Some gnubg-mk stats
Murat Kalinyaprak
To my dear friends and foes, here are
some stats and observations that you
may find interesting...
I saved 14 files of 25-point matches
against Gnubg, with the results as
below (in 2 sets of 6 and 7 matches):
Murat Gnubg total Murat Gnubg
points points games won won
------ ------ ----- ----- -----
32 00 01 01 00
38 05 09 07 02
72 00 02 02 00
25 08 14 07 07
26 00 04 04 00
21 26 15 10 05
------ ------ ----- ----- -----
214 39 45 31 14
28 00 03 03 00
36 18 07 05 02
25 00 04 04 00
36 08 09 06 03
22 26 20 12 08
33 04 07 06 01
13 25 14 06 08
------ ------ ----- ----- -----
193 81 64 42 22
====== ====== ===== ===== =====
407 120 109 73 36
After the last match, for some reason
I began to think that Gnubg was trying
too hard to win... Well, of course, no
such thing as gnubg's cheating can even
be suspected but noone can blame me for
suspecting that I may have supernatural
powers either, can they...? :)
So, I decided to make a rule that if at
any "extremely critical" point during a
game I could guess an unlikely roll/play
(i.e. way more than 50%) then I would get
to cancel the whole match, regardless of
who may be ahead at that moment...
Well, it has been perhaps more than a
dozen of matches cancelled one after
another (except for one that I won a
match 64-00 in just one game (oops:)...
I know, I know; astrology doesn't interest
you guys either, but fine, let's just look
at the 13 recorded matches above:
As far as matches, Murat is ahead by 10 to
3... Even though the sample is small, such
a feat is still nothing to snear at, no...?
As far as "cubeful points", Murat wins by
a total of 407 to 120, quite a large margin
wouldn't you say...? Does that mean that a
2050 FIBS rated robot doesn't know what to
do with that little device called "cube"...?
How come I am the one who makes all those
terrible cube errors and still win by that
margin against an extra-terrestrial "cube"
player...?
Maybe I'm taking advantage of some peculiar
"weaknesses of robots" by "jacking-up" the
cube, etc. (i.e. I am defeating a robot's
superior cube play by jacking-up the cube?)
Well, let's look at the number of games won:
Holy shit! Murat is running circles around
Gnubg with a score of 73 against 36 games...!
How come I am the one who makes all those
checker play errors and still win by such a
margin against an extra-terrestrial rated
checker-play master robot...?
The matches Murat lost are among the longest
ones and with scores of 10-05, 12-08, 06-08;
Murat is way ahead in number of games won in
2 matches (i.e. 22-13) and ahead by 28-21 in
overall number of games won...
Hey, perhaps that damn robot is jacking up
the cube...? And/or Murat is superior to it
both in checker play (which he claims to be
good at) and cube skill (which he admits to
be clueless about but calls it "bad-science
bullshit")...??
Well, I don't know the answers my friends
and foes, but I am willing to reproduce such
results in front of cameras for whomever may
want to shove it up their world-class ass...
Sorry... :( For a change, I thought I was
going to poste an article without profanities
or even strong words, but I just couldn't
resist... :( You all have a nice day now...
MK
-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 90,000 Newsgroups - 16 Different Servers! =-----
Graham Price
"Murat Kalinyaprak" <mu...@compuplus.net> wrote in message
news:3bbb...@post.newsfeeds.com...
> *** post for FREE via your newsreader at post.newsfeeds.com ***
>
> To my dear friends and foes, here are
> some stats and observations that you
> may find interesting...
ATTRIBUTION: Benjamin Disraeli (1804–1881)
Like playing poker for matchsticks.
Graham.
Murat Kalinyaprak
Graham Price wrote news:3bbf7...@news1.prserv.net
> Murat Kalinyaprak wrote news:3bbb...@post.newsfeeds.com
>> To my dear friends and foes, here are some stats
>> and observations that you may find interesting...
> QUOTATION: There are three kinds of lies: lies,
> damned lies and statistics. ATTRIBUTION:
> Benjamin Disraeli (1804-1881) Like playing
> poker for matchsticks.
The numbers I posted bother you, don't they...?
Of course, I already knew that they would bother
you and that's exactly why I had already offered
that I would be glad to duplicate them for you all
in front of cameras...
I'm fairly sure I can produce similar results with
any of its dice rollers (I was using "Isaac"), but I
could post the match files here, and if you were
willing to reroll the same rolls for me and gnubg,
I probably would be able to duplicate the results
even more closely. (I hope you won't argue that I
can memorize/remember 1000's of dice rolls)...
Another long/cold Montana winter is coming up.
I'll have lots of idle time. Why don't you guys make
it a little convenient for me to demonstrate what I
have been claiming about JF and now gnubg...?
I'll even risk to lose some of my own sweet money
betting against you all and perhaps can find a few
more people to join me in this, to make it exciting
enough even for the hard-core "gamblers" among
you...
I know, I know; now that gnubg is a non-profit bot
with more "honorable" sounding names behind its
development, the question "Why should it cheat?"
will be even more of a proof for you than has been
about JF...
Hey, I don't know... Maybe there are no sick sob's
who walk into a post office and gun down a dozen
innocent people, or who piss on and torture their
sex partners to derive pleasure, or who... etc. etc.
