.0021 0.0080 0.0031 - 0.0021 0.0022 0.0004 0.0043 0.0043
			Full cubeful rollout with var.redn.
			1296 games, Mersenne Twister dice gen. with seed 835177876 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	2	R	12/8	+0.1166 ( -0.0289)
			0.5625 0.2419 0.0401 - 0.4375 0.0930 0.0019 +0.1166 +0.1166 0.0022 0.0072 0.0035 - 0.0022 0.0025 0.0005 0.0044 0.0044
			Full cubeful rollout with var.redn.
			1295 games, Mersenne Twister dice gen. with seed 835177876 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
•	3	R	13/9	+0.0927 ( -0.0528)
			0.5519 0.2307 0.0413 - 0.4481 0.1048 0.0030 +0.0927 +0.0927 0.0020 0.0070 0.0040 - 0.0020 0.0027 0.0006 0.0043 0.0043
			Full cubeful rollout with var.redn.
			1295 games, Mersenne Twister dice gen. with seed 835177876 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Position 10:

GNU gave me another ding here as I played 7/6 7/3 and tried to kill my numbers as fast as possible while maintaining a 4 point board but not exposing blots for him to get the second checker. GNU says: I shouldn't be worried about the blot until he hits me. The best way to maintain a 4 point board is to make the 4 point, not the 6 point where I still have 5s and 4s to play. Either 8/4 2/1 or 8/4 7/6 is likely better than my play, but by somewhat less than .02.

The score (after 7 games) is: Jeb Horton 6, Matt Cohn-Geier 8 (match to 9 points, Crawford game)

Move number 42: Matt Cohn-Geier to play 41

Jeb Horton	91
Matt Cohn-Geier	82
Position ID: 2bYdIAC7YxsAEA Match ID: 8AkmAWAAQAAA

• Matt Cohn-Geier moves 7/6 7/3

Alert: doubtful move ( -0.004)

#		Ply	Move	Equity
	1	R	8/4 2/1	-0.8823
			0.0638 0.0073 0.0001 - 0.9362 0.0810 0.0029 -0.8823 -0.8823 0.0006 0.0008 0.0000 - 0.0006 0.0014 0.0002 0.0014 0.0014
			Full cubeful rollout with var.redn.
			4174 games, Mersenne Twister dice gen. with seed 834790729 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
•	2	R	7/6 7/3	-0.8867 ( -0.0044)
			0.0603 0.0071 0.0000 - 0.9397 0.0711 0.0019 -0.8867 -0.8867 0.0006 0.0008 0.0000 - 0.0006 0.0013 0.0001 0.0012 0.0012
			Full cubeful rollout with var.redn.
			4174 games, Mersenne Twister dice gen. with seed 834790729 and quasi-random dice
			Play: world class 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Messages In This Thread

• MCG vs. Jeb Horton
  Matt Cohn-Geier -- Wednesday, 1 October 2008, at 4:14 p.m.
  • MCG vs. Jeb Horton -- Crawford
    Matt Cohn-Geier -- Wednesday, 1 October 2008, at 4:26 p.m.
  • my guess...
    KenB -- Wednesday, 1 October 2008, at 5:48 p.m.
  • MCG vs. Jeb Horton -- Rollouts and Evals
    Matt Cohn-Geier -- Thursday, 2 October 2008, at 4:51 a.m.
    • MCG vs. Jeb Horton -- Rollouts and Evals
      Stick -- Thursday, 2 October 2008, at 8:35 a.m.
      • MCG vs. Jeb Horton -- Rollouts and Evals
        Matt Cohn-Geier -- Thursday, 2 October 2008, at 7:21 p.m.
  • so many problems,
    garyo -- Thursday, 2 October 2008, at 4:19 p.m.