Search This Blog

December 31, 2014

Early 65: Run or Split?

 I.  Theory

65 run 24/13 is often automatic on the first or second move.  But from the 3rd move move, sometimes in the early game it is right to run 24/13 while other times it is right to split 24/18, 13/8.

How to decide?  Stick covered this concept in a recent Gammon Village article.  He provided several main factors to consider.

(1) Race: 
  • Down in the race --> Split
  • Up in the race --> Race
Discussion

The race is always an important factor to consider.  The logic here is if you are dwonin the race you want contact while if you are up the race you want to race.  24/13 safely escapes a man, furthering the race game plan.  Splitting with 24/18 seeks contact when you are down in the race, either for an exchange of hits or to simply make the Bar point anchor next turn.

(2) Boardage:
  • Better board --> Split
  • Worse board --> Run
Discussion 

A blot hitting contest will tend to favor the side with the better board.  Usually in these positions the boardage is 2 to 1 one way or the other.  This may not seem like much but consider you are four times more likely to dance on a 2 point board than a 1 point board.

Also keep in mind that getting hit on opponent's bar point only costs seven pips while if you score a return hit on the opponent's bar point it costs him 18 pips.  Thus an exchange of hits will gain pips in the race.

(3) Midpoint & 8 Point Distribution:
  • Stacked midpoint (5 or more) & stripped 8 point (only 2) --> Split
  • Midpoint not stacked & 8 point not stripped --> Run
Discussion

This simple observation struck me as rather obvious when Stick wrote it yet I'd never thought of it before! 

Running with 24/13 is less attractive if you are creating a stack on the midpoint.  This is especially true if you are putting a sixth man on the midpoint.

Similarly, Splitting is especially attractive the 8 point is stripped.  Here, playing the 5 from 13/8 reinforces the 8 point and usually creates a visually appealing "nice" balance of builders.

(4) Opponent Outfield Blot:
  • Opponent has extra outfield blot or point --> Split
  • Opponent does not have outfield blot or point -->  Run
Discussion

Stick writes "the blot or even point in the ourfield" can be a target if they don't use it in some fashion."

I would add splitting to the bar is the classic distraction play.  If the opponent is busy hitting you on his bar point, at least he is distracted from making an inner point.

(5) Other minor Factors:
  • Stick mentioned that you are more likely to split when opponent own 10 point than 9 point.  Reason is the general duplication of 3's from the 10 point.
  • Opponent advanced anchor.  If opponent has an advanced anchor than you generally want one too.  So splitting may be more desirable than usual.  Please use this one with caution since this is my own rule and not from Stick.

 II. Practice

**************
(1) 63S-31P-61@-65

is Player 2

score: 0
pip: 151
Unlimited Game
Jacoby Beaver
pip: 163
score: 0

is Player 1
XGID=-----BDbB---dE--ac-e----B-:0:0:1:56:0:0:3:0:10
to play 56

1.XG Roller++24/18 13/8eq: -0.215
Player:
Opponent:
44.26% (G:8.16% B:0.25%)
55.74% (G:11.22% B:0.29%)
2.XG Roller++24/13eq: -0.271 (-0.056)
Player:
Opponent:
43.04% (G:6.23% B:0.17%)
56.96% (G:10.98% B:0.22%)
3.3-ply13/2eq: -0.296 (-0.082)
Player:
Opponent:
41.84% (G:9.43% B:0.25%)
58.16% (G:14.47% B:0.54%)
eXtreme Gammon Version: 2.10

This was Stick's first example from his article.

Here all the factors point towards splitting.  Mainly you are down in the race and have the better board so you want contact.  Further, the midpoint/8-point interplay visually looks bad after 24/13 when you get 6 on midpoint and 2 on 8 point.  Contrast that with 24/18, 13/8 when you get a more balanced distro of 5 on midpoint and 3 on 8 point.  Additionally, there is a blot on the 9 point so you get the distraction bonus by splitting.  Finally, note my criterion on anchor also points to splitting as opponent is anchored on your bar point and you are stuck on the ace point.

