Discussion:
Gammon loss alert!
Add Reply
Tim Chow
2020-07-07 01:30:25 UTC
Reply
Permalink
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10

X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31

---
Tim Chow
Paul Epstein
2020-07-07 06:19:52 UTC
Reply
Permalink
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
Definitely maintain the position of the rear checker.
Not only are 61/51 shotleavers for the opponent but leaving
would give the opponent a more efficient bearoff.
Note also that there is no backgammon danger, and no favourable
hitting rolls for O.

Once the stay decision is made, 7/6 7/4 is clear.
Given the stay decision, it's the only play with two crossovers,
and an empty point is occupied.

7/6 7/4 with 95% confidence (which is my max confidence for
positions that are not amenable to exact computation).

Paul
Peter
2020-07-07 07:42:03 UTC
Reply
Permalink
Post by Paul Epstein
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
Definitely maintain the position of the rear checker.
Not only are 61/51 shotleavers for the opponent but leaving
would give the opponent a more efficient bearoff.
Note also that there is no backgammon danger, and no favourable
hitting rolls for O.
Once the stay decision is made, 7/6 7/4 is clear.
Agreeeeeeeed.
Post by Paul Epstein
Given the stay decision, it's the only play with two crossovers,
and an empty point is occupied.
7/6 7/4 with 95% confidence (which is my max confidence for
positions that are not amenable to exact computation).
Paul
Paul Epstein
2020-07-07 09:20:34 UTC
Reply
Permalink
Post by Peter
Post by Paul Epstein
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
Definitely maintain the position of the rear checker.
Not only are 61/51 shotleavers for the opponent but leaving
would give the opponent a more efficient bearoff.
Note also that there is no backgammon danger, and no favourable
hitting rolls for O.
Once the stay decision is made, 7/6 7/4 is clear.
Agreeeeeeeed.
Post by Paul Epstein
Given the stay decision, it's the only play with two crossovers,
and an empty point is occupied.
7/6 7/4 with 95% confidence (which is my max confidence for
positions that are not amenable to exact computation).
Paul
The DMP play is even easier. At DMP, we assume that we will hit, and
position ourselves as well as possible for the post-hit game.
The slot 7/4 is obvious and 6/5 probably creates the best building
opportunities. So 7/4 6/5 at DMP.
Does this contradict the DMP rule because it's different to the money play?
Or does this vindicate the DMP rule because both plays stay back?

Paul
b***@gmail.com
2020-07-08 09:00:37 UTC
Reply
Permalink
Post by Paul Epstein
Post by Peter
Post by Paul Epstein
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
Definitely maintain the position of the rear checker.
Not only are 61/51 shotleavers for the opponent but leaving
would give the opponent a more efficient bearoff.
Note also that there is no backgammon danger, and no favourable
hitting rolls for O.
Once the stay decision is made, 7/6 7/4 is clear.
Agreeeeeeeed.
Post by Paul Epstein
Given the stay decision, it's the only play with two crossovers,
and an empty point is occupied.
7/6 7/4 with 95% confidence (which is my max confidence for
positions that are not amenable to exact computation).
Paul
The DMP play is even easier. At DMP, we assume that we will hit, and
position ourselves as well as possible for the post-hit game.
The slot 7/4 is obvious and 6/5 probably creates the best building
opportunities. So 7/4 6/5 at DMP.
Does this contradict the DMP rule because it's different to the money play?
Or does this vindicate the DMP rule because both plays stay back?
Paul
This is an obvious exception.

Stick
Michael
2020-07-07 20:21:49 UTC
Reply
Permalink
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
---
Tim Chow
I was attracted to this by the term "alert" in the tittle.
Usually other verdicts don't convince me, unless I do my own thinking and after doing so, I found Paul's reasoning excellent.
Some more points to mention:
a)filling the 4 point gap is good for saving the gammon anyway.
b)the move gets 2 checkers in which is again fine for the purpose of saving the gammon
c)O may delay more than one exchange to clear her 6 point so 61, or 51 may get repeated.
d)we have some possibility of winning the game if we hit (51 isa very bad roll for O leaving so many blots around). So most propably the gammon to Single tradeoff is in our favor.

