Weird JF moves.
Murat Kalinyaprak
won 9 out of 10 matches of 25 points aginst JF level
7/1000. Since then I have played quite a number of 25
point matches (unfortunately haven't kept statistics
but I would claim that I won well over half of them).
One thing that kind of stands out in my mind is that
I did very often get a big lead like 14-18 to 0 (even
in the matches I lost) and not necessarily because of
the cube going very high. However I observed something
really strange when the cube was owned by JF at 32 and
it was the last roll before it lost (and could even get
gammoned if he didn't bear off a piece). At that cube
value it doesn't matter, but JF seems to deliberately
not bear off a piece during his last move when it can.
For example, look at this position:
---------
OOOOOO|O cube=32
OOOOOO|
O |
O |
|
|
X |
---------
JF rolled 54 and played 7/2 6/2...! Oh, BTW, only
JF 3.5 does this, JF 3.0 plays 7/2 4/0.
Here is another example:
---------
OOOOO| cube=32
OOOOO|
OO |
OO |
O |
|
|
X |
---------
JF rolled 54 here also and played 6/1 6/2...! In the
same position JF 3.0 bears off 2 pieces.
If some of you would remember, I prefer to report
match results indicating the actual points as 34-8,
64-17, etc. instead of rounding down all such scores
to match length like 25-17 in a 25 point match. It
almost looks like JF is trying to please me regarding
my preference about this... LOL.. :)
The first couple of times when I noticed this, I
thought it was weird but didn't make much out of
it thinking it must be a really rare case. But
the third time I wanted to see if the previous
version of JF would play the same way. It didn't.
Yet it had been claimed that nothing had changed
from version 3.0 to 3.5 other than a few changes
to the user interface, bug fixes, etc. So what
now...??? Why does JF 3.5 move differently than
JF 3.0...? I'm sure it has nothing to do with the
dice bug being fixed or anything of the sort but
I'll be curious to see how his highness Mr. Dahl
(or his feet-kissers?) will explain this one...?
If I sound a litter bitter here, it's because I am.
I had asked 4-5 questions each to Mr. Dahl and OASYA
team and never got answers to them. I don't mind not
getting answers from them because often my questions
imply an answer in case no other answer is offered.
What was/is really disappointing was/is that nobody
ever joined me in wanting answers to them. Even if we
assumed the absolute impossible case that nobody in
this group was bright enough to understand what those
questions or the answers to them could imply, someone
could still have said: "Hey Mr. Dahl, hey OASYA team,
this guy here keeps asking question; why don't you
give him some (preferably) straight answers?". Noone
did. Sad... And even sadder may be the fact that many
people came after me, scolding me for asking questions,
asking from me proof for what the questions may have
implied, etc. instead... Anyway, let's go back to our
other subject.
Notice that in JF 3.5 "light" version one can't take
moves back or save matches/games/positions. So, in
order to replay/verify the last move in JF 3.5 and to
replay it to see if JF 3.0 would move the same, I had
to remember/repeat all my 40 moves exactly the same,
which I was able to do on the first try... Why say
this at the risk of being taken as patting myself on
the back? Because it may be an indication of either
an extremely consistent play for a "humanling" or a
pretty darn good memory (which may help one see or
"imagine:)" patters, etc.) So, if some people have
hard time believing that I can beat JF by being a
very good player, they can maybe believe I can do so
by having "decoded" JF through observing. I guess I
can't help making such comments because I desire some
common credibility even if I don't need it...
I'll append the mat file at the end of this post so
that others can check whether this happens in their
JF 3.5 light. I suppose the full version would allow
just editing a position and rolling 54 manually to
see if it would move the same way also.
Oh, one more self-patting on the back before I stop:).
I don't know if JF makes such moves every time it is
in a similar position. I would think that it doesn't.
So, in order for me to notice 4 (maybe 5) occurrences
of this oddity, I must have put JF in positions where
it was on the brink of being gammoned while it owned
the cube at 32 much more often than just 4-5 times...
Just begging:) for some little credibility that I may
indeed be whipping JF's butt pretty good here...
While at it, I should probably take this opportunity
to make a comment about the cube also. Now, 25 point
games are a little different than the 15 point ones
I used to play aginst JF, where I would at times
backgammon it with the cube at 32, because it would
turn the cube to 16 anyway in desperate cases. In 25
point matches it does it when 16 points isn't enough
for either side to reach 25 points. So, my conclusion
is that JF wrongly doubles (i.e. doesn't know how to
use the cube even as much as I do:) and loses 1x, 2x
and even 3x either after doubling or taking a redouble.
Ah, there is another possibility, of course, and that
is the cube and the skill that is supposedly needed to
use it are useless...
Anyway, would be nice to hear if anybody else had
noticed any similar oddities in JF's moves or just
feedback on the subject.
MK
Here is the mat file:
--------------------------------------------------
25 point match
Game 1
JellyFish : 0 JellyFish : 0
1) 53: 8/3 6/3
2) 64: 8/2 6/2 11: 8/7 8/7 6/5 6/5
3) 21: 24/21 Doubles => 2
4) Takes 63: 13/4*
5) 43: 25/21* 24/21 63: 25/22 13/7
6) 32: 13/8 51: 22/16
7) 62: 13/5 63: 16/10 13/10
8) 33: 13/10 8/5 6/3 6/3 63: 24/15*
9) 62: 25/17 44: 15/7 13/9 13/9
10) 41: 17/12 53: 7/2 4/21
11) 32: 13/10 12/10 64: 9/3 9/5
12) 33: 10/4* 10/4 52:
13) Doubles => 4 Takes
14) 32: 8/5 8/6 52:
15) 51: 21/16 13/12 42:
16) 31: 16/13 12/11 31: 25/24 7/4*
17) 66: Doubles => 8
18) Takes 22: 10/4 5/3
19) 51: 25/24 6/1* 41: 25/24* 10/6
20) 54: 42: 6/2 3/1*
21) 31: 25/24* 43:
22) 31: 25/24 13/10 41: 25/24 7/3
23) Doubles => 16 Takes
24) 53: 10/5 11/8 21: 7/5 3/2
25) 64: 24/18 8/4 61: 24/18 2/1*
26) 55: Doubles => 32
27) Takes 61: 24/17
28) 65: 21: 3/1 17/16
29) 63: 62: 18/12 16/14
30) 41: 52: 14/7*
31) 63: 65: 12/6 7/2
32) 41: 42: 6/2 5/3
33) 53: 61: 6/0 6/5
34) 32: 61: 5/0 2/1
35) 61: 25/19 55: 5/0 5/0 4/0 4/0
36) 31: 32: 3/0 2/0
37) 22: 43: 3/0 3/0
38) 66: 25/7 19/13 65: 2/0 2/0
39) 33: 13/4 7/4 52: 1/0 1/0
40) 54: 5/0 4/0 66: 1/0
Wins 32 points and the match
------------------------------------------------
PS: Another Microsoft trash software Outlook express 5.0
really messed up the alignment of the file I inserted
above and I had to reformat things manually. I hope it's
still usable as a JF .mat file (I don't know how picky
JF is about spaces, line feeds, etc.)
David Montgomery
<At that cube
<value it doesn't matter, but JF seems to deliberately
<not bear off a piece during his last move when it can.
<
<OOOOOO|
< O |
< O |
< |
< |
< X |
<---------
<
<JF rolled 54 and played 7/2 6/2...! Oh, BTW, only
<JF 3.5 does this, JF 3.0 plays 7/2 4/0.
<
< OOOOO|
< OO |
< OO |
< O |
< |
< |
<X |
<---------
<
<JF rolled 54 here also and played 6/1 6/2...! In the
<same position JF 3.0 bears off 2 pieces.
Probably JF generates a list of moves by considering
moving the backmost checkers first. This is a natural
way to do things. Then as it goes along generating
moves it saves the best move found so far. Here, no
better move after the first one is found, so
the first move generated -- the one moving the
backmost checkers -- is the one selected.
This appears to be true for both version 3.0 and 3.5
(with 3.0, look at the best moves list). Apparently
3.0 has some additional logic which makes it bear
checkers off in these situations. Perhaps it does
this when the evaluations are equivalent.
The important thing is to realize that 3.0 and 3.5's
moves, although different, are both correct and
completely equivalent.
David Montgomery
mo...@cs.umd.edu
monty on FIBS and GG
Rodrigo Andrade
OOOOOO|O cube=32
OOOOOO|
O |
O |
|
|
X |
---------
JF rolled 54 and played 7/2 6/2...! Oh, BTW, only
JF 3.5 does this, JF 3.0 plays 7/2 4/0.
Here is another example:
---------
OOOOO| cube=32
OOOOO|
OO |
OO |
O |
|
|
X |
---------
JF rolled 54 here also and played 6/1 6/2...! In the
same position JF 3.0 bears off 2 pieces.
**********************************************************************
You say you're not jacking up the cube here, but you are. Holy cow, you're
deciding a 25-point match in 1 game!!!!! That's pretty much the same as
playing a 1-pointer, which is your strength, right?
As for the above positions: regardless of the matchscore and given the cube
value, O will lose the game and the match no matter what it plays.
--
RODRIGO
===========================================================
"All religions of a spiritual nature are inventions of man. He has
created an entire system of gods with nothing more than his carnal brain.
Just because he has an ego and cannot accept it, he has had to externalize
it into some great spiritual device he calls 'God.'"
- The Satanic Bible
Anton Szandor LaVey
Murat Kalinyaprak
>In <7gbgme$e5...@taisp3.in-tch.com> Murat Kalinyaprak writes:
><At that cube
><value it doesn't matter, but JF seems to deliberately
><not bear off a piece during his last move when it can.
><OOOOOO|O cube=32
><OOOOOO|
>< O |
>< O |
>< |
>< |
>< X |
><---------
><JF rolled 54 and played 7/2 6/2...! Oh, BTW, only
><JF 3.5 does this, JF 3.0 plays 7/2 4/0.