Of course, we would never go as far to call anyone
here a sick sob and we could safely assume that
creating cheating bots, conducting psychological
experiments, trolling, etc. etc. wouldn't be all that
satisfying for any of us here anyway...
So, if I have your attention by now, the only choice
you have remaining will be that I'm better than JF
and gnubg, while being a very bad checker and/or
cube player at the same time...
But your real problem will not start until I show you
that I can make seemingly bad checker moves or
even worse cube decisions (according to those
extra-terrestrial level players) and still run circles
around them...
That would really ruin your whole fantasy-world and
perhaps hurt bg book sales, bg player/analyser bot
sales, etc...? Come on folks, have some guts. Let's
get to the bottom of this. I'm sure many of you spend
lots of time and money on many useless things like
even backgammon... Why don't we make a minimal
effort to organize an event to show what we all got...?
Perhaps then, a few more "gutless-brains" may find
the courage to join me on the unpopular side of the
argument... Whad da ya all say...? I say let's do it...!
Ian Shaw
I'd be interested in having a look at a couple of those match files where
you won with a huge cube. Any chance of emailing me a couple? It might be
quite instuctive.
Ian
Christopher D. Yep
> After the last match, for some reason
> I began to think that Gnubg was trying
> too hard to win... Well, of course, no
> such thing as gnubg's cheating can even
> be suspected but noone can blame me for
> suspecting that I may have supernatural
> powers either, can they...? :)
>
> So, I decided to make a rule that if at
> any "extremely critical" point during a
> game I could guess an unlikely roll/play
> (i.e. way more than 50%) then I would get
> to cancel the whole match, regardless of
> who may be ahead at that moment...
>
This statement above seems to invalidate the whole study. How do you
define "extremely critical"? If "extremely critical" just means "when
one's frustration level has reached a maximum" then there is a high
correlation between an "extremely critical" situation and being behind
in the game or match. As far as cancelling a match due to an unlikely
play, do you mean that if you deem the situation critical and if Gnubg
makes an "unlikely" play (e.g. bearing off only one checker instead of
two checkers when Gnubg is either 0% or 100% to win seems to qualify
as an unlikely play in your mind), you will cancel the match?
Wouldn't it be better to include all matches? If I can correctly
guess my opponent's next roll in a tournament match, I can't cancel
the match. Similarly I can't cancel a match if I can correctly guess
my opponent's checker play given the dice roll!
> Well, it has been perhaps more than a
> dozen of matches cancelled one after
> another (except for one that I won a
> match 64-00 in just one game (oops:)...
>
This is a fatal flaw in your study. You simply have to include all
matches for your study to be valid. Even if we ignore for the moment
that competitive backgammon doesn't involve cancelling "unlikely"
matches, from a fairness point of view such a scheme is unfair to
Gnubg since Gnubg is not given the same opportunity to cancel matches
based on its determination of "extremely critical" moments, unlikely
rolls/plays, etc.
Chris
Murat Kalinyaprak
Christopher wrote news:7685eec1.01101...@posting.google.com
> This statement above seems to invalidate the whole study. How do you
> define "extremely critical"? If "extremely critical" just means "when
> one's frustration level has reached a maximum" then there is a high
> correlation between an "extremely critical" situation and being behind
> in the game or match. As far as cancelling a match due to an unlikely
> play, do you mean that if you deem the situation critical and if Gnubg
> makes an "unlikely" play (e.g. bearing off only one checker instead of
> two checkers when Gnubg is either 0% or 100% to win seems to qualify
> as an unlikely play in your mind), you will cancel the match?
Hi Christopher,
I wasn't keying on "moves" but "rolls"; and my criteria was like "less than
12 out of 36" (i.e. 33% or less)... Almost invariably, those were positions
where "gnudung" had one last chance before getting gammoned and/or
perhaps losing the match. It's so much more predictable than even what
I claimed about JF, that it's plain amuzing... :) or should I say sad...?
:((
>> Well, it has been perhaps more than a dozen of matches cancelled
>> one after another (except for one that I won a match 64-00 in just one
>> game (oops:)...
> This is a fatal flaw in your study. You simply have to include all
> matches for your study to be valid.
For that specific purpose and that period of time, I sure did include all
matches... What I meant is that: if I failed to "guess" an unlikely roll
(i.e.
33% or less), then the game would continue... And in one such a rare
case, I ended up winning 64 to 0 anyway...
> Even if we ignore for the moment that competitive backgammon doesn't
> involve cancelling "unlikely" matches, from a fairness point of view such
> a scheme is unfair to Gnubg since Gnubg is not given the same
> opportunity to cancel matches based on its determination of "extremely
> critical" moments, unlikely rolls/plays, etc.
You have to take it all as my fooling around to my hearts content... I just
felt that I observed something and invented a method to prove/disprove
it to my own satisfaction. I'm just relating what I did and I respect your
judgement about its validity and/or value...
Ian Shaw
> As far as matches, Murat is ahead by 10 to
> 3... Even though the sample is small, such
> a feat is still nothing to snear at, no...?
>
your opponent is trying to beat you on.
> As far as "cubeful points", Murat wins by
> a total of 407 to 120, quite a large margin
> wouldn't you say...? Does that mean that a
> 2050 FIBS rated robot doesn't know what to
> do with that little device called "cube"...?
It knows, for example, that it might as wll lose by 20 points as by 1. It
considers them both equal. Do you?
> Well, let's look at the number of games won:
> Holy shit! Murat is running circles around
> Gnubg with a score of 73 against 36 games...!