****************
(2) 53P-33B-65

is Player 2

score: 0
pip: 155
Unlimited Game
Jacoby Beaver
pip: 159
score: 0

is Player 1
XGID=---Bb-D-B---cE-b-c-e----B-:0:0:1:56:0:0:3:0:10
to play 56

1.XG Roller++24/18 13/8eq: -0.189
Player:
Opponent:
46.02% (G:9.25% B:0.33%)
53.98% (G:14.52% B:0.45%)
2.XG Roller++24/13eq: -0.209 (-0.020)
Player:
Opponent:
45.75% (G:7.51% B:0.24%)
54.25% (G:13.57% B:0.36%)
3.3-ply13/8 13/7eq: -0.287 (-0.098)
Player:
Opponent:
42.67% (G:9.55% B:0.29%)
57.33% (G:15.40% B:0.72%)
eXtreme Gammon Version: 2.10

This is Stick's other example from the article.

Here the criteria are mixed but the preponderance are pointing towards splitting.

You are up in the race argues for running but everything else points towards splitting.

You have the better board, the Midpoint/8-point distro and extra outifeld point all points towards splitting.  My extra criterion (opponent's anchor) also points to the split.

******************
(3) 53P-44B-65

is Player 2

score: 0
pip: 151
Unlimited Game
Jacoby Beaver
pip: 159
score: 0

is Player 1
XGID=---B-bD-B---cE--bc-e----B-:0:0:1:56:0:0:3:0:10
to play 56

1.XG Roller++24/13eq: -0.352
Player:
Opponent:
42.36% (G:5.52% B:0.16%)
57.64% (G:14.05% B:0.30%)
2.XG Roller++24/18 13/8eq: -0.356 (-0.004)
Player:
Opponent:
42.25% (G:6.79% B:0.21%)
57.75% (G:15.38% B:0.41%)
3.3-ply13/8 13/7eq: -0.435 (-0.082)
Player:
Opponent:
39.30% (G:7.30% B:0.23%)
60.70% (G:16.59% B:0.69%)
eXtreme Gammon Version: 2.10

I found this one on the forums. You can find discussion from this thread. http://www.bgonline.org/forums/webbbs_config.pl?read=114804

This is similar to the above problem. Splitting is probably correct despite the above evaluation.  Dmitriy posted a rollout showing split correct by .017

Probably the reason the split is less correct than in #2 is opponent has the 9 point rather than 10 point.

Dmitriy writes:  (Note Blue refers to you and White the opponent)

This is XG1 rollout.

Race is basically even after the roll, so running play doesn't gain much. If Blue runs, he hopes that he will roll another running number before White primes him in. By splitting he stays connected and puts a spare from overstocked mid point to stripped 8pt. White has an advanced anchor,so by splitting Blue makes a bid for an advanced anchor himself, which would equalize the game. 18pt blot is exposed to plenty of shots, but often he has return hits from the bar or rolls like 54 and 43 that make an anchor. If he splits and White rolls 52 or 41 he is getting hit, but White would make 4 or 5pt if Blue chose to run instead. If White rolls 42 or 31 he should not hit [hit is a blunder], but build his home board instead. It is easy to see that after these sequences split gains a lot as he will have plenty of numbers to anchor on 18pt. Split also gives White opportunity to make more mistakes. With 32 he should not hit, but I think a lot of humans will. With 63 White should hit and slot 4pt [he would be better of by making 3pt if Blue run].

***************
(4) 31P-21S-65

is Player 2

score: 0
pip: 164
Unlimited Game
Jacoby Beaver
pip: 163
score: 0

is Player 1
XGID=-aa--BD-B---dEa--c-e----B-:0:0:1:56:0:0:3:0:10
to play 56

1.XG Roller++24/18 13/8eq: +0.327
Player:
Opponent:
57.60% (G:16.82% B:0.82%)
42.40% (G:10.53% B:0.38%)
2.XG Roller++24/13eq: +0.319 (-0.008)
Player:
Opponent:
58.30% (G:13.63% B:0.57%)
41.70% (G:9.58% B:0.28%)
3.4-ply13/2* eq: +0.240 (-0.087)
Player:
Opponent:
54.84% (G:19.08% B:0.94%)
45.16% (G:11.58% B:0.65%)
eXtreme Gammon Version: 2.10

This is an old one from the forums (2007).

http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=114706

The rollout in the link seems to match the ++ eval.  Namely Split but essentially a toss up.