>< OOOOO| cube=32
>< OOOOO|
>< OO |
>< OO |
>< O |
>< |
>< |
><X |
><---------
><JF rolled 54 here also and played 6/1 6/2...! In the
><same position JF 3.0 bears off 2 pieces.
>Probably JF generates a list of moves by considering
>moving the backmost checkers first. This is a natural
>way to do things.
Oh...?
>Then as it goes along generating
>moves it saves the best move found so far. Here, no
>better move after the first one is found, so
>the first move generated -- the one moving the
>backmost checkers -- is the one selected.
I appreciate your responding but are you serious...?
The stage of moving checkers is over, now it's bear
off time...! No move can be said to be better than
bearing off a piece when possible in such positions.
Please ridicule me and explain why *even the first*
move 6/1 is better than 5/0...?
>This appears to be true for both version 3.0 and 3.5
>(with 3.0, look at the best moves list). Apparently
>3.0 has some additional logic which makes it bear
>checkers off in these situations.
Yes, I guess that much was apparent to me also... :)
>Perhaps it does
>this when the evaluations are equivalent.
>The important thing is to realize that 3.0 and 3.5's
>moves, although different, are both correct and
>completely equivalent.
Heh, heh... No, no, no...! The important thing is
the question "Why do 3.0 and 3.5 play differently
when there has been no further neural-net training?"
What necessitated this additional or subtractional:)
logic in JF 3.5...?
Come on, guys, stop burying your heads in the sand.
It's always good/interesting to ask questions as one
never knows where the answers may end up leading us.
And, of course, it's always better to get an answer
from the horse's mouth than speculating (as I often
do:)...
MK
Murat Kalinyaprak
>Here is another example:
>---------
> OOOOO| cube=32
>You say you're not jacking up the cube here, but you are.
I guess we'll have to define what "jacking up the cube"
really means someday...
>Holy cow, you're deciding a 25-point match in 1 game!!!!!
No, I'm not deciding anything by myself!!! It takes two
to tango and one player alone can't turn the cube from
1 to 32. Why is JF not jacking up the cube when it gives
it to me at 16...? Why shouldn't I shove it back to JF
if I'm thinking that I'm the one who has the advantage...?
How JF, you, me or some world-class players determine who
has the advantage at any stage in a match may be different
but I decided that I had it in this case and did win. This
is how the cookie crumbles and that's all there is to it.
If the result doesn't mean anything, then what is any use
of all that blabbering in this forum about the cube skill?
If the results should mean something, we must conclude
that at least within the scope of those ten 25-pt matches
JF didn't do as well with the cube as a middle-easterner
who learned (err, actually hasn't even learned yet:) how
to use it less than a year ago...
>That's pretty much the same as playing a 1-pointer, which
>is your strength, right?
Well, not really because backgammons don't count in
1-pointers but you're right to the extent that they
don't involve using the cube. Even so, what you are
saying sounds like accepting my tentative argument
that the cube can be trivialized/"eliminated" from
the game by doing as little as simply ignoring it
(i.e. keep taking and/or doubling back, which BTW
wasn't quite my approach in those ten 25-pt matches
against JF) and focusing on one's checker play (i.e.
the real game skill)...
>As for the above positions: regardless of the matchscore
>and given the cube value, O will lose the game and the
>match no matter what it plays.
Of course, but the issue I was raising was with the
fact that JF 3.0 and 3.5 were playing certain positions
differently (despite claims that changes from one to
the other revision were nothing but cosmetic)...! And
no, the idea that this could be related to the dice bug
and how it was fixed didn't even cross my mind... :) :)
MK
Daniel Murphy
<mu...@compuplus.net> wrote:
>David Montgomery wrote in news:7gcicu$f...@krackle.cs.umd.edu...
>
>>Perhaps it does this when the evaluations are equivalent.
>>The important thing is to realize that 3.0 and 3.5's
>>moves, although different, are both correct and
>>completely equivalent.
>
>Heh, heh... No, no, no...! The important thing is
>the question "Why do 3.0 and 3.5 play differently
>when there has been no further neural-net training?"
>What necessitated this additional or subtractional:)
>logic in JF 3.5...?
>
> Come on, guys, stop burying your heads in the sand.
> It's always good/interesting to ask questions as one
> never knows where the answers may end up leading us.
It's also good -- when one has a question -- to state it compactly,
completely and clearly visible somewhere above, not under, a mound of
extemporaneous verbiage.
Your question is: In a position where all possible moves give
JellyFish 0.00 match equity, why do JF 3.0 and 3.5 pick different
moves?
That's an interesting albeit minor question: interesting, because
Effect Software has said that JF 3.5 and 3.0 have the same "engine";
minor, because as David points out "3.0 and 3.5's moves, although
Donald Kahn
>David Montgomery wrote in news:7gcicu$f...@krackle.cs.umd.edu...
>
>>In <7gbgme$e5...@taisp3.in-tch.com> Murat Kalinyaprak writes:
>
>><At that cube
>><value it doesn't matter, but JF seems to deliberately
>><not bear off a piece during his last move when it can.
>
>><OOOOOO|O cube=32
>><OOOOOO|
>>< O |
>>< O |
>>< |
>>< |
>>< X |
>><---------
>
>><JF rolled 54 and played 7/2 6/2...! Oh, BTW, only
>><JF 3.5 does this, JF 3.0 plays 7/2 4/0.
>
>>< OOOOO| cube=32
>>< OOOOO|
>>< OO |
>>< OO |
>>< O |
>>< |
>>< |
>><X |
>><---------
>
>><JF rolled 54 here also and played 6/1 6/2...! In the
>><same position JF 3.0 bears off 2 pieces.
>
>>Probably JF generates a list of moves by considering
>>moving the backmost checkers first. This is a natural
>>way to do things.
>
>Oh...?
>
>>Then as it goes along generating
>>moves it saves the best move found so far. Here, no
>>better move after the first one is found, so
>>the first move generated -- the one moving the
>>backmost checkers -- is the one selected.
>
>I appreciate your responding but are you serious...?
>The stage of moving checkers is over, now it's bear
>off time...! No move can be said to be better than
> bearing off a piece when possible in such positions.
>Please ridicule me and explain why *even the first*
>move 6/1 is better than 5/0...?
>
>>This appears to be true for both version 3.0 and 3.5
>>(with 3.0, look at the best moves list). Apparently
>>3.0 has some additional logic which makes it bear
>>checkers off in these situations.
>
>Yes, I guess that much was apparent to me also... :)
>
>>Perhaps it does
>>this when the evaluations are equivalent.
>>The important thing is to realize that 3.0 and 3.5's
>>moves, although different, are both correct and
>>completely equivalent.
>
>Heh, heh... No, no, no...! The important thing is
>the question "Why do 3.0 and 3.5 play differently
>when there has been no further neural-net training?"
>What necessitated this additional or subtractional:)
>logic in JF 3.5...?
>
> Come on, guys, stop burying your heads in the sand.
> It's always good/interesting to ask questions as one
> never knows where the answers may end up leading us.
>And, of course, it's always better to get an answer
>from the horse's mouth than speculating (as I often
>do:)...
>
>MK
>
>
In my JF 3.0. you can produce moves like 7/2 6/2 in the first diagram
by setting the level at 1 or 2. Changing the seed and counter also
seems to make a difference, though I don't see why it should.
Ian Shaw
Donald Kahn wrote in message <373aa6ca....@news.newsguy.com>...
>>><value it doesn't matter, but JF seems to deliberately
>>><not bear off a piece during his last move when it can.
>>
>>><OOOOOO|
>>>< O |
>>>< O |
>>>< |
>>>< |
>>>< X |
>>><---------
>>
>>><JF rolled 54 and played 7/2 6/2...! Oh, BTW, only
>>><JF 3.5 does this, JF 3.0 plays 7/2 4/0.
>>
>>
>In my JF 3.0. you can produce moves like 7/2 6/2 in the first diagram
>by setting the level at 1 or 2. Changing the seed and counter also
>seems to make a difference, though I don't see why it should.
Oh dear, oh dear! Those of us who believe that JF doesn't cheat have always
used as a cornerstone of our argument that the value of seed and counter
make no difference to the play selected. Now you've found a counter example.
Is it possible that at low levels JF plays less well by randomly picking
from a number of candidate moves?
--
Regards
Ian Shaw (ian on FIBS)
Daniel Murphy
25-point matches and has kindly shared his lessons with us r.g.bg
readers. But Rodrigo Andrade impudently told Mr. K:
>You say you're not jacking up the cube here, but you are.
To which Murat replied:
>I guess we'll have to define what "jacking up the cube" really means someday...
I offer the following definition, together with some illustrating
examples provided by Murat.
Jacking up the cube:
(1) Intentionally making incorrect cube decisions
in order to raise the cube to higher levels
than are justified by the actual equity and
volatility of a position, e.g., doubling
positions that are not good enough to double,
or taking doubles which should be dropped.
(2) Unintentionally making incorrect cube decisions
which raise the cube to higher levels than are
justified by the actual equity and volatility
of a position, e.g., "he thought he was a huge
favorite when he doubled, but he was only
jacking up the cube."
(3) In a more general sense, raising the value of
the cube by doubling.
The following discussion pertains primarily to definition (1).
Jacking up the cube is a hallmark of players who see backgammon
primarily as an opportunity to gamble and who enjoy the excitement of
large cubes and large swings on the scoresheet. Jacking up the cube is
also a hallmark of the 'steamer' who, losing badly, wants to 'win it
all back in one game' -- with the usual result of compounding his
losses. Such players, provided they pay when they lose, are generally
welcome opponents, and are sometimes referred to as 'fish' or, in more
genteel parlance, 'friendly players.'
Assuming good play by both players, jacking up the cube is a costly
error. The effect of incorrect cube actions is to offer or agree to
disadvantageous odds and/or cube position for the remainder of the
game or match in progress, resulting in a loss of equity in comparison
with correct cube action.