Irrelevant. GnuBg isn't trying to maximise thenumber of games it wins. Were
you?
> The matches Murat lost are among the longest
> ones and with scores of 10-05, 12-08, 06-08;
> Murat is way ahead in number of games won in
> 2 matches (i.e. 22-13) and ahead by 28-21 in
> overall number of games won...
I would tentatively draw the inference that these longer matches are the
ones where the cube action approached what most of us would consider
"normal". The implication is that GnuBg is better in these situations.
Murat Kalinyaprak
Ian Shaw wrote news:2QYz7.455$ur6.12573@wards
>> As far as matches, Murat is ahead by 10 to
>> 3... Even though the sample is small, such
>> a feat is still nothing to snear at, no...?
> This is the only statistic I would pay any attention to. It is
> the only one your opponent is trying to beat you on.
Yes, but the rest are relevant to preempt arguments
about my taking advantage of the robot by jacking
up the cube, etc. bs... I'm glad that you do at least
argue based on accepting that I'm being honest in
what I'm reporting here. I would certainly be happy
to take a 10-3 score against gnubg and call it the
end of the story. But I'm affraid others wouldn't be
satisfied/happy with just that...
>> As far as "cubeful points", Murat wins by
>> a total of 407 to 120, quite a large margin
>> wouldn't you say...? Does that mean that a
>> 2050 FIBS rated robot doesn't know what to
>> do with that little device called "cube"...?
> It knows, for example, that it might as wll lose by 20
> points as by 1. It considers them both equal. Do you?
Not true, Ian. We are not talking here about the cube
going up to 64 in a 25 point match. But the cube can
go up to 32 reasonably in a 25-point match. If I'm 4 to
16 behind for example and succeed in bringing up the
cube to 16 and win 36 to 16 by gammoning it, good
for me wouldn't you say...? And if it's gnubg who is
doing the same thing in a reverse case and loses by
48 to 4 by a gammon, that shoul also mean something
doesn't it...? Perhaps it doubles hopelessly...? Perhaps
it doesn't know what to do with that cube...? How can
you say that this isn't material...? With the cube at 8
and without a gammon, it couldn't win the match. Why
does it double to 16 unless it thinks it can win...? And
if it's wrong, shouldn't be discredited for having doubled
wrongly...? I just don't understand you guys... :(
In none of the games had the cube gone over 32 and
a total cubeful score of 407 to 120 is very material in
my opinion. The difference is almost 3.5 times...!
Please think about it more...
>> Well, let's look at the number of games won:
>> Holy shit! Murat is running circles around
>> Gnubg with a score of 73 against 36 games...!
> Irrelevant. GnuBg isn't trying to maximise thenumber
> of games it wins. Were you?
Irrelevant...? How could anything be more relevant in
the face of the arguments that I can only beat these
robots by short matches with high cubes...? I'm trying
to do exactly what gnubg is trying to do, which is to
win the matches by a combination of checker play
and cube play skills. I gave you both stats separately
to show that I'm ahead of it in both cases. I think you
are in denial my friend... :(
>> The matches Murat lost are among the longest
>> ones and with scores of 10-05, 12-08, 06-08;
>> Murat is way ahead in number of games won in
>> 2 matches (i.e. 22-13) and ahead by 28-21 in
>> overall number of games won...
> I would tentatively draw the inference that these longer
> matches are the ones where the cube action
> approached what most of us would consider "normal".
> The implication is that GnuBg is better in these situations.
I think you are looking for ways to find consolation. I
had concluded that I had not been as much of an easy
prey for gnubg as it had been for me. The average
cube value per game for the above mentioned three
matches would be 1.66, 1.25 and 1.78, which means
there was not much cube action in those matches. My
conclusion then would be that when there is cube action,
gnubg gets lost... I would hate to acknowledge that
gnubg was better at checker play and that I was better
at cube play during those 3 matches, but that's about
the only conclusion we draw from it I think...
Anyway, nice discussing with you. Take care...
Murat Kalinyaprak
Ian Shaw wrote:
>> What I meant is that: if I failed to "guess" an unlikely roll
>> (i.e. 33% or less), then the game would continue... And
>> in one such a rare case, I ended up winning 64 to 0 anyway...
> You call 33% unlikely? I would call 33% reasonably likely.
I guess we differ on this. To me, 33% as opposed to 66%
is unlikely. Also, notice that I had said "33% or less"... So,
unlikely rolls would range from 33% to 3%, perhaps the
average being at around 18%...? Would that be unlikely
enough for you then...?
> If a five comes up (which it will about 30% of the time with
> fair dice), you get to cancel the match. If not, you continue,
> but reserve the right to have another guess.
No, then that game would continue. Perhaps this is a silly
game I had invented but kept me entertained for a while.
More than how would the match end, I was interested in
how many times I could guess right. If you took a 100
guesses where you had 33% or less chance of randomly
being right but let's say you were right in 56% of the time,
wouldn't you think something may be fishy...? Personally,
I find playing against gnubg very entertaining if not quite
useful, especially at a low setting like 0-ply... :)
Ian Shaw
> *** post for FREE via your newsreader at post.newsfeeds.com ***
>
>
> Ian Shaw wrote news:2QYz7.455$ur6.12573@wards
>
> > points as by 1. It considers them both equal. Do you?