Looking at Stick's criteria:

Up in the race points to racing

Better board, Mid/8 Distro, and outfield blot point to splitting.

****************
(5) 31P-53P-65

is Player 2

score: 0
pip: 159
Unlimited Game
Jacoby Beaver
pip: 163
score: 0

is Player 1
XGID=-b---BD-B---eE---b-d--b-B-:0:0:1:56:0:0:3:0:10
to play 56

1.XG Roller++24/13eq: +0.255
Player:
Opponent:
56.68% (G:14.25% B:0.63%)
43.32% (G:11.39% B:0.35%)
2.XG Roller++24/18 13/8eq: +0.251 (-0.003)
Player:
Opponent:
56.26% (G:16.89% B:0.92%)
43.74% (G:13.20% B:0.50%)
3.3-ply13/8 13/7eq: +0.143 (-0.112)
Player:
Opponent:
53.67% (G:17.12% B:0.79%)
46.33% (G:13.18% B:0.77%)
eXtreme Gammon Version: 2.10

Another old one from the forums.  An old GNU rollout has running best by .014

http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=114632

Looking at Stick's criteria:

Race and no extra outfield blots/points argue for racing.

"Better" board and mid/8 point distro argue for splitting so it's close.

I put better board in quotes since the actual boardage is 2 to 2.  But surely your 5 point is better than opponent's 3 point. 

No wonder it's close but note if you count the boardage as a wash than the preponderance of criteria is slightly in favor of running, consistent with the rollout.



December 19, 2014

Presser

Below are selected errors from a match against David Presser. This is from the Chicago Open in May and represents my worst PR performance of the year. Meaning I have plenty of errors to choose from.


9 away 9 away, 54 to play. Hit or not?

is David Presser

score: 0
pip: 156
9 point match
pip: 165
score: 0

is Bill
XGID=----b-DAD---cCb--baeB--A--:0:0:1:54:0:0:0:9:10
to play 54

1.XG Roller++23/18* 7/3eq: +0.036
Player:
Opponent:
50.62% (G:10.76% B:0.32%)
49.38% (G:9.46% B:0.33%)
2.XG Roller++8/3 7/3eq: -0.042 (-0.078)
Player:
Opponent:
48.24% (G:9.48% B:0.29%)
51.76% (G:8.48% B:0.28%)
3.XG Roller+23/18* 13/9eq: +0.033 (-0.003)
Player:
Opponent:
50.39% (G:11.78% B:0.42%)
49.61% (G:10.38% B:0.36%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I made the point with 8/3, 7/3.

Better was to hit 23/18*, 7/3.

Looking at this six months later, everything points to the hit.  I am down in the race and have an anchor.  The man on the 23 point is horrible and hitting takes away half the roll.  That's plenty of upside.

Perhaps I was thinking the hit is unlikely to work since I get back with most 3's, 4's and 6's.  But he has a one point board so there is really no danger (little downside) and only upside.  Probably the best argument for hitting is that I give David the whole roll to do something good his turn like make the bar or an inner point.

****************

Later in the same game and I am being doubled.  Take or Drop?

is David Presser

score: 0
pip: 129
9 point match
pip: 132
score: 0

is Bill
XGID=---B-bE-D---bC--b-bbc--b-A:0:0:-1:00:0:0:0:9:10
on roll, cube action?

Analyzed in XG Roller++ No double Double/Take
  Player Winning Chances: 71.06% (G:25.57% B:0.23%) 71.17% (G:26.51% B:0.26%)
  Opponent Winning Chances: 28.94% (G:4.12% B:0.14%) 28.83% (G:3.90% B:0.14%)
  Cubeless Equities +0.635 +1.332
Cubeful Equities
     No double:+0.895 (-0.105)
     Double/Take:+1.117 (+0.117)
xg Double/Pass:+1.000
Best Cube action: Double / Pass
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I took a bad one.