Given extreme differences in skill between two players, however,
jacking up the cube may be justified. A much weaker player may, for
example, choose to accept disadvantageous odds in a single position to
compensate for poor long-run chances of winning.
An example of justifiable 'cubejacking' by a much weaker player:
24 23 22 21 20 19
|----------------- Woody Worldbeater -0
| 0
| 0
| 7-point match
| X
| X
|----------------- Barney Beginner - 0
1 2 3 4 5 6 Barney holds cube on 4
In the position above X is not a favorite to win the game but should
nevertheless redouble to 8 and win the 7-point match 17/36 of the time
if X's chances of winning the match after not doubling -- and either
leading 4-0 or trailing 0-4 against his expert opponent -- are less
than 47.22%.
A much stronger player may justifiably take or double 'incorrectly' if
opponent's subsequent misplays and erroneous cube (in)actions can be
expected to give sufficient compensation for the stronger player's
intrinsically inaccurate cube action.
However, if the stronger player's expectations are accurate, such
'incorrect' cube decisions will in fact be correct, and should
therefore not be considered to be jacking up the cube. And done with
appropriate gusto, such 'incorrect' actions may even encourage one's
opponent to indulge his own penchant for cubejacking.
The following 25-point match between the backgammon program JellyFish
3.0 and Murat, a merely human player, includes several instances of
jacking up the cube -- none committed by the computer.
------------------------------------------------
25 point match
Game 1
JellyFish (O): 0 Murat (X): 0
1) 53: 8/3 6/3
2) 64: 8/2 6/2 11: 8/7 8/7 6/5 6/5
3) 21: 24/21 Cube action??
Should Murat (X) double? Should JellyFish (O) take?
+24-23-22-21-20-19-+---+18-17-16-15-14-13-+
| X O O | | O X |
| X O O | | O X |
| O | | X |
| O | | X |
| | | X |
| | | | Cube: 64
| | | |
| | | O |
| | | O |
| | | O |
| X X X | | X O |
| O X O X X | | X O |
+-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
3) ... Doubles => 2
JF L7 Evaluation: Wins Gammons Backgammons
JellyFish 39.0% 9.4% 0.2%
Murat 61.0% 18.0% 0.5%
Equity Murat: 0.308
Volatility: 0.174
Cube action: No double / Take
1296 L6 rollouts: Wins Gammons Backgammons
JellyFish 40.7% 10.3% 0.5%
Murat 59.3% 18.5% 0.8%
Equity Murat: 0.270
Standard deviation: 0.008
The first example of "jacking up the cube." With this optimistic
double, X apparently plans to roll 33 -- and prays 0 fans.
4) Takes 63: 13/4*
5) 43: 25/21* 24/21 63: 25/22 13/7
6) 32: 13/8 51: 22/16
7) 62: 13/5 63: 16/10 13/10
8) 33: 13/10 8/5 6/3 6/3 63: 24/15*
9) 62: 25/17 44: 15/7 13/9 13/9
10) 41: 17/12 53: 7/2 24/21
Better is 24/16. Murat's should try to bring the last checker all the
way home.
11) 32: 13/10 12/10 64: 9/3 9/5
12) 33: 10/4* 10/4 52: can't move
13) Cube action??
Should JellyFish (O) double? Should Murat (X) take?
+24-23-22-21-20-19-+---+18-17-16-15-14-13-+
| O O O O O | | O | Cube: 2
| O O O O O | | O |
| | | |
| | | |
| | | |
| | X | |
| | | |
| | | |
| | | |
| X X | | X |
| X O X X | | X X |
| X X O X X | | X X O |
+-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
13) Doubles => 4 Takes
JF L7 Evaluation: Wins Gammons Backgammons
JellyFish 78.4% 17.7% 0.4%
Murat 21.6% 3.3% 0.0%
Equity JellyFish: 0.716
Volatility: 0.250
Cube action: Double / Drop
1296 L6 rollouts: Wins Gammons Backgammons
JellyFish 82.5% 14.3% 0.2%
Murat 17.5% 3.1% 0.1%
Equity Murat: 0.763
Standard deviation: 0.006
Second example of "jacking up the cube." On the bar against JF's 5
point board, Murat should drop, but if he doesn't take here, how will
the cube ever get to 32?
14) 32: 8/5 8/6 52: can't move
15) 51: 21/16 13/12 42: can't move
16) 31: 16/13 12/11 31: ??
Murat (X) to play 31
+24-23-22-21-20-19-+---+18-17-16-15-14-13-+
| O O O O O | | O |
| O O O O O | | |
| O O | | |
| | | |
| | | |
| | X | |
| | | |
| | | |
| | | |
| X X | | X |
| X X X | | X X |
| X X O X X | | X X O | Cube: 4
+-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
16) ... 31: 25/24 7/4*
25/24 7/4* is a blunder. 25/24 5/2 leaves only 1 blot, gives JellyFish
5s and 4s to hit or escape, and leaves Murat with a reasonable
position if JellyFish doesn't hit or escape. The actual play leaves
three blots and 23 ways for JellyFish to immediately put a 2nd checker
behind a 5-prime and put Murat on the gammon trail.
After 25/24 7/4*:
JF L7 Evaluation: Wins Gammons Backgammons
JellyFish 72.9% 38.2% 1.8%
Murat 27.1% 6.0% 0.2%
Equity JellyFish: 0.797
After 25/24 5/2:
JF L7 Evaluation: Wins Gammons Backgammons
JellyFish 68.7% 12.0% 0.3%
Murat 31.3% 4.4% 0.2%
Equity JellyFish: 0.451
17) 66: can't move Cube action??
JellyFish misses the blots!
Should Murat (X) double? Should JellyFish (O) take?
+24-23-22-21-20-19-+---+18-17-16-15-14-13-+
| X O O O O O | | O |
| O O O O O | | |
| O O | | |
| | | |
| | | |
| | O | |
| | | |
| | | |
| | | |
| X X | | |
| X X X | | X X |
| X X X X X | | X X O | Cube: 4
+-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
17) ... Doubles => 8
Third example of "jacking up the cube." Ever the optimist, Murat's
well on the way to a 32-cube.
JF L7 Evaluation: Wins Gammons Backgammons
JellyFish 44.3% 12.9% 0.6%
Murat 55.7% 16.8% 0.7%
Equity Murat: 0.154
Volatility: 0.730
Cube action: No double / take
1296 L6 rollouts: Wins Gammons Backgammons
JellyFish 43.9% 11.4% 0.5%
Murat 56.1% 18.8% 0.7%
Equity Murat: 0.197
Standard deviation: 0.007
18) Takes 22: 10/4 5/3
10/4 5/3, making a 4 point board and leaving one blot, is a blunder.
Clearly best is 5/1 4/2 3/1, which makes a 5 point board and leaves no
blots.
19) 51: 25/24 6/1* 41: 25/24* 10/6
20) 54: can't move 42: 6/2 3/1*
21) 31: 25/24* 43: can't move
22) 31: 25/24 13/10 41: 25/24 7/3
23) Cube action??
Should JellyFish (O) double? Should Murat (X) take?
+24-23-22-21-20-19-+---+18-17-16-15-14-13-+
| X O O O O O | | O O | Cube: 8
| X O O O O O | | |
| O | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| X | | |
| X | | |
| O X X X X X | | |
| O X X X X X | | X |
+-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
23) Doubles => 16 Takes
A powerful double. JF has a great, gammonish game with an immediate 6,
in many sequences of no-6 by JF and no-6 by Murat, and even some
sequences of no-6 by JF and a 6 by Murat. Murat correctly takes.
JF L7 Evaluation: : Wins Gammons Backgammons
JellyFish 68.6% 22.6% 1.2%
Murat 31.4% 7.4% 0.3%
Equity JellyFish: 0.533
Volatility: 0.443
Cube action: Double / take
1296 L6 rollouts: Wins Gammons Backgammons
JellyFish 70.8% 22.2% 1.7%
Murat 29.2% 6.0% 0.3%
Equity JellyFish: 0.592
Standard deviation: 0.007
JF does better in the cubeless rollouts, but after seeing the rollout
results JF still thinks Murat should take.
24) 53: 10/5 11/8 21: 7/5 3/2
25) 64: 24/18 8/4 61: 24/18 2/1*
26) 55: can't move Cube action??
What excitement! With 61 Murat escapes one back checker and hits, and
JF fans! And that cube is itching to be turned to 32 ...
Should Murat (X) double? Should JellyFish (O) take?
+24-23-22-21-20-19-+---+18-17-16-15-14-13-+
| X O O O O O | | X |
| O O O O O | | |
| O O | | |
| O | | |
| | | |
| | O | |
| | | |
| | | |
| | | |
| X X | | |
| X X X X X | | |
| X X X X X X | | O | Cube: 16
+-1--2--3--4--5--6-+---+-7--8--9-10-11-12-+
26) ... Doubles => 32
JF L7 Evaluation: Wins Gammons Backgammons
JellyFish 26.0% 6.3% 0.3%
Murat 74.0% 22.1% 0.7%
Equity Murat: 0.638
Volatility: 0.480
Cube action: Double / Take
1296 L6 rollouts: Wins Gammons Backgammons
JellyFish 25.4% 5.0% 0.2%
Murat 74.6% 24.1% 0.6%
Equity Murat: 0.688
Standard deviation: 0.007
27) Takes 61: 24/17
28) 65: can't move 21: 3/1 17/16
29) 63: can't move 62: 18/12 16/14
30) 41: can't move 52: 14/7*
31) 63: can't move 65: 12/6 7/2
32) 41: can't move 42: 6/2 5/3
33) 53: can't move 61: 6/0 6/5
34) 32: can't move 61: 5/0 2/1
35) 61: 25/19 55: 5/0 5/0 4/0 4/0
36) 31: can't move 32: 3/0 2/0
37) 22: can't move 43: 3/0 3/0
38) 66: 25/7 19/13 65: 2/0 2/0
39) 33: 13/4 7/4 52: 1/0 1/0
40) 54: 5/0 4/0 66: 1/0
Wins 32 points and the match
------------------------------------------------
The inescapable conclusion is how easy it is to beat JellyFish or
another expert player:
Jack up the cube and get lucky!