>
> Not true, Ian. We are not talking here about the cube
> going up to 64 in a 25 point match. But the cube can
> In none of the games had the cube gone over 32 and
> a total cubeful score of 407 to 120 is very material in
> my opinion. The difference is almost 3.5 times...!
> Please think about it more...
>
point cube in a 25 point match is perfectly reasonable. A 64 cube would not
be. I've no quibbles with you here. I queried it in order to rule out one
possible explanation for the results.
The difference of 407 to 120 may or may not be significant, depending on how
the individual games went. I recently lost an 11-point tournament match
16-1, in two games. It doesn't show that my opponent is much, much better
than me. If I'd lost 12-1 in eleven games, for example, the evidence might
be stronger. I don't think there is a useful conclusion to be drawn here.
> >> Well, let's look at the number of games won:
> >> Holy shit! Murat is running circles around
> >> Gnubg with a score of 73 against 36 games...!
>
> > Irrelevant. GnuBg isn't trying to maximise thenumber
> > of games it wins. Were you?
>
> Irrelevant...? How could anything be more relevant in
> the face of the arguments that I can only beat these
> robots by short matches with high cubes...? I'm trying
> to do exactly what gnubg is trying to do, which is to
> win the matches by a combination of checker play
> and cube play skills. I gave you both stats separately
> to show that I'm ahead of it in both cases. I think you
> are in denial my friend... :(
>
If the only result that matters is who wins the match, it is irrelevant how
many games were won by each player. That is what I was trying to point out.
As long as you are not delibrately trying to maximise the number of games
you win over and above winning the match, then the results are valid.
You have now explained that you were quoting these numbers as evidence for a
particular style of play. I don't have a problem with using them for that
purpose. Let's examine these numbers in a bit more detail. I will compare it
with a 21 point match I played recently, which is the longest match I have
ever played (I've never played to longer than 11 before). I and my opponent
are both reasonably good players, and I would subjectively describe the cube
actions as normal for players of our abililty, playing with traditional cube
useage techniques. Lets see, the match lasted 21 games. The distribution of
final cube values were:
1: 13
2: 7
4: 1
I'm sure we both made mistakes but this looks like a typical spread to me.
You played 14 25-point matches, a total of 350 points. This took 109 games,
so the average is 0.311 games per points of match length. My 21-point match
lasted 21 games, averaging 1.00 games per point of match length.
Obviously, since I only have a single match as a comparison, the results are
not statistically significant. Nevertheless, I find it interesting. To me
this would tend to support those who argue "that I can only beat these
robots by short matches with high cubes".
If your cube strategy is working for you, you should continue to use it (at
least until you find something even better!) What interests me is whether
your cube strategy is valid against GnuBg for anybody, which is why I am
looking at the matches you emailed me. Even more interesting would be if
your strategy would work against a live opponent.
Denial? Who knows? If I am in denial, my subconcious isn't going to own up.
I hope I'm trying to examine the facts carefully before I reach a
conclusion.
> >> The matches Murat lost are among the longest
> >> ones and with scores of 10-05, 12-08, 06-08;
> >> Murat is way ahead in number of games won in
> >> 2 matches (i.e. 22-13) and ahead by 28-21 in
> >> overall number of games won...
>
> > I would tentatively draw the inference that these longer
> > matches are the ones where the cube action
> > approached what most of us would consider "normal".
> > The implication is that GnuBg is better in these situations.
>
[Snip] My
> conclusion then would be that when there is cube action,
> gnubg gets lost...
Our conclusions are equally valid interpretaions of this statistic.
Ian Shaw
> guesses where you had 33% or less chance of randomly
> being right but let's say you were right in 56% of the time,
> wouldn't you think something may be fishy...?
A statistician could tell you how likely you were to achieve this success by
chance, but I'm not one. But yeah, it looks fishy to me. I still have doubts
about your accurate recollection of your success. If you wrote down your
predictions as you made them so that you could accurately measure your
results it would be better. Otherwise I would suspect your memory of playing
tricks on you.
Murat Kalinyaprak
Ian Shaw wrote news:skSD7.2401$ur6.145660@wards
> You have now explained that you were quoting these numbers
> as evidence for a particular style of play. I don't have a problem
> with using them for that purpose.
Frankly Ian, I really don't know what to make out of those
numbers and that's why I'm trying to discuss them here.
Here are some more numbers that may surprise you, as
they did surprise me. I just reviewed those 14 matches,
tallying gammons and backgammons. Do you have any
guesses as to what the ratios may be...?
Murat gammons 20 times vs. gnubg's 7...
Murat backgammons 1 time vs. gnubg's 1...
Is 29 out of 109 games being gammons/backgammons
average (27% for both of us)...?
For me it's 21 out of 73 games that I had won (28%) and
for gnubg it's 8 out of 36 (22%). Are these typical...? What
I find interesting here is that the ratios are fairly proportional...
What do these stats mean, if anything...? I don't really know.
Feel free to comment, speculate, etc...
> You played 14 25-point matches, a total of 350 points. This
> took 109 games, so the average is 0.311 games per points
> of match length.
Maybe not very important and agains I don't know what to
nake of it, but the correct numbers are 527 points in 109
games, giving 0.206 games per point... Is this average...?
Good, bad, interesting, boring...??
> If your cube strategy is working for you, you should continue
> to use it (at least until you find something even better!) What
> interests me is whether your cube strategy is valid against
> GnuBg for anybody, which is why I am looking at the matches
> you emailed me. Even more interesting would be if your
> strategy would work against a live opponent.