This position is similar to #48 of the recent study group.

I am down in all phases of the game;  outboarded, down in the race, and on the bar possibly against a four point board next time.

Probably I was thinking the race is close, it's only a 3 point board and I have counterplay against the men on the 5 anchor.

While this is true, the key seems to be the gammons.  I am looking at a net 25% differential here, much more than I would have guesses.   I am creating an on the fly rule of thumb that these types of positions are low 70's win percent.  To take, therefore you need the gammons to be very low.  Meaning I have way to many outside crossovers here to think about taking.  The only way you take something like this is if the position is analygous to a containment cube.  You got hit bearing in for example.  In a scenario like that you have almost no outside crossovers and are way up in the race.  Of course your opponent wouldn't be cubing then since he has a lot of work to do but the point is made nethertheless.  Gammons would be low so we can take with low 20% winning chances in that scenario.

Another broad lesson is beware of taking a cube when you are behind in all phases of the game.

***************

I lead 7 away 9 away and have 65 to play.

is David Presser

score: 0
pip: 152
9 point match
pip: 156
score: 2

is Bill
XGID=-a-a--E-C--AcD-A-bce----A-:0:0:1:65:2:0:0:9:10
to play 65

1.XG Roller++15/9 8/3* eq: -0.323
Player:
Opponent:
44.31% (G:11.60% B:0.46%)
55.69% (G:17.69% B:1.00%)
2.XG Roller++15/9 11/6eq: -0.433 (-0.110)
Player:
Opponent:
40.92% (G:8.94% B:0.30%)
59.08% (G:16.19% B:0.55%)
3.3-ply15/9 13/8eq: -0.344 (-0.021)
Player:
Opponent:
43.12% (G:9.84% B:0.36%)
56.88% (G:16.26% B:0.71%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I made the hopeless play 15/9, 11/6.  This play is so obviously bad I'm not sure why I am even writing this up.  Perhaps I was thinking as follows:  I am up in the race and don't want to get hit.  So leave the six shot, which is a lot, but less than a double shot.

Best is 15/9, 8/3*  I have to leave a direct anyway so I should to strive to do something constructive.

My move 11/6 is the opposite of constructive as I place a sixth man on the 6 point.
Also my move let's David have his whole roll to work with.  He either hits, makes an inside point, anchors, runs.  Everything plays well.  Not good.

On the other hand hitting at least takes half the roll away.  Now David has some bad rolls (like entering sixes or anything that misses).  Also if he is hitting at least he isn't making an inner board point and at any rate David only has a one point board right now so getting hit isn't the end of the world.

******************

I lead 7 away 8 away.  61P-64R-61S-53??

is David Presser

score: 1
pip: 153
9 point match
pip: 157
score: 2

is Bill
XGID=-aa---E-C---cEA--bce----A-:0:0:1:53:2:1:0:9:10
to play 53

1.XG Roller++8/3 6/3eq: -0.169
Player:
Opponent:
47.90% (G:11.50% B:0.48%)
52.10% (G:15.32% B:0.48%)
2.XG Roller++14/11 13/8eq: -0.255 (-0.086)
Player:
Opponent:
45.95% (G:10.25% B:0.37%)
54.05% (G:15.36% B:0.47%)
3.3-ply24/16eq: -0.220 (-0.051)
Player:
Opponent:
47.25% (G:8.15% B:0.32%)
52.75% (G:15.18% B:0.48%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I moved 14/11, 13/8.

Best was making the 3 point 8/3, 6/3.

It's usually right to make an inner point rather than clean up that loose outside blot early in the game.

Making a point is always constructive.  He is favored to miss so most of the time you don't get hit.  Even if I do get hit there is some good two duplication at play here.   at least he isn't making a point to hit and is stripping the midpoint.