Double a no-double position, take a drop, redouble to 8 in a no-double
position, recover from a blunder or two along the way, take a 16 cube
with 29% winning chances, recube to 32 and voilá -- another "proof" of
the unimportance of doubling cube strategy in modern backgammon.
David desJardins
> Jacking up the cube: (1) Intentionally making incorrect cube decisions
> in order to raise the cube to higher levels than are justified by the
> actual equity and volatility of a position, e.g., doubling positions
> that are not good enough to double, or taking doubles which should be
> The following discussion pertains primarily to definition (1) ....
> Given extreme differences in skill between two players, however,
> jacking up the cube may be justified.
But in that case, the cube decisions are not in fact "incorrect".
It seems to me that Murat has a single valid point: Jellyfish doesn't
realize that it's a much stronger player than Murat is. So it doesn't
take skill differences into account when doubling. If Murat does, then
he can achieve a higher win percentage (especially in long matches) than
his relative strength would otherwise predict. Although of course much
below 50%. There's nothing wrong with such a strategy.
David desJardins
David desJardins
> Do you agree that to improve results by adopting a theoretically
> suboptimal strategy requires (1) the other player have exploitable
> weaknesses and (2) one is able to identify them?
No. Your opponent doesn't have to have any weaknesses in order for you
to improve your results against it by adopting a "theoretically
suboptimal" strategy. It only has to be the case that adopting your
"theoretically suboptimal" strategy makes the other weaknesses of your
game less significant.
Suppose I challenge you to the following game. Either we play a 5-point
backgammon match, winner take all, or you can choose instead to just
roll a single die, and if you roll 1 or 2 you win otherwise you lose.
The "theoretically optimal" strategy for this game is to choose to play
the backgammon match, which (being a fair game) gives you a theoretical
50% chance. However, if you are a terrible backgammon player and I'm a
great one, you might in practice win the match only 25% of the time. So
you would be better off to choose the 33% chance of winning with the die
roll. This choice on your part, and its execution, doesn't require any
"exploitable weaknesses" on my part at all.
> But what about Murat vs. JellyFish? What *abnormal* strategy should
> Murat adopt that would enable him to do better than "his relative
> strength would otherwise predict"?
Taking doubles that a player just as strong as Jellyfish would drop, is
a good example. Since his personal match equity table against JF is
different than the theoretical match equity table between equal players,
his correct double and take points will be different. He can also
double JF and get JF to take in positions where it would be correct for
JF to take against an equal opponent, but where JF would do better to
drop if JF were taking into account Murat's actual strength.
> Can he take "drops"? Can he drop "takes"? -- in effect, adjusting his
> own take window to account for his own weaknesses?
Yes. (Mostly the former---by taking positions that should theoretically
be drops, he reduces his expectation but increases his variance, which
is good for the weaker player. I don't know if there are any positions
which are theoretical takes but which a weaker player should instead
drop. I don't think so.)
> If he could make that adjustment, would that improve his results?
Yes.
> And is he skillful enough to identify the positions in which he should
> make that adjustment? Or is it possible to make that adjustment
> without knowing when he should?
I think it's possible to make the practical observation that taking more
and doubling more than theoretical calculations would predict will
improve your match winning chances, and then to implement that
observation, without having to have a detailed understanding of the
theory. Of course, it's also possible to overdo it.
David desJardins
Daniel Murphy
<de...@math.berkeley.edu> wrote:
>Daniel Murphy <rac...@cityraccoon.com> writes:
>> Given extreme differences in skill between two players, however,
>> jacking up the cube may be justified.
>
>But in that case, the cube decisions are not in fact "incorrect".
You must be right, because I said so myself :)
>It seems to me that Murat has a single valid point: Jellyfish doesn't
>realize that it's a much stronger player than Murat is. So it doesn't
>take skill differences into account when doubling. If Murat does, then
>he can achieve a higher win percentage (especially in long matches) than
>his relative strength would otherwise predict. Although of course much
>below 50%. There's nothing wrong with such a strategy.
It's not clear to me, David, exactly what that strategy would be in
the specific case of Murat vs. JellyFish, and I'd be interested to
hear more.
Do you agree that to improve results by adopting a theoretically
suboptimal strategy requires (1) the other player have exploitable
weaknesses and (2) one is able to identify them?
A stronger player can exploit the weaknesses of a weaker player by:
Doubling "early" inducing an incorrect drop.
Doubling "late" inducing incorrect takes.
Adjusting doubling windows to reflect actual game and match equity
based on opponent's capacity for play and cube errors.
If JF could adjust its strategy when playing Murat, JF would do even
better than it does.
A weaker play can in theory exploit the weaknesses of a stronger
player IF the stronger player is weak enough to have identifiable play
and/or cube errors AND the weaker player is strong enough to
capitalize on them. An example is the ease with which weaker players
could in theory "steal points" if only they'd toss the cube more often
in scary gammonish positions against stronger players who aren't quite
as strong as they think they are! I would claim, though, that *most*
"stronger" players who adjust their normal cube strategy to account
for perceived superiority over their weaker opponent are more likely
to adjust wrongly than rightly.
But what about Murat vs. JellyFish? What *abnormal* strategy should
Murat adopt that would enable him to do better than "his relative
strength would otherwise predict"?
He can't double late -- JellyFish will drop correctly. Can he refrain
from doubling in positions which are theortical doubles but aren't for
Murat because he misplays them? I suppose, but then his results should
be *what* his relative strength would predict, not better, no?
He can't double early -- a take is a take is a take, and doubling
early must cost equity.
Can he take "drops"? Can he drop "takes"? -- in effect, adjusting his
own take window to account for his own weaknesses? If he could make
that adjustment, would that improve his results? And is he skillful
Robert-Jan Veldhuizen
Murphy) wrote:
>Murat Kalinyaprak has been teaching JellyFish how to use the cube in
>25-point matches and has kindly shared his lessons with us r.g.bg
>readers. But Rodrigo Andrade impudently told Mr. K:
>
>>You say you're not jacking up the cube here, but you are.
>
>To which Murat replied:
>
>>I guess we'll have to define what "jacking up the cube" really means someday...
>
>I offer the following definition, together with some illustrating
>examples provided by Murat.
Thanks for the educational humor or should I write humorous education?
Isn't it great to learn while you laugh? :-)
--
Robert-Jan/Zorba
** New email address: R.J.Vel...@cable.A2000.nl **
Murat Kalinyaprak
>On Wed, 12 May 1999 "Murat Kalinyaprak" wrote:
>>David Montgomery wrote in news:7gcicu$f...@krackle.cs.umd.edu...
>>>The important thing is to realize that 3.0 and 3.5's
>>>moves, although different, are both correct and
>>>completely equivalent.
>>Heh, heh... No, no, no...! The important thing is
>>the question "Why do 3.0 and 3.5 play differently
>>when there has been no further neural-net training?"
>>What necessitated this additional or subtractional:)
>>logic in JF 3.5...?
>It's also good -- when one has a question -- to state it
>compactly, completely and clearly visible somewhere above,
>not under, a mound of extemporaneous verbiage.
I may have to agree with the "mound of extemporaneous
verbiage" remark. Perhaps I tend to do it for fear
that the same content in more condensed words could
present a challenge to some dense heads...
>Your question is: In a position where all possible moves
>give JellyFish 0.00 match equity, why do JF 3.0 and 3.5
>pick different moves?
Why is all your "mound of extemporaneous verbiage" about
match equity, bla, bla...? The question could be further
shortened to: "Why JF 3.0 and JF 3.5 move differently in
identical circumstances"...?
>That's an interesting albeit minor question: interesting,
>because Effect Software has said that JF 3.5 and 3.0 have
>the same "engine"; minor, because as David points out "3.0
>and 3.5's moves, although different, are both correct and
>completely equivalent."
Let me ask you all: are you guys a bunch of mental and/or
social wimps...? Why are you trying to trivialize something
before even you know what it is...? Why don't you guys have
any guts to go after answers to such questions from F. Dahl,
Oasya, etc...? Are you afraid of them, of government, of
Jesus, of God, etc...? Or are you afraid that you may not
like the answers you may be given...?
Let's first hear the stupid answer and then we can further
express opinions as to whether we think it's a minor or a
major thing. How about it folks...? Would all the smart and
sane people in this newsgroup please have the guts to repeat
after me: "Let's hear F. Dahl's answer to this question"...!
MK
Murat Kalinyaprak
David desJardins wrote...
>Daniel Murphy writes:
>> Jacking up the cube: (1) Intentionally making incorrect cube
>> decisions in order to raise the cube to higher levels than are
>> justified by the actual equity and volatility of a position...
>> Given extreme differences in skill between two players, however,
>> jacking up the cube may be justified.
>But in that case, the cube decisions are not in fact "incorrect".
>It seems to me that Murat has a single valid point: Jellyfish
>doesn't realize that it's a much stronger player than Murat is.
What if JF is smarter than you and it realizes that Murat
is a stronger player than itself while you are not smart
enough to realize the same...? Couldn't JF admirers give
JF this much credit when the time seems so right...? :)
>So it doesn't take skill differences into account when doubling.
>If Murat does, then he can achieve a higher win percentage
>(especially in long matches) than his relative strength would
>otherwise predict. Although of course much below 50%. There's
>nothing wrong with such a strategy.
Are you guys' above comments jokes by chance or written when you
were drunk or something...?
MK
Murat Kalinyaprak
>David desJardins wrote:
>Do you agree that to improve results by adopting a theoretically
>suboptimal strategy requires (1) the other player have exploitable
>weaknesses and (2) one is able to identify them?