This is very interesting to me also. It has been a dilemma
for me with JF and now it is with gnubg... I wouldn't easily
dare to declare myself a world-class player and I'm sure I
would be very scared to make the same checker or cube
plays against a human oppnent. But I feel like I can beat
the shit out of those bot, more and more casually. It's like
I can "read" them...
Many times I wish I were geographically closer to the bg
circles to find out what the real bg world is like out there.
In fact, distance is probably not the biggest problem, if
only some qualified people would listen to me, test me
out and help me out with contributing what I may have to
contribute to this game... If I knew I was just imagining
things, deceiving myself, etc. I could at least stop wasting
my (and perhaps other's time) me on it...
> Our conclusions are equally valid interpretaions of this
> statistic.
I'm not sure about what's valid but I do at least appreciate
a rare civil (and hopefully intelliegnt) discussion... Thanks...
Ian Shaw
> Murat backgammons 1 time vs. gnubg's 1...
>
> Is 29 out of 109 games being gammons/backgammons
> average (27% for both of us)...?
>
> For me it's 21 out of 73 games that I had won (28%) and
> for gnubg it's 8 out of 36 (22%). Are these typical...? What
> I find interesting here is that the ratios are fairly proportional...
>
> What do these stats mean, if anything...? I don't really know.
> Feel free to comment, speculate, etc...
>
end in gammons, so your numbers look typical.
> Ian Shaw wrote news:skSD7.2401$ur6.145660@wards
> > You played 14 25-point matches, a total of 350 points. This
> > took 109 games, so the average is 0.311 games per points
> > of match length.
>
> Maybe not very important and agains I don't know what to
> nake of it, but the correct numbers are 527 points in 109
> games, giving 0.206 games per point... Is this average...?
> Good, bad, interesting, boring...??
>
I was calculating based on the "official" length of the matches. The idea
was to calculate the average number of games to complete a match of a given
length. For this purpose it is correct to ignore the "overage" of points won
in excess of the number required. Using your figure of 0.206 games per point
exagerates the evidence in support of the claim that you are winning in
short matches with high cubes. I think my method is fairer to you, since
winning with lots of overage is perfectly acceptable.
> I'm sure I
> would be very scared to make the same checker or cube
> plays against a human oppnent. But I feel like I can beat
> the shit out of those bot, more and more casually. It's like
> I can "read" them...
>
After a brief look at your matches it looks like you are very quick to take
the cube and redouble almost straight away, even when you are an underdog
according to the bots analysis, (which I'm sure you would dispute!) If the
bot is right, good players would attempt to take advantage of this tendency
in a way the bot can't. They might wait and wait to double, thinking that
you would take when other players would surely drop.
They might double early, knowing that you would quickly recube. They would
then be playing holding a 4 cube which only they could use. You would have
to win all games by bearing off your last checker. They could do this, or
double when they are far enough ahead that they are sure you will drop. Say
this is when they are 95% ceratin to win. They never have to go through the
pain of playing to a conclusion and losing 5% of those games. This is an
illustration of the advantage of holding the cube.
Murat Kalinyaprak
Ian Shaw wrote u7PF7.3604$ur6.216763@wards
> Murat Kalinyaprak wrote 3be3...@post.newsfeeds.com
>> Maybe not very important and agains I don't know what to
>> nake of it, but the correct numbers are 527 points in 109
>> games, giving 0.206 games per point... Is this average...?
>> Good, bad, interesting, boring...??
> I was calculating based on the "official" length of the matches.
> The idea was to calculate the average number of games to
> complete a match of a given length. For this purpose it is
> correct to ignore the "overage" of points won in excess of the
> number required. Using your figure of 0.206 games per point
> exagerates the evidence in support of the claim that you are
> winning in short matches with high cubes. I think my method
> is fairer to you, since winning with lots of overage is perfectly
> acceptable.
Ian, I either can't express myself enough or you guys can't
understand or refuse to understand the point I'm trying to
make here. What you call "overage" is unintentional...! As
long as the last cube action doesn't jack it up above what
would be needed to win the match at that point (i.e. pushing
the cube to 32 when even 16 points are enough to win) the
extra points earned are significant simply because those
same points could also be lost...!
The cube does nothing more than to increase the value of
a given game within a match and what you call "overage"
is nothing more than the necessary result of the concocted
"bullshit" that they call "cube-skill"... Even then, what you
call "overage" is as meaningful as the rest of the "bullshit"
called "cube-skill"...
We can, of course, reduce that "overage" and in fact we
can eliminate it altogether by playing bg the old fashioned
way (i..e without the cube)... But, since the whole current
"bg world" is convinced that the "cube skill" is a "god sent"
improvement to bg, let's stick to reducing the "overage",
instead of eliminating completely...
The "overage" in cube-skill is exponential just because it
is called "doubling cube"... What if we call it "increasing
cube" instead...? What if we say that a player can "raise
the cube" by at least 1 point but by not more than twice its
current value...?
(If no such game already exists, I think it would be quite
appropriate to propose such a new kind of cubeful bg
game and I suggest it to be called "muratgammon"... :))
Suppose I'm leading 20-0... I own the cube at 4... I think I
can win the game, but 4 points aren't enough to win the
entire match. So in order to take advantage of my position,
I would have to "double" to 8, which then would give my
opponent a chance to double to 16 (even in a "desperate"
attempt).. This gives my opponent 6 free points...!