***********

Later in the same game.  32 play it to take it.


is David Presser

score: 1
pip: 131
9 point match
pip: 144
score: 2

is Bill
XGID=-aa---EBC---aD---acdbb--A-:0:0:1:32:2:1:0:9:10
to play 32

1.XG Roller++13/8eq: -0.692
Player:
Opponent:
38.40% (G:8.36% B:0.39%)
61.60% (G:25.86% B:0.55%)
2.XG Roller++24/22 13/10eq: -0.886 (-0.194)
Player:
Opponent:
38.05% (G:8.03% B:0.46%)
61.95% (G:33.09% B:0.81%)
3.2-ply6/1* eq: -0.877 (-0.186)
Player:
Opponent:
35.76% (G:7.67% B:0.30%)
64.24% (G:28.60% B:0.83%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I panicked, moving 24/22, 13/10.

Best was the simple 13/8.

Apparently I wanted to challenge by getting to the edge of the prime immediately. This is the wrong idea here. I don't want to initiate a fight on David's side of the board, where I am outgunned 13 to 1. Better to stay back on the ace where the prime versus prime gameplan has some chance of success. If he does attack me at least it's deep where he'd rather not be.

Also my move exposes another blot which is a bad idea since I will likely be on the bar next time. Further, when I scoot up his only bad rolls are sixes really. But by moving up now 62 and 63 hit the outside man while 64 and 65 hit loose on his inner board. So I am diversifiying his numbers.

All in all a terrible play.

*******************

Next roll and David cubes Take or Drop?

is David Presser

score: 1
pip: 131
9 point match
pip: 139
score: 2

is Bill
XGID=-aa---EBC-A-aC---acdbbA---:0:0:-1:00:2:1:0:9:10
on roll, cube action?

Analyzed in Rollout No double Double/Take
  Player Winning Chances: 61.89% (G:33.86% B:0.92%) 61.74% (G:34.64% B:0.94%)
  Opponent Winning Chances: 38.11% (G:8.78% B:0.62%) 38.26% (G:9.28% B:0.96%)
  Cubeless Equities +0.512 +1.108
Cubeful Equities
     No double:+0.623 (-0.241)±0.017 (+0.606..+0.640)
xg Double/Take:+0.863±0.026 (+0.838..+0.889)
     Double/Pass:+1.000 (+0.137)
Best Cube action: Double / Take
Rollout details
1296 Games rolled with Variance Reduction.
Moves and cube decisions: 1-ply
Double Decision confidence:100.0%
Take Decision confidence:100.0%
Duration: 11.7 seconds
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I was crapping my pants and dropped this.

The computer shows this is takeable by a sizable margin.

I don't feel too bad about this one though as I find this hard to judge. We know we get gammoned an awful lot. What is more difficult to see is that we win this game almost 40% of the time. That makes up for a lot of gammons.

****************

7 away 7 away and I am on roll. Cube action?

is David Presser

score: 2
pip: 140
9 point match
pip: 127
score: 2

is Bill
XGID=a-aBCBCA----bB-a--bdbb--B-:0:0:1:00:2:2:0:9:10
on roll, cube action?

Analyzed in XG Roller++ No double Double/Take
  Player Winning Chances: 60.53% (G:40.12% B:1.86%) 60.65% (G:40.29% B:1.78%)
  Opponent Winning Chances: 39.47% (G:11.84% B:0.93%) 39.35% (G:11.99% B:0.93%)
  Cubeless Equities +0.510 +1.063
Cubeful Equities
     No double:+0.620 (-0.124)
xg Double/Take:+0.744
     Double/Pass:+1.000 (+0.256)
Best Cube action: Double / Take
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I rolled, probably confused as to what is going on here. Is a four point black "more than you know" or "less than you think?"
Being confused translate to double per Woolsey's law if nothing else.

Would you want to face a cube here? Doubling would have given David a tough decision. Even though he is an excellent player he still could make a mistake by dropping - but only if I double.

Bottom line is the position is too volatile. Things are going to clear up very soon one way or the other. And a significant amount of the time they clarify with me winning a G-ball. So I simply must cube and see what happens.