I'm rather pleased to see some interesting comments/questions
but can't understand why the persistence in calling any such
strategies still "suboptimal"... How does inserting the term
"theoretically" make a strategy suboptimal to a another one
which produces even lesser results...?
>But what about Murat vs. JellyFish? What *abnormal* strategy should
>Murat adopt that would enable him to do better than "his relative
>strength would otherwise predict"?
>He can't double late -- JellyFish will drop correctly....
>He can't double early -- a take is a take is a take....
I agree with these sane remarks of Daniel...
>Can he take "drops"? Can he drop "takes"? -- in effect, adjusting
>his own take window to account for his own weaknesses? If he could
>make that adjustment, would that improve his results? And is he
>skillful enough to identify the positions in which he should make
>that adjustment?
Obviously I don't know enough to even begin discussing on
these questions but would be very intereting to see them
being discussed among the cube-masters of this newsgroup.
To me, in line with my previous comments about what I
think of the cube, the questions don't even make any sense
to begin with... :(
>Or is it possible to make that adjustment without knowing
>when he should?
Aha... I like this one... :) "Without knowing"...! I sure
won't mind beating JF without knowing how to use the cube
or even without knowing how to move my checkers around for
the rest of my life... :)
MK
Murat Kalinyaprak
>Daniel Murphy writes:
>> Do you agree that to improve results by adopting a theoretically
>> suboptimal strategy requires (1) the other player have exploitable
>> weaknesses and (2) one is able to identify them?
>No. Your opponent doesn't have to have any weaknesses in order
>for you to improve your results against it by adopting a
>"theoretically suboptimal" strategy....
>Suppose I challenge you to the following game. Either we play
>a 5-point backgammon match, winner take all, or you can choose
>instead to just roll a single die, and if you roll 1 or 2 you
>win otherwise you lose. The "theoretically optimal" strategy
>for this game is to choose to play the backgammon match, which
>(being a fair game) gives you a theoretical 50% chance. However,
>if you are a terrible backgammon player and I'm a great one,
>you might in practice win the match only 25% of the time. So
>you would be better off to choose the 33% chance of winning with
>the die roll. This choice on your part, and its execution,
>doesn't require any "exploitable weaknesses" on my part at all.
I quoted the entire above paragraph so that people who
read it (perhaps a second time now) can see what kind
of irrelevant bullshit some writers are capable of in
this newsgroup of elite minds...
>> But what about Murat vs. JellyFish? What *abnormal* strategy
>> should Murat adopt that would enable him to do better than
>> "his relative strength would otherwise predict"?
>Taking doubles that a player just as strong as Jellyfish would
>drop, is a good example. Since his personal match equity table
>against JF is different than the theoretical match equity table
>between equal players, his correct double and take points will
>be different. He can also double JF and get JF to take in
>positions where it would be correct for JF to take against an
>equal opponent, but where JF would do better to drop if JF were
>taking into account Murat's actual strength.
You guys can't imagine how happy I am to hear such words
right out of your own mouths... I had advocated that if
the cube was used by either one or both players in a
pre-agreed manner (i.e. double after every 10 moves), the
player better with his checkers would win in the long run.
However, using the cube in such ways would be "improper".
And now we are hearing arguments that "improper" usage of
the cube can benefit a player if done (un-)knowingly and
consistently. Not only this, but we hear the argument that
such a startegy could benefit the weaker player. Well guys,
I hadn't laughed this good for a long time. Read David's
last sentence above a few times over if you need to, and
after that, if you are not laughing then maybe you should
be crying...
>I think it's possible to make the practical observation that
>taking more and doubling more than theoretical calculations
>would predict will improve your match winning chances, and
>then to implement that observation, without having to have a
>detailed understanding of the theory. Of course, it's also
>possible to overdo it.
Here we go again... Somebody does something right again
and again "without understanding", "without knowing"... :)
It appears that you folks invented some fancy cube rules
which you teach and learn, write books about, etc... And
as long as everybody "plays the game of doubling/dropping"
as they are supposed to do, everything self-validates. But
when somebody says "to hell with the cube" and starts using
it in a manner that a cube-master would deem haphazardly,
then the magic doesn't work anymore...
Here is where we start. You and I are equal strength players.
With fair/equal dice and without the cube each would win 50%
of the time. Now suppose I double after the first roll each
and every game and the cube is never used again in any of the
games. Each of us will still win 50% of the time because the
roll after the double will make it "right" or "wrong" 50% of
the time. Now extrapolate this to doubling at later/different
stages of the games and then try to figure out the compounded
effect of it if the limitation of using the cube only once
per game is lifted. You will find out that you won't be able
to figure things out unless you assume that subsequent cube
actions will be taken "correctly". (BTW, that's exactly why
you guys are so tempted to consider ridiculous possibilities
like somebody soing something right repeatedly without knowing
how to do it right:). Back to the previous thought; in other
words, looking ahead won't work unless players execute future
takes/drops at X%/Y% winning/losing probabilities like robots
or brainwashed monkeys... In yet other words, by defying what
is expected as far as cube action, one can to a great extent
render the cube irrelevant to the game, in which case players
with the "staying-power" (i.e. superior checker skills) will
consistently win... If you don't understand this it may be
due to my language limitations but if you do and if you are
expecially one of the people who had written books about cube,
you may consider going back to revise a few paragraphs...
MK
David Montgomery
>The question could be further
>shortened to: "Why JF 3.0 and JF 3.5 move differently in
>identical circumstances"...?
>
>social wimps...? Why are you trying to trivialize something
>before even you know what it is...? Why don't you guys have
>any guts to go after answers to such questions from F. Dahl,
>Oasya, etc...? Are you afraid of them, of government, of
>Jesus, of God, etc...? Or are you afraid that you may not
>like the answers you may be given...?
Doh! Because I don't care!
I don't care what moves JF makes when the moves don't matter.
I also haven't spent any time pondering why JF3.5 has a different
default color scheme, or why it's 3.5 and not 3.4 or 3.6.
You ask:
>"Why JF 3.0 and JF 3.5 move differently in identical circumstances"...?
But so far you have only given examples where IT DOESN'T MATTER.
Since IT DOESN'T MATTER, I don't care. Come back with an example
where it matters and maybe I'll care.
Actually, when it has mattered, I have written Fredrik, and I've
learned several undocumented things about JF that way. I've
found Fredrik very responsive. But these positions... who
(besides you) cares?
Murat Kalinyaprak
> What was/is really disappointing was/is that nobody ever
> joined me in wanting answers to them. Even if we assumed
> the absolute impossible case that nobody in this group
> was bright enough to understand what those questions or
> the answers to them could imply, someone could still have
> said: "Hey Mr. Dahl, hey OASYA team, this guy here keeps
> asking question; why don't you give him some (preferably)
> straight answers?".
And somebody with a familiar name in rgb sent
an email saying:
"I like seeing you not get answers and get frustrated.
"It's sort of amusing."
I guess when I had used words like "sad" in my previous
posting, I wasn't too far off. I had myself defined my
reaction to people like him using terms like bitter or
disappointed, but I'm not sure if I'm "frustrated" for
not receiving answers to my questions. Apparently this
guy isn't bright enough to have noticed that in my old
articles on these subjects it was either other people
who had detected implied answers and/or it was myself
who hinted at things and/or acknowledged other people's
such comments. The idea that my questions had "implied"
answers is clearly expressed even in the paragraph above
that he was replying but obviously he is imcapable of
reading or understanding what he reads.
Who should really be frustrated are people like this
guy who tries to butt heads with me without having what
it would take to do it or people who have to live with
my implied answers until they can come up with others
that they may like better. Whether they are absolutely
accurate or not, I'm fairly content/satisfied with what
I believe to know as answers to my own questions. Maybe
what confuses people like this guy is that for the sake
of being more proper (and giving you a chance to work
your own brains:) I don't speculate/opinate outright as
much as I used to do earlier...
I can understand why I didn't/don't/won't get answers
from JF's or SW's developers and I hope that at least
a few other readers can also see why not... Far from
frustrating me, their silence can only be taken as
proof of their inability to deny what may be implied
by those questions. For their own sake, what the
general readership of this newsgroup needs to realize
is that those questions were asked by a member of this
newsgroup on behalf of and for the benefit of its
entire membership. Their not responding to the group
can only indicate that in their eyes this group may
not deserve a shit's worth of an answer. And yes, this
group includes "you", "you" and "you" also...
Maybe you avoid feeling worthless by deceiving yourself
that you are not curious about such things and that you
don't care or need to know the answers, etc. But what
if you did want to know...? Do you think you could get
any better answers from them than what my questions may
have implied...? I bet you couldn't...! Go ahead prove
me wrong. Do what you need to do to get some better
answers. Send private emails, kiss their hands, kiss
their feet, kiss their ass or do whatever you may think
would help you with getting answers... And then come
post them in this newsgroup.
In the meantime, you may not like or agree with my
opinions/speculations/etc. but at least you don't need
to feel like you have to kiss my ass before you can
even ask me a question, let alone getting an answer.
Why can't you show the decency to acknowledge at least
this much if not to appreciate my efforts in trying to
make you think, look, see, etc...
MK
Robert-Jan Veldhuizen
<mu...@compuplus.net> wrote:
>Murat Kalinyaprak had written:
So when can we expect an interesting article of yours appearing here?
--
Robert-Jan/Zorba
Daniel Murphy
<mu...@compuplus.net> wrote:
>Here is where we start. You and I are equal strength players.
>With fair/equal dice and without the cube each would win 50%
>of the time. Now suppose I double after the first roll each
>and every game and the cube is never used again in any of the
>games. Each of us will still win 50% of the time because the
>roll after the double will make it "right" or "wrong" 50% of
>the time. Now extrapolate this to doubling at later/different
>stages of the games and then try to figure out the compounded
>effect of it if the limitation of using the cube only once
>per game is lifted. You will find out that you won't be able
>to figure things out unless you assume that subsequent cube
>actions will be taken "correctly". (BTW, that's exactly why
>you guys are so tempted to consider ridiculous possibilities
>like somebody soing something right repeatedly without knowing
>how to do it right:). Back to the previous thought; in other
>words, looking ahead won't work unless players execute future
>takes/drops at X%/Y% winning/losing probabilities like robots
>or brainwashed monkeys... In yet other words, by defying what
>is expected as far as cube action, one can to a great extent
>render the cube irrelevant to the game, in which case players
>with the "staying-power" (i.e. superior checker skills) will
>consistently win.