Let me explain a little more... If I could raise the cube by
only 1 point to 5, which is all I would need to win, instead
of having to raise it by 4 points to 8; my opponent could
only raise it back to 6 (i.e. 5+1) or to 10 (i.e. 2x5) instead
of being able to raise it to 16, which gives him/her 6 extra
points and put me at the risk of losing 6 more points than
I should really have to risk...
With these rules, if I ended up winning anyway, I would only
end up with a surplus of 5 points (i.e. winning 30-0 instead
of 25-0), while under the existing cube rules I woud win with
a surplus of 11 points (i.e. winning 36-0 instead of 25-0)...
So, since the "surplus" points are very significant to the
trailing/losing side, they should be just as significant to
the leading/winning side...
If I'm leading 20-0 and lose after my opponent doubles
to 16, my opponent gains a whole 16 points toward
winning the match as compared to my gaining only 5
points towards winning the match. But if I were leading
by only 20-19, then he would only gain 6 points whether
the cube was at 8, 16, 32 or 64; as compared to my
winning only 5 points towards winning the match...
What I'm suggesting here would deter/prevent what we
called "jaking up the cube" (i.e. by winning and should I
also say losing short matches by huge cube values)...
This is exactly why I had coined the term "jackgommon"
for the cubeful backgammon with the current rules and
as a compromise, I'm suggesting "muratgammon" that
would allow a player to increase the cube by as little as
1 or as much as twice its current value...
(Of course, the 1-point cubeless mathces which I had
called "lackgammon" have absolutely no resemblance
to backgammon at all)...
I guess I'm a "benevolent" person to not charge you all
$19.96 for telling you all these and I would be more than
happy enough if at least some of you understood what
I'm talking about...
Daniel Murphy
At one extreme, frequent rec.games.backgammon poster Murat Kalinyaprak
writes (in r.g.bg, 10 November 2001):
"The cube does nothing more than to increase the value of a given game
within a match and what you call 'overage' is nothing more than the
necessary result of the concocted 'bullshit' that they call
'cube-skill'... Even then, what you call 'overage' is as meaningful as
the rest of the "bullshit" called "cube-skill."
On the other hand, two-time American Backgammon Tour champion Neil
Kazaross writes (in Ortega & Kleinman's "Backgammon With the Giants:
Neil Kazaross," San Jose, C.R.: Editorama, 2001):
"I much prefer match play to money play, because there's much more to
it. In tournaments, the match score is everything, especially with
regard to cube strategy.... One of the reasons I've done so well in
matches is my understanding of how to use the cube at different
scores.... I cannot control the dice, but I can maneuver my pieces and
the doubling cube to try to increase my probability of winning."
Any questions?
Daniel Murphy
Raccoon on FIBS/GamesGrid
--------
"I should spend more time thinking about this instead of writing
down ideas that come to me." -- hmh, in rec.games backgammon, 10 Nov. 2001
Murat Kalinyaprak
*** Reposting due to distribution problems reported by
people who wanted to continue certain discussions ***
--------------------------------------------------------------------------
Ian Shaw wrote u7PF7.3604$ur6.216763@wards
> Murat Kalinyaprak wrote 3be3...@post.newsfeeds.com
>> Maybe not very important and agains I don't know what to
>> nake of it, but the correct numbers are 527 points in 109
>> games, giving 0.206 games per point... Is this average...?
>> Good, bad, interesting, boring...??
> I was calculating based on the "official" length of the matches.
> The idea was to calculate the average number of games to
> complete a match of a given length. For this purpose it is
> correct to ignore the "overage" of points won in excess of the
> number required. Using your figure of 0.206 games per point
> exagerates the evidence in support of the claim that you are
> winning in short matches with high cubes. I think my method
> is fairer to you, since winning with lots of overage is perfectly
> acceptable.
Ian, I either can't express myself enough or you guys can't
understand or refuse to understand the point I'm trying to
make here. What you call "overage" is unintentional...! As
long as the last cube action doesn't jack it up above what
would be needed to win the match at that point (i.e. pushing
the cube to 32 when even 16 points are enough to win) the
extra points earned are significant simply because those
same points could also be lost...!
The cube does nothing more than to increase the value of
a given game within a match and what you call "overage"
is nothing more than the necessary result of the concocted
"bullshit" that they call "cube-skill"... Even then, what you
call "overage" is as meaningful as the rest of the "bullshit"
Murat Kalinyaprak
Daniel Murphy wrote 3bed25da....@news.get2net.dk
> At one extreme, frequent rec.games.backgammon poster
> Murat Kalinyaprak writes (in r.g.bg, 10 November 2001):
> "The cube does nothing more than to increase the value of
> a given game within a match and what you call 'overage' is
> nothing more than the necessary result of the concocted
> 'bullshit' that they call 'cube-skill'...
> On the other hand, two-time American Backgammon Tour
> champion Neil Kazaross writes (in Ortega & Kleinman's
> "Backgammon With the Giants: Neil Kazaross," San Jose,
> C.R.: Editorama, 2001):
> "I much prefer match play to money play, because there's
> much more to it. In tournaments, the match score is everything,
> especially with regard to cube strategy.... One of the reasons
> I've done so well in matches is my understanding of how to
> use the cube at different scores.... I cannot control the dice,
> but I can maneuver my pieces and the doubling cube to try to
> increase my probability of winning."