As for Paul playing a slightly different move at DMP I think there shouldn't be much equity difference between the two.
Michael
2020-07-07 20:32:32 UTC
Reply
Permalink
invalid regarding the blots-->(51 isa very bad roll for O leaving so many blots around)
BlueDice
2020-07-08 08:05:07 UTC
Reply
Permalink
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
---
Tim Chow
7/6 7/4
If we run we have very little chance of saving the gammon and zero chance of winning the game. Staying back we have some winning chance and better gammon saving chance.
--
BD
Tim Chow
2020-07-14 02:35:18 UTC
Reply
Permalink
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10

X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31

Paul's analysis of these kinds of plays are excellent. At DMP, XG favors
7/4 and then either 11/10 or 6/5.

1. Rollout¹ 7/6 7/4 eq:-1.539
Player: 1.79% (G:0.00% B:0.00%)
Opponent: 98.21% (G:57.76% B:0.48%)
Confidence: ±0.002 (-1.541..-1.537) - [100.0%]

2. Rollout¹ 11/8 7/6 eq:-1.630 (-0.090)
Player: 1.74% (G:0.00% B:0.00%)
Opponent: 98.26% (G:66.62% B:0.52%)
Confidence: ±0.002 (-1.632..-1.628) - [0.0%]

3. Rollout¹ 11/10 7/4 eq:-1.642 (-0.102)
Player: 1.98% (G:0.00% B:0.00%)
Opponent: 98.02% (G:68.48% B:0.41%)
Confidence: ±0.002 (-1.643..-1.640) - [0.0%]

4. Rollout¹ 20/17 7/6 eq:-1.657 (-0.118)
Player: 0.00% (G:0.00% B:0.00%)
Opponent: 100.00% (G:65.69% B:0.00%)
Confidence: ±0.001 (-1.658..-1.656) - [0.0%]

5. Rollout¹ 7/4 6/5 eq:-1.665 (-0.126)
Player: 1.98% (G:0.00% B:0.00%)
Opponent: 98.02% (G:70.90% B:0.44%)
Confidence: ±0.002 (-1.667..-1.663) - [0.0%]

6. Rollout¹ 7/4 5/4 eq:-1.667 (-0.128)
Player: 1.93% (G:0.00% B:0.00%)
Opponent: 98.07% (G:70.94% B:0.44%)
Confidence: ±0.002 (-1.669..-1.665) - [0.0%]

7. Rollout¹ 7/4 3/2 eq:-1.670 (-0.131)
Player: 1.86% (G:0.00% B:0.00%)
Opponent: 98.14% (G:70.99% B:0.44%)
Confidence: ±0.002 (-1.672..-1.668) - [0.0%]

8. Rollout¹ 7/4 2/1 eq:-1.671 (-0.131)
Player: 1.89% (G:0.00% B:0.00%)
Opponent: 98.11% (G:71.09% B:0.46%)
Confidence: ±0.002 (-1.673..-1.669) - [0.0%]

¹ 1296 Games rolled with Variance Reduction.
Dice Seed: 271828
Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.19.207.pre-release

---
Tim Chow
Paul Epstein
2020-07-14 21:03:51 UTC
Reply
Permalink
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
Paul's analysis of these kinds of plays are excellent. At DMP, XG favors
7/4 and then either 11/10 or 6/5.
I'm not convinced that XG is undecided between these two plays at DMP.
To ascertain this, one needs a DMP rollout.
In some of the variations where we would win at DMP, we are forced to run
to save the backgammon or gammon.
So, assuming the 1.98% win figure is accurate, at DMP, we will definitely
win more games than that.

Having said all that, I wouldn't be surprised if 11/10 is as good or
even best. That play creates more cover shots for the 4 but 6/5 looks more
natural. As for Michael saying that 7/6 7/4 doesn't lose much equity at
DMP, he might be right because there aren't that many winning choices to lose.
But one can only be shocked, mortified and struck with horror at someone
willing to entertain such an egregious play as 7/6 7/4 which doesn't even
attempt to cover the 4 point blot.
There are two types of blundering in games of skill like chess or backgammon.
One is that you lose a lot of equity (or in chess, your expected win/draw/lose
result is much worse). Another way a move is a blunder is that the consequence
of your mistake isn't that large (you don't lose much equity) but the fact that
it's wrong is glaringly obvious -- even to a Trump advisor who has been woken
up by a phone call from Jared Kushner at 3 am after having fallen asleep
in a drunken haze after an excess of whisky.
For example, it's a type of blunder to play 13/8 13/10 with an opening 53
because everyone loves the three point but you probably don't lose much equity.
7/6 7/4 is the Trumpian move.