You´re kidding, right?
Let us play a series of games. In every game, you shall double before
your first roll. I may redouble if and whenever I please, and you of
course can then take or drop my redouble. We shall make no assumptions
that either of us make cube decisions "correctly" or "like robots or
brainwashed monkeys."
If we further assume that you and I are equals at checker play, I
assure you without a doubt that "the compounded effect of it if the
limitation of using the cube only once per game is lifted" will be:
(1) you will win fewer than 50% of all games, and
(2) you will win fewer points than I, your "equal" opponent.
________________________________________________
Daniel Murphy www.cityraccoon.com/
Humlebæk Backgammon Klub www.hbgk.dk/
Raccoon on FIBS www.fibs.com/
Raccoon on GamesGrid too
Murat Kalinyaprak
>Murat Kalinyaprak wrote:
>>Here is where we start. You and I are equal strength players.
>>With fair/equal dice and without the cube each would win 50%
>>of the time. Now suppose I double after the first roll each
>>and every game and the cube is never used again in any of the
>>games. Each of us will still win 50% of the time because the
>>roll after the double will make it "right" or "wrong" 50% of
>>the time. Now extrapolate this to doubling at later/different
>>stages of the games and then try to figure out the compounded
>>effect of it if the limitation of using the cube only once
>>per game is lifted. You will find out that you won't be able
>>to figure things out unless you assume that subsequent cube
>>actions will be taken "correctly". (BTW, that's exactly why
>>you guys are so tempted to consider ridiculous possibilities
>>like somebody soing something right repeatedly without knowing
>>how to do it right:). Back to the previous thought; in other
>>words, looking ahead won't work unless players execute future
>>takes/drops at X%/Y% winning/losing probabilities like robots
>>or brainwashed monkeys... In yet other words, by defying what
>>is expected as far as cube action, one can to a great extent
>>render the cube irrelevant to the game, in which case players
>>with the "staying-power" (i.e. superior checker skills) will
>>consistently win.
>You´re kidding, right?
Not intentionally... :)
>Let us play a series of games. In every game, you shall double
>before your first roll. I may redouble if and whenever I please,
>and you of course can then take or drop my redouble. We shall
>make no assumptions that either of us make cube decisions
>"correctly" or "like robots or brainwashed monkeys."
>If we further assume that you and I are equals at checker play,
>I assure you without a doubt that "the compounded effect of it
>if the limitation of using the cube only once per game is lifted"
>will be:
> (1) you will win fewer than 50% of all games, and
> (2) you will win fewer points than I, your "equal" opponent.
By clarifying/redefining/refining the conditions, you
can make some valid counter arguments to what I had said
and when I started reading what you wrote above I thought
you were going to do so but at the end I'm disappointed a
little.
The number one assertion above is plainly wrong because
the cube does't make you win or lose games among equal
players. However, it does raise the value of each game
and thus the number two assertion can be correct but
only with further clarified conditions.
With fair dice and equal checker strength, you won't
get to double properly any more than 50% of the time.
And if we are of equal strength in cube skill, then
you won't win any more points either because I would
take or drop properly. In other words, whenever you
reached 51% winning chances you would double and the
monkey brainwashed to drop at 50+% would properly drop.
Since you would never reach the cube point any more
often than 50% of the time, all of your doubles would
be dropped and you would never win any more points
per game.
I was, of course, talking about "defying" the cube,
etc. and you could have developed an argument in that
direction, which you haven't even began to to. Such
an argument would necessarily be much more complex
than what you tried above. For example, if N% of all
games end in gammon, if your opponent wrongly takes
X% of your correct doubles, and if these two coincide
more often then normal, you may quicly lose a lot of
what you gained by correct cube+correct checker play
to your opponent's wrong cube+correct checker play...
From the little I know about the cube, I would guess
that even the best robots couldn't properly consider
the overall combined effect of many factors (such as
gammons) when looking ahead. To that, we would have
to add strategies based on intentionally recycling
your pieces, etc. (which I haven't seen any robot be
able do so far) which would make the process even
more complicated and perhaps outright impossible. I
sure would like to see this subject being discusses
further among people more "knowledgeable" about the
cube stuff.
MK
Phill Skelton
>
> Daniel Murphy wrote:
>
> >Let us play a series of games. In every game, you shall double
> >before your first roll. I may redouble if and whenever I please,
> >and you of course can then take or drop my redouble. We shall
> >make no assumptions that either of us make cube decisions
> >"correctly" or "like robots or brainwashed monkeys."
> >If we further assume that you and I are equals at checker play,
> >I assure you without a doubt that "the compounded effect of it
> >if the limitation of using the cube only once per game is lifted"
> >will be:
>
> > (1) you will win fewer than 50% of all games, and
> > (2) you will win fewer points than I, your "equal" opponent.
> With fair dice and equal checker strength, you won't
> get to double properly any more than 50% of the time.
Not true. You are assuming that he never gets to double in half the
games. The fact is that some of the games you would win (if the cube
were not in use at all) are games in which he gets to the ideal
doubling point.
Lets say that your probabiloty of winning a game is w. Initially, w
is 0.5. You win when w reaches 1 (by definition) and lose when it
reaches 0.
Lets say that your opponent has the doubling cube, and will double
when you reach a probility of winning of w=t, and that it is
always correct to drop at this chance of winning (because you lose more
than 1 point per game if you take on average).
The probability of you winning (w->1) before your opponent gets to
double (w->t), starting from the inital w=0.5 is:
0.5 - t
-------
1.0 - t
If t=0 (your opponent never doubles either) then this is 0.5 - you
have a 50% chance of winning the game.
If t=0.25 (the highest it can be if you never double) then your
winnig chance becomes 1/3 - your opponent then wins 2/3 of the games.
i.e. *his* winning chance has gone from 50% to 66.7% due to him
having acces to the cube and you not using it.
> And if we are of equal strength in cube skill, then
> you won't win any more points either because I would
> take or drop properly. In other words, whenever you
> reached 51% winning chances you would double and the
> monkey brainwashed to drop at 50+% would properly drop.
> Since you would never reach the cube point any more
> often than 50% of the time, all of your doubles would
> be dropped and you would never win any more points
> per game.
Wrong, because the chance of reaching 51% starting from 50% before
reaching 0% is:
0.50 - 0.0
---------- = 0.98
0.51 - 0.0
So your monkey would lose 98% of games by dropping when slightly
behind.
Phill
(Incidentally it is possible to extend the above bits of maths to
account for both players using the cube, different skills using
the cube, volatility and match play - haven't included gammons in
it yet though)
Daniel Murphy
<mu...@compuplus.net> wrote:
>Daniel Murphy wrote:
>
>
>>Let us play a series of games. In every game, you shall double
>>before your first roll. I may redouble if and whenever I please,
>>and you of course can then take or drop my redouble. We shall
>>make no assumptions that either of us make cube decisions
>>"correctly" or "like robots or brainwashed monkeys."
>>If we further assume that you and I are equals at checker play,
>>I assure you without a doubt that "the compounded effect of it
>>if the limitation of using the cube only once per game is lifted"
>>will be:
>
>> (1) you will win fewer than 50% of all games, and
>> (2) you will win fewer points than I, your "equal" opponent.
>
>The number one assertion above is plainly wrong because
>the cube does't make you win or lose games among equal
>players.
The only way in a cubeful game to make the cube have no affect on the
number of games you win against an equal opponent is if you (1) do not
double yourself (other than the initial gift of the cube after your
first roll) and (2) take all doubles. That's not much of a strategy,
but if that's your strategy, my apologies. I thought you thought you
had something more interesting to say.
So, as long as the player (YOU) who is spotting his opponent the cube
in every game has some semblance of a rational strategy on deciding
when to take redoubles, the first assertion above is indeed true. If
two players are equal at checker play, the more masterful cube user
will win more games through cube use. If two equal players play AND
one gets to hold the cube, as you hypothesized, I daresay the player
holding the cube doesn´t even need to be better at cube play than the
other player in order to win more games.
>However, it does raise the value of each game
>and thus the number two assertion can be correct but
>only with further clarified conditions.
>With fair dice and equal checker strength, you won't
>get to double properly any more than 50% of the time.
>And if we are of equal strength in cube skill, then
>you won't win any more points either because I would
>take or drop properly. In other words, whenever you
>reached 51% winning chances you would double and the
>monkey brainwashed to drop at 50+% would properly drop.
>Since you would never reach the cube point any more
>often than 50% of the time, all of your doubles would
>be dropped and you would never win any more points
>per game.
See Phil Skelton's post which explains in more detail why the argument
and math in the paragraph above is wrong, but just think a moment
about typical backgammon games. It often happens in a SINGLE game that
BOTH players FREQUENTLY reach 51% or more winning chances at various
points in the game. If you or your monkey drop a cube any time your
winning chances drop below 50%, it stands to reason that the only
games you will ever win are games in which your opponent NEVER has
more than 50% winning chances. That just doesn´t happen very often.
>From the little I know about the cube, I would guess
>that even the best robots couldn't properly consider
>the overall combined effect of many factors (such as
>gammons) when looking ahead.
They certainly do consider them.
>to add strategies based on intentionally recycling
>your pieces, etc. (which I haven't seen any robot be
>able do so far)
JellyFish can and does recirculate checkers -- not as often as it
should, but then it doesn't need to do it very often.
Murat Kalinyaprak
>Jacking up the cube:
>(1) Intentionally making incorrect cube decisions in order
>to raise the cube to higher levels than are justified by
>the actual equity and volatility of a position, e.g.,
>doubling positions that are not good enough to double,
>or taking doubles which should be dropped.