> Any questions?
Yes, indeed Daniel, there seems to be quite a few questions
that you, your likes and your idols are continiously ignoring...
And then you all still have the guts to come try to talk down to
me on this issue and make "case closed" kind of comments...?
Let me ask you all once more, as clearly as I can and itemize:
1 - Is there any considerable disagreement among the world
class players about cube skill and all the rest of the god given
rules in modern backgammon...? If not, how would we know
the difference...? A few days ago I wrote about this and I asked
things like what if everybody adhered to bullshit cube skill rules
like doubling after the 5th roll on Mondays...? Do you think that
some world class player would gain or lose from such a "cube
skill teaching" as compared to another world class player...?
Now, please, stop wasting my time with endless and worthless
mumblings and give a direct answer to this question. If you
can't; then maybe you can ask Kazaross to give an answer...
2 - How does cube skill become more or less quantifiable as
the length of the match increases...? Suppose I'm an1800 elo
(another worthless bullshit) rated player playing against a 2000
rated player and double at the first chance I get (i.e. on my 1st
or 3rd roll, depending on who starts). Now, would you please
quantify that for me or if you can't; you are again welcome to
to seek Kazaross' help....
2a- Suppose I'm a 1800 rated player, playing against a 2000
rated player and I take each and every one of his/her/its doubles
(and I mean this! i.e. even during bearinbg off, when he/she/it is
down to 1 piece left to bear off!)... Could you please quantify
this for me also or if you can't; you can once more seek help
from world class cube gurus like Kazaross' to feed us more
cube skill bullshit...
There were other questions of mine that had gone unanswered
but let's leave it it this for today... Come on, Daniel, you guys
know it all and you have all them world class gurus behind you...
Is it too much to ask from you to answer the above questions
before you argue any more on this issue on this forum...? If so,
then let me know so that I can say something like "or else,
fuck off...!!!", which will give you an excuse to argue that you
don't debate with rude people, to put me in your kill-file, etc.
and "fuck off" one way or another... While I'm being on a binge
of being rude, why not dare Kazaross (or whatever else idols
you may have) to give me some answers or also "fuck off"...!
Ian Shaw
> Ian Shaw wrote u7PF7.3604$ur6.216763@wards
>
> > I was calculating based on the "official" length of the matches.
> > The idea was to calculate the average number of games to
> > complete a match of a given length. For this purpose it is
> > correct to ignore the "overage" of points won in excess of the
> > number required. Using your figure of 0.206 games per point
> > exagerates the evidence in support of the claim that you are
> > winning in short matches with high cubes. I think my method
> > is fairer to you, since winning with lots of overage is perfectly
> > acceptable.
>
> Ian, I either can't express myself enough or you guys can't
> understand or refuse to understand the point I'm trying to
> make here. What you call "overage" is unintentional...! As
> long as the last cube action doesn't jack it up above what
> would be needed to win the match at that point (i.e. pushing
> the cube to 32 when even 16 points are enough to win) the
> extra points earned are significant simply because those
> same points could also be lost...!
>
with lots of overage is perfectly acceptable". As I wrote in another post,
"A 32 point cube in a 25 point match is perfectly reasonable. A 64 cube
would not be. I've no quibbles with you here."
> The cube does nothing more than to increase the value of
> a given game within a match and what you call "overage"
> is nothing more than the necessary result of the concocted
> "bullshit" that they call "cube-skill"... Even then, what you
> call "overage" is as meaningful as the rest of the "bullshit"
> called "cube-skill"...
>
I find your use of perjoratives rearding cube-skill unhelpful. When you
insult something I enjoy I feel angry and defensive, and my instinct is
defend my point of view without consdiring all the possibilities. It is then
not so easy to engage in sensible discsussion.
When you call cube-skill "bullshit", I understand you to mean that cube
usage does not involve skill. Is this what mean by these statements? Slang
usage may vary in meaning between England and the US (it certainly varies
around England!), so please clarify your meaning.
You are correct to state that the cube increases the value of a given game
within a match. I argue that this is where the skill comes in. On a
take/pass descsion, a player has to judge whether they are better off
playing out the current game at the increased stake, or giving up the game,
conceding a smaller number of points but giving themselves more time to make
up the deficit (if they are trailing the match), or preserving their lead
(if they are leading the match.
I believe that this requires skill in two areas:
Being able to judge what percentage of the time you are likely to win from
the current psotion.
Ability to judge how likely you are to win the match if you pass or take,
depending on the match scores.
Checker play skill involves judging the RELATIVE merits of several possible
moves and making the one which gives or the best chance to win the game. (Or
a gammon, or whatever.)
Cube skill involves judging the ASOLUTE likelihood of winning a game (as a
percentage orequivalnet measure).
They are two different types of skill, but both are skills. If you don't
value one of them, or do not enjoy games involving one of them, that's fine.
If other people find pleasure in using both skills, that's fine too.
> We can, of course, reduce that "overage" and in fact we
> can eliminate it altogether by playing bg the old fashioned
> way (i..e without the cube)... But, since the whole current
> "bg world" is convinced that the "cube skill" is a "god sent"
> improvement to bg, let's stick to reducing the "overage",
> instead of eliminating completely...