Paul
b***@gmail.com
2020-07-15 00:54:17 UTC
Reply
Permalink
Post by Paul Epstein
Post by Tim Chow
XGID=--AB-BEC---A-------bAaa-b-:1:1:1:31:0:0:0:0:10
X:Player 1 O:Player 2
Score is X:0 O:0. Unlimited Game
+13-14-15-16-17-18------19-20-21-22-23-24-+
| | | O X O O O |
| | | O O |
| | | |
| | | |
| | | |
| |BAR| |
| | | X |
| | | X |
| X | | X | +---+
| X | | X X X | | 2 |
| X X | | X X X X | +---+
+12-11-10--9--8--7-------6--5--4--3--2--1-+
Pip count X: 100 O: 21 X-O: 0-0
Cube: 2, X own cube
X to play 31
Paul's analysis of these kinds of plays are excellent. At DMP, XG favors
7/4 and then either 11/10 or 6/5.
I'm not convinced that XG is undecided between these two plays at DMP.
To ascertain this, one needs a DMP rollout.
In some of the variations where we would win at DMP, we are forced to run
to save the backgammon or gammon.
So, assuming the 1.98% win figure is accurate, at DMP, we will definitely
win more games than that.
Having said all that, I wouldn't be surprised if 11/10 is as good or
even best. That play creates more cover shots for the 4 but 6/5 looks more
natural. As for Michael saying that 7/6 7/4 doesn't lose much equity at
DMP, he might be right because there aren't that many winning choices to lose.
But one can only be shocked, mortified and struck with horror at someone
willing to entertain such an egregious play as 7/6 7/4 which doesn't even
attempt to cover the 4 point blot.
There are two types of blundering in games of skill like chess or backgammon.
One is that you lose a lot of equity (or in chess, your expected win/draw/lose
result is much worse). Another way a move is a blunder is that the consequence
of your mistake isn't that large (you don't lose much equity) but the fact that
it's wrong is glaringly obvious -- even to a Trump advisor who has been woken
up by a phone call from Jared Kushner at 3 am after having fallen asleep
in a drunken haze after an excess of whisky.
For example, it's a type of blunder to play 13/8 13/10 with an opening 53
because everyone loves the three point but you probably don't lose much equity.
7/6 7/4 is the Trumpian move.
Paul
I rolled the two plays out at dmp for you since it takes all of the blink of an eye. The plays are interchangeable.

Stick
Tim Chow
2020-07-16 03:19:31 UTC
Reply
Permalink
Post by Paul Epstein
I'm not convinced that XG is undecided between these two plays at DMP.
To ascertain this, one needs a DMP rollout.
I did one (might even have been 5184 trials) before making that claim,
but didn't bother posting the results.

---
Tim Chow
Paul Epstein
2020-07-17 08:08:39 UTC
Reply
Permalink
Post by Tim Chow
Post by Paul Epstein
I'm not convinced that XG is undecided between these two plays at DMP.
To ascertain this, one needs a DMP rollout.
I did one (might even have been 5184 trials) before making that claim,
but didn't bother posting the results.
---
Tim Chow
Sorry, I didn't realise that. In what you posted, the plays were exactly
tied at unlimited, and I thought that's what you were describing.

I should have known better than to suspect such a basic mistake.
If you happen to know, the GWC at DMP (with one of the best plays), I'd
be interested to see how it compares with unlimited.
But don't bother if this is any trouble -- I just thought you might have
that stat already at hand.

Why don't I just get a bot myself?
Because I don't want to tempt myself to spend too much time at
backgammon, which isn't my job, and almost certainly never will be my job.

If I had the personality type who was more disciplined and was a time
management expert, then yes, downloading a bot would be a good idea.

Paul

Loading...