>(2) Unintentionally making incorrect cube decisions...
>(3) In a more general sense, raising the value of....
>The following discussion pertains primarily to definition (1).
Good idea to just expand on the first one since the other
two are nonsensical anyway. Let me beging by asking who's
to decide whether a cube decision correct or incorrect...?
You? Joe the worl champion? Jellyfish? Snowie? God...?
>Jacking up the cube is a hallmark of players who see backgammon
>primarily as an opportunity to gamble and who enjoy the
>excitement of large cubes and large swings on the scoresheet.
This isn't me...
>Jacking up the cube is also a hallmark of the 'steamer' who,
>losing badly, wants to 'win it all back in one game' -- with
>the usual result of compounding his losses.
Obviously, in this case, this isn't me either...
>Such players, provided they pay when they lose, are generally
>welcome opponents, and are sometimes referred to as 'fish' or,
>in more genteel parlance, 'friendly players.'
I don't know if you punned intentionally but this sure
sounds like the case of Jellyfish here... :)
>Given extreme differences in skill between two players, however,
>jacking up the cube may be justified. A much weaker player may,
>for example, choose to accept disadvantageous odds in a single
>position to compensate for poor long-run chances of winning.
The player you call "weaker" jacks up the cube and the
other side is stupid enough to allow it, and you still
insist in calling it the "stronger/better" player? I'm
truely puzzled by this, especially when your "weaker"
player is the one who is winning. I'm not admitting to
doing what you describe nor would it matter even if it
were true. Why is the denial of reality/results? And if
you really need to deny it, why not take an easier way
out such as attributing it to luck, lack of sufficient
statistical data, etc...?
>A much stronger player may justifiably take or double
>'incorrectly' if opponent's subsequent misplays and erroneous
>cube (in)actions can be expected to give sufficient
>compensation for the stronger player's intrinsically
>inaccurate cube action.
>However, if the stronger player's expectations are accurate,
>such 'incorrect' cube decisions will in fact be correct, and
>should therefore not be considered to be jacking up the cube.
>And done with appropriate gusto, such 'incorrect' actions may
>even encourage one's opponent to indulge his own penchant for
>cubejacking.
Player "A" uses the cube based on his own way of reading
the board to decide whether he or his opponent the player
"B" is more likely to win and further makes a deliberate
attempt to increase the weight of his puch by "baiting?"
player "B" to make bad takes as described above (as well
as baiting it to make certain moves). Player "A" seems to
also exploit very well any early lead of a certain size
which often results in matches won by wide gaps in score
or even against zero at an unthinkable rate. All this is
fine and well, except that player "A" can never be Murat
and player "B" can never be Jellyfish... Oh no, no sir...
Impossible...
>The following 25-point match between the backgammon program
>JellyFish 3.0 and Murat, a merely human player, includes
>several instances of jacking up the cube -- none committed
>by the computer.
>------------------------------------------------
>25 point match
>Game 1
>JellyFish (O): 0 Murat (X): 0
>
>1) 53: 8/3 6/3
>2) 64: 8/2 6/2 11: 8/7 8/7 6/5 6/5
>3) 21: 24/21 Cube action??
>......................
>The inescapable conclusion is how easy it is to beat JellyFish
>or another expert player:
>Jack up the cube and get lucky!
>Double a no-double position, take a drop, redouble to 8 in a
>no-double position, recover from a blunder or two along the
>way, take a 16 cube with 29% winning chances, recube to 32
>and voilá -- another "proof" of the unimportance of doubling
>cube strategy in modern backgammon.
It sounds like you're trying to mock at something but
I don't quite know at what? Of the 10 matches of 25
points against Jellyfish that I had made a subject of,
only one had ended in a single game. I would've liked
nothing better than "getting lucky":) all the time and
keep winning 25 point matches in a single game each but
unfortunately I never get so lucky... :( If I remember
correctly, I had indicated that those matches had lasted
an average of 9 games. Even if I had won those matches
by winning all games, that would work out to about 2.8
points per game. Considering that JF had won a portion
of the games, average cube value per game must have been
higher than that. I would say that it's not a trivial
feat for a guy to sustain wins while keeping the cube
high, to a final result of winning 9 out of 10 matches.
Perhaps you can attribute some of it to luck but how
much of it...? Whatever may remain after what you would
be willing to belittle as luck, still would be a major
feat against such an overpraised player as Jellyfish.
I have the matches saved and can make them available by
email, by download from a web site, etc. if anybody is
interested in analysing the real thing instead of making
up imaginary scenarios. I'd have no problem with accepting
that whatever I may be doing right I do "unknowingly":) if
saying so would help you overcome your denial and possibly
learn something from whatever I may be doing...
MK
Paul Tanenbaum
"Murat Kalinyaprak" <mu...@compuplus.net> wrote:
> ...
> With fair dice and equal checker strength, you won't
> get to double properly any more than 50% of the time.
> And if we are of equal strength in cube skill, then
> you won't win any more points either because I would
> take or drop properly. In other words, whenever you
> reached 51% winning chances you would double and the
> monkey brainwashed to drop at 50+% would properly drop.
> Since you would never reach the cube point any more
> often than 50% of the time, all of your doubles would
> be dropped and you would never win any more points
> MK
Questions for Mr. Kalinyaprak: Are you familiar with the 25%
rule for accepting a double?
If so, have any opinion on it?
And are you a fan of Georges Mabardi?
---
Paul T.
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Murat Kalinyaprak
><7ja6no$a...@taisp3.in-tch.com> Murat Kalinyaprak writes:
>>Let me ask you all: are you guys a bunch of mental and/or
>>social wimps...? Why are you trying to trivialize something
>>before even you know what it is...? Why don't you guys have
>>any guts to go after answers to such questions from F. Dahl,
>>Oasya, etc...? Are you afraid of them, of government, of
>>Jesus, of God, etc...? Or are you afraid that you may not
>>like the answers you may be given...?
>Doh! Because I don't care!
It looks like you care enough to participate in the
thread (in general) and I think you would care more
(in specific) if you weren't missing the point...
>I don't care what moves JF makes when the moves don't matter.
Neither do I care what moves JF makes even when they
do matter. The issue here is two different revisions
of JF moving differently in identical circumstances,
and it all goes to whether JF is rigged or not. That
is the reason why I care about it...
>You ask:
>>"Why JF 3.0 and JF 3.5 move differently in identical
>>circumstances"...?
>But so far you have only given examples where IT DOESN'T MATTER.
>Since IT DOESN'T MATTER, I don't care. Come back with an example
>where it matters and maybe I'll care.
David, if I have to one up, let me say that I really
don't care whether you care or not about something
that you don't even understand. What happened between
the two JF versions besides the cosmetic changes...?
The dice roll generating bug was fixed and in a very
drastic way limiting valid seed range to 1-65535...!
No further neural-net training and identical set of
external data files... Yet, the newer revision plays
differently "FOR NO APPARENT GOOD REASON"...! Doesn't
this make you wonder at all...? You're one of the
people I take some liking to in this newsgroup, so
please tell me that you're not one of the herd with
castrated minds...
>Actually, when it has mattered, I have written Fredrik,
>and I've learned several undocumented things about JF
>that way. I've found Fredrik very responsive. But these
>positions... who (besides you) cares?
After the above clarification, I hope you will also
care. And since Fredric has been responsive to you
in the past, maybe you can get a response from him
on this issue also. In a previous posting I had dared
people to get answers to my questions from JF's and
SW's developers, insinuating that they should feel
free to resort to anything including kissing their
asses. But since I kind of like you and since I think
I know the answers which you and others assert they
don't care to know, I'd say don't go as far to kiss
his ass in case he is not readily responsive to you
on this issue... We'll leave that task to somebody
else in this newgsroup who may like me more than you
do... :)
MK
Murat Kalinyaprak
>In <7jl7v7$1p...@taisp3.in-tch.com> Murat Kalinyaprak wrote:
>> With fair dice and equal checker strength, you won't
>> get to double properly any more than 50% of the time.
>> And if we are of equal strength in cube skill, then
>> you won't win any more points either because I would
>> take or drop properly. In other words, whenever you
>> reached 51% winning chances you would double and the
>> monkey brainwashed to drop at 50+% would properly drop.
>> Since you would never reach the cube point any more
>> often than 50% of the time, all of your doubles would
>> be dropped and you would never win any more points
>> per game...
>Questions for Mr. Kalinyaprak:
>Are you familiar with the 25% rule for accepting a double?
No.
>If so, have any opinion on it?
N/A
>And are you a fan of Georges Mabardi?
Never even heard of him before.
Now I'm curious to know why you asked the above
questions, if you wouldn't mind explaining even
if briefly...?
MK
Robert-Jan Veldhuizen
<mu...@compuplus.net> wrote:
>Paul Tanenbaum wrote news:7jpiqh$ng$1...@nnrp1.deja.com...
>>Questions for Mr. Kalinyaprak:
>
>>Are you familiar with the 25% rule for accepting a double?
>
>No.
Then do yourself (and others) a favour and go read some good bg book
about the doubling cube, instead of trolling the newsgroup with your
mindless drivel.
BTW, are you having much success with your bg tactics of "compensating
for perceived biases"?
--
Robert-Jan/Zorba
Robert-Jan Veldhuizen
<mu...@compuplus.net> wrote:
>It looks like you care enough to participate in the
>thread (in general) and I think you would care more
>(in specific) if you weren't missing the point...
Do YOU care then, is the more interesting question.
--
Robert-Jan/Zorba
Paul Tanenbaum
"Murat Kalinyaprak" <mu...@compuplus.net> wrote:
> Paul Tanenbaum wrote news:7jpiqh$ng$1...@nnrp1.deja.com...
>
>
> >> With fair dice and equal checker strength, you won't
> >> get to double properly any more than 50% of the time.