***************

I lead 5 away 7 away and have double twos. Your play?

is David Presser

score: 2
pip: 138
9 point match
pip: 110
score: 4

is Bill
XGID=-Ba-B-DbC---bD--aa-bbbb---:0:0:1:22:4:2:0:9:10
to play 22

1.XG Roller++13/11(4)eq: +0.789
Player:
Opponent:
71.72% (G:13.98% B:0.13%)
28.28% (G:3.59% B:0.08%)
2.XG Roller++6/2*(2) eq: +0.491 (-0.298)
Player:
Opponent:
64.79% (G:16.24% B:0.14%)
35.21% (G:5.84% B:0.14%)
3.2-ply13/9(2)eq: +0.531 (-0.258)
Player:
Opponent:
66.95% (G:14.04% B:0.18%)
33.05% (G:5.63% B:0.14%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I went for the attack with 6/2*(2).  I was probably thinking that by putting him on the bar I give him 16 dancing rolls.  If he dances I can double (which actually isn't even true but I probably thought that).  And if he comes in at least I flush him forward.

But this is the wrong gameplan.  I am up in the race so I want to race.  Double twos allows me to disengage the men on the midpoint and get closer to home.  This was the correct play.  Also note I have some spares aiming at the open five point and can hope to close it with a future roll.


memo: position after dance.

is Player 2

score: 2
pip: 140
9 point match
pip: 102
score: 4

is Player 1
XGID=aBB-B-BbC---bD--aa-bbbb---:0:0:1:00:4:2:0:9:10
on roll, cube action?

Analyzed in XG Roller++ No double Double/Take
  Player Winning Chances: 64.03% (G:21.64% B:0.19%) 64.29% (G:21.65% B:0.21%)
  Opponent Winning Chances: 35.97% (G:6.22% B:0.12%) 35.71% (G:6.07% B:0.12%)
  Cubeless Equities +0.424 +0.875
Cubeful Equities
No double:+0.537
Double/Take:+0.478 (-0.059)
Double/Pass:+1.000 (+0.463)
Best Cube action: No double / Take
Percentage of wrong pass needed to make the double decision right: 10.2%
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

*******************

Score is 5 away 5 away. I have already cubed and have 31 to play.

is David Presser

score: 4
pip: 169
9 point match
pip: 108
score: 4

is Bill
XGID=a-BB-BBcC-A-cB--Ac-bc-----:1:-1:1:31:4:4:0:9:10
to play 31

1.XG Roller++16/15 13/10eq: +0.855
Player:
Opponent:
74.61% (G:41.09% B:0.36%)
25.39% (G:3.83% B:0.15%)
2.XG Roller++10/6eq: +0.832 (-0.023)
Player:
Opponent:
74.83% (G:38.64% B:0.25%)
25.17% (G:3.47% B:0.12%)
3.XG Roller++16/13 10/9eq: +0.742 (-0.113)
Player:
Opponent:
73.38% (G:34.46% B:0.23%)
26.62% (G:3.51% B:0.11%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I made the "obvious" and pedestrian 16/13, 10/9. How could safetying the back checker be a whopper?

To be honest I don't know. The computer likes 16/15, 13/10. I suppose you are taking the chance to clear the midpoint now while he is on the bar against a 4 point board. So basically if David doesn't rolla joker now he is toast. A tough one in my opinion.

Note the second best play keeps the midpoint but duplicates fours.

******************

I trail 5 away Crawford. 65 to play. Break the anchor or not?

is David Presser

score: 8
pip: 108
9 point match
Crawford
pip: 110
score: 4

is Bill
XGID=--ABBBBBA-A-bb---c-cBbba--:0:0:1:65:4:8:1:9:10
to play 65

1.XG Roller++10/5 8/2eq: +0.018
Player:
Opponent:
50.62% (G:5.91% B:0.10%)
49.38% (G:2.35% B:0.06%)
2.XG Roller++20/9eq: -0.043 (-0.061)
Player:
Opponent:
47.33% (G:8.88% B:0.29%)
52.67% (G:3.49% B:0.11%)
3.XG Roller+7/2 7/1eq: +0.027 (+0.009)
Player:
Opponent:
51.10% (G:5.53% B:0.09%)
48.90% (G:2.37% B:0.06%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I bolted with 20/9. Not the worst play in the world but wrong.

Better was 10/5, 8/2. I am not up enough in the race to disengage and the roll plays well so I am better to wait.