> >> And if we are of equal strength in cube skill, then
> >> you won't win any more points either because I would
> >> take or drop properly. In other words, whenever you
> >> reached 51% winning chances you would double and the
> >> monkey brainwashed to drop at 50+% would properly drop.
>
> >Questions for Mr. Kalinyaprak:
>
> >Are you familiar with the 25% rule for accepting a double?
>
> No.
>
>
> Never even heard of him before.
>
> Now I'm curious to know why you asked the above questions,
> if you wouldn't mind explaining even if briefly...?
> MK
Certainly.
The "25% rule" is accepted as proper doubling technique among
modern backgammon players. It states that one should accept a double
even with as little as 25% winning chances. This is not based on
gambling, i.e. "I could still get lucky", but in the statistical sense
of minimizing one's losses.
The fact that you are unaware of this explains much of the
communications gap between yourself and most others here.
Mr. Mabardi authored a book shortly after the game's renaissance in
the US, which coincided with the invention of the doubling cube. As
for his views on the cube... well, let's put it this way - I've never
been a believer in reincarnation, but there's occasionally a piece of
supporting evidence which makes me wonder...
Rodrigo Andrade
"If two absolutely perfect players engaged in a match, there would never be
an accepted double."
And that's just the beginning...
"The less it [the doubling cube] is used the better."
"Rarely should a double be made then [in the early game]."
"The former [the doubling cube] is merely a new way of scoring."
The author recommends to move BOTH runners with the rolls 64, 63, 62, 43 (I
sometimes play it that way), 41, and 32... ouch!
He recommends a hopeless roll of 21 to run 24/21 with no mention of 24/23
13/11 as either good or bad.
If I had a time machine, I'd go back to 1930 and become a backgammon world
champ!
Thanks America we have return policy.
--
RODRIGO
===========================================================
"All religions of a spiritual nature are inventions of man. He has
created an entire system of gods with nothing more than his carnal brain.
Just because he has an ego and cannot accept it, he has had to externalize
it into some great spiritual device he calls 'God.'"
- The Satanic Bible
Anton Szandor LaVey
Murat Kalinyaprak
> Murat Kalinyaprak wrote:
>> Daniel Murphy wrote:
>>>Let us play a series of games. In every game, you shall double
>>>before your first roll. I may redouble if and whenever I please,
>>>and you of course can then take or drop my redouble. We shall
>>>make no assumptions that either of us make cube decisions
>>>"correctly" or "like robots or brainwashed monkeys."
>>>If we further assume that you and I are equals at checker play,
>>>I assure you without a doubt that "the compounded effect of it
>>>if the limitation of using the cube only once per game is lifted"
>>>will be:
>>> (1) you will win fewer than 50% of all games, and
>>> (2) you will win fewer points than I, your "equal" opponent.
>> With fair dice and equal checker strength, you won't
>> get to double properly any more than 50% of the time.
> Not true. You are assuming that he never gets to double in
> half the games.
Do you mean a bad double when he shouldn't be doubling?
I'm a little lost on this...
> The fact is that some of the games you would win (if the
> cube were not in use at all) are games in which he gets
> to the ideal doubling point.
The same is true for the other player...
> Lets say that your probabiloty of winning a game is w.
> Initially, w is 0.5. You win when w reaches 1 (by
> definition) and lose hen it reaches 0.
I think you explained things well and that I understand
them but don't they apply to both side equally...?
>> And if we are of equal strength in cube skill, then
>> you won't win any more points either because I would
>> take or drop properly. In other words, whenever you
>> reached 51% winning chances you would double and the
>> monkey brainwashed to drop at 50+% would properly drop.
>> Since you would never reach the cube point any more
>> often than 50% of the time, all of your doubles would
>> be dropped and you would never win any more points
>> per game.
>Wrong, because the chance of reaching 51% starting from
>50% before reaching 0% is:
>0.50 - 0.0
>---------- = 0.98
>0.51 - 0.0
>So your monkey would lose 98% of games by dropping when
>slightly behind.
Without arguing anything else, skill and dice being
equal, isn't the same true for your monkey as for my
monkey...? I wonder if we're talking about the same
thing here...?
MK
Phill Skelton
>
> On 1999/06/09 Phill Skelton wrote:
>
> > Murat Kalinyaprak wrote:
> >>
> >> With fair dice and equal checker strength, you won't
> >> get to double properly any more than 50% of the time.
>
> > Not true. You are assuming that he never gets to double in
> > half the games.
>
> Do you mean a bad double when he shouldn't be doubling?
> I'm a little lost on this...
I mean a correct double. Lets assume that I can double correctly with
75% cubless winning chance (roughly in accordance with common belief).
How often will I get to double? Lets say that my chance of getting to
the doubling point from the initial position (50% winning chance) is
X.
Lets ignore the existence of the cube for the moment. If I get to
at least 75% cpw (cubeless percentage wins) in X % of games, then I
will win 0.75 x X % of games. There are no games that I win in which
I don't get to >75% (since a win, at 100%, is >75%). So my total
number of wins is 0.75 x X, and since I know that I win 50% of games
over all (since I am playing an equal opponent), then:
0.75 x X = 50 %
X = 50 / 0.75 = 66.7%.
So I only have a 1/3 chance of losing the game before getting to double,
and a 2/3 chance of getting to double in any game.
So of course does my opponent, which might confuse some people.
The net result is that I win 1/3 of games with a double/drop or
double/take/win. My opponent wins 1/3 of games the same way. The
remaining 1/3 of the time, we both get a chance to double (unless one
of us drops the initial double). If we both have access to the cube
then everything is equal.
The original context for this was Daniel Murphys claim that if two
players (call them A and B) of equal checker play ability play a large
number of games, but player A starts off owning the cube, so player B
can't double, then player A will win more than 50% of the games, and
will win more points than B overall.
In the light of the above explanation, it can be seen that player B
will lose 0.33 points per game played (1/3 of the time he wins 1 point,
2/3 of the time he is correctly doubled and loses 1 point (through some
drops, some 2 point wins for B but 3 times as many 2 point wins for A)
Just for the record we can estimate the number of games won or lost as
well. Roughly 35% of doubles should be dropped (empirical estimate),
so 65% should be taken. Player B wins 33% of games easily, and
67% x 0.65 x 0.25 = 11% (double/take/win sequence), and so wins
44% of all games played while player A wins 56% (and player A has more
2 point wins than B)
All the above assumes that player B never recubes once he owns the
cube. If he does, then the only difference is that A needs to get to
perhaps 78% to double instead of 75%. Most of the leverage in A's
favour comes from the fact that A has access to the cube at the start
and B doesn't, and the question of whether B recubes or not is
irrelevant.
> >> And if we are of equal strength in cube skill, then
> >> you won't win any more points either because I would
> >> take or drop properly. In other words, whenever you
> >> reached 51% winning chances you would double and the
> >> monkey brainwashed to drop at 50+% would properly drop.
> >> Since you would never reach the cube point any more
> >> often than 50% of the time, all of your doubles would
> >> be dropped and you would never win any more points
> >> per game.
>
> >Wrong, because the chance of reaching 51% starting from
> >50% before reaching 0% is:
>
> >0.50 - 0.0
> >---------- = 0.98
> >0.51 - 0.0
>
> >So your monkey would lose 98% of games by dropping when
> >slightly behind.
>
> Without arguing anything else, skill and dice being
> equal, isn't the same true for your monkey as for my
> monkey...? I wonder if we're talking about the same
> thing here...?
If we are talking about two equally moronic monkeys playing then they
both win 50% (there is a 50% chance of getting to 51% cpw before
getting to 49% cpw). But a someone who understood cube behaviour
could take the monkey to the cleaners. If the monkey doubles at 49%
and drops at 51% then I would win 1.42 point per game off the simian
gold-mine, or more if it's chekecr play was worse than mine.
I've forgotten how this all started now - wasn't it something to do
with you claiming that correct use of the cube made no difference to
the number of games won or lost or the points won or lost?
Phill
Murat Kalinyaprak
> Murat Kalinyaprak wrote:
>> Daniel Murphy wrote:
>>>Let us play a series of games. In every game, you shall double
>>>before your first roll. I may redouble if and whenever I please,
>>>and you of course can then take or drop my redouble. We shall
>>>make no assumptions that either of us make cube decisions
>>>"correctly" or "like robots or brainwashed monkeys."
>>>If we further assume that you and I are equals at checker play,
>>>I assure you without a doubt that "the compounded effect of it
>>>if the limitation of using the cube only once per game is lifted"
>>>will be:
>>> (1) you will win fewer than 50% of all games, and
>>> (2) you will win fewer points than I, your "equal" opponent.
>> With fair dice and equal checker strength, you won't
>> get to double properly any more than 50% of the time.
> Not true. You are assuming that he never gets to double in
> half the games.
Do you mean a bad double when he shouldn't be doubling?
I'm a little lost on this...
> The fact is that some of the games you would win (if the
> cube were not in use at all) are games in which he gets
> to the ideal doubling point.
The same is true for the other player...
> Lets say that your probabiloty of winning a game is w.
> Initially, w is 0.5. You win when w reaches 1 (by
> definition) and lose hen it reaches 0.
I think you explained things well and that I understand
them but don't they apply to both side equally...?
>> And if we are of equal strength in cube skill, then
>> you won't win any more points either because I would
>> take or drop properly. In other words, whenever you
>> reached 51% winning chances you would double and the
>> monkey brainwashed to drop at 50+% would properly drop.
>> Since you would never reach the cube point any more
>> often than 50% of the time, all of your doubles would
>> be dropped and you would never win any more points
>> per game.
>Wrong, because the chance of reaching 51% starting from
>50% before reaching 0% is:
>0.50 - 0.0
>---------- = 0.98
>0.51 - 0.0
>So your monkey would lose 98% of games by dropping when
>slightly behind.
Without arguing anything else, skill and dice being
equal, isn't the same true for your monkey as for my
monkey...? I wonder if we're talking about the same
thing here...?
MK