Speaking personally, I enjoy playing with or without the cube. I learnt the
game cubeless and still often play the game that way. Since being online, I
have also found enjoyment in cube skills.
> The "overage" in cube-skill is exponential just because it
> is called "doubling cube"... [snip]...
> Suppose I'm leading 20-0... I own the cube at 4... I think I
> can win the game, but 4 points aren't enough to win the
> entire match. So in order to take advantage of my position,
> I would have to "double" to 8, which then would give my
> opponent a chance to double to 16 (even in a "desperate"
> attempt).. This gives my opponent 6 free points...!
>
I argue that there is skill in judging whether it is worth your while to
double and risk losing those sixteen points, or whether you are better off
playing for a lower number of points in this game and trying to win the rest
of the necesary points later.
> I'm suggesting "muratgammon" that
> would allow a player to increase the cube by as little as
> 1 or as much as twice its current value...
I've wondered along these lines too, like what would be effect of a cube
that just went 1, 2, 3, 4, 5, 6 etc. I'd have to give your suggestion more
thought before making a comment. My initial impression is that it would
increase the complexity of cube descisions. Some people would enjoy extra
complexity, others wouldn't.
> (Of course, the 1-point cubeless mathces which I had
> called "lackgammon" have absolutely no resemblance
> to backgammon at all)...
>
I enjoy these, too. The priority is simply winning the game, without
worrying about gammons and backgammons. Many countries play without counting
backgammons. What is wrong with not counting gammons either? The checker
movement rules are still the same, so there is certainly a "resemblence to
backgammon". Or are you being sarcastic?
Ian Shaw
>
> Let me ask you all once more, as clearly as I can and itemize:
>
> 1 - Is there any considerable disagreement among the world
> class players about cube skill and all the rest of the god given
> rules in modern backgammon...? If not, how would we know
> the difference...?
I'm not an expert, but I'll try and answer, more politley than you care to
pose the questions.
World class players pretty much agree on good cube strategy. I'm sure they
disagree on individual examples.
Know the difference beteen what?
A few days ago I wrote about this and I asked
> things like what if everybody adhered to bullshit cube skill rules
> like doubling after the 5th roll on Mondays...? Do you think that
> some world class player would gain or lose from such a "cube
> skill teaching" as compared to another world class player...?
> Now, please, stop wasting my time with endless and worthless
> mumblings and give a direct answer to this question. If you
> can't; then maybe you can ask Kazaross to give an answer...
>
I believe that any world class player would lose from such a "cube skill
teaching" because it's so obviously not in accordance with what we
understand about the laws of physics. Someone who taught that would lost MY
respect. They'd lose out on book sales. They'd only lose out in games if
they actually followed it!! I don't believe that the day of the week you
play affects bg strategy. If this strategy was in common use, I'd try to
play a lot on Mondays because I believe I would have an advantage over
players following this strategy.
There's an answer for you. What good does it do?
> 2 - How does cube skill become more or less quantifiable as
> the length of the match increases...? Suppose I'm an1800 elo
> (another worthless bullshit) rated player playing against a 2000
> rated player and double at the first chance I get (i.e. on my 1st
> or 3rd roll, depending on who starts). Now, would you please
> quantify that for me or if you can't; you are again welcome to
> to seek Kazaross' help....
>
> 2a- Suppose I'm a 1800 rated player, playing against a 2000
> rated player and I take each and every one of his/her/its doubles
> (and I mean this! i.e. even during bearinbg off, when he/she/it is
> down to 1 piece left to bear off!)... Could you please quantify
> this for me also or if you can't; you can once more seek help
> from world class cube gurus like Kazaross' to feed us more
> cube skill bullshit...
According to the FIBS formaula, a two hundred point advantage means you
should win 58% of 2-point matches. Let's
assume that this is accurate for any single game within a match. (I'm not
aware of any specific figures.) I'll also assume that 20% of all games end
in gammons.
Strategy 2.
In a 2-point match, your strategy is correct according to current theory.
There will only be one game in the match so the better player (Player A
henceforth) should win 58% of the matches.
In a 3-point match;
Player A wins in one game 58% if he wins a gammon. P = 0.58 * 0.2 = 0.116
Player A wins on second game if:
He is ahead 2-0 and wins again .58 * .8 * .58 = 0.26912
Player B can't double in this game because it is the Crawford game.
When Player B wins if the Crawford game, player A still wins the match if he
wins the next game.
Player A wins on third game if:
He loses the first game (not gammoned) and then wins the next two. 0.42 *
0.8 * 0.58 * 0.58 = 0.1130304
He wins the first (not gammon), loses the second and wins the third = 0.58 *
0.8 * 0.42 * 0.58 = 0.1130304
The rest of the time Player b wins the match.
Adding up these probabilities: Player A wins 61.11% of matches.
This has taken rather a long time. It has also ignored Player As ability to
double in game 1. You also haven't specified player Bs strategy for
redoubles, which would make a difference.
If you want to go further you have to build up from short length matches to
longer ones, figuring in all the various options. I won't enjoy doing that,
and I don't believe I will benefit from it, so that's as much time as I'm
prepared to give the matter. I've given you an idea how it's done, you can
do it for yourself if you are truly interested. Feel free to email me for
advice. If you're not interested enough to do it, I don't think you can
expect other people to do the work for you.
I doubt whether anyone has worked out the values you want to know, since
they don't apply to the situations we come across. I think it is probably
theoretically possible to come up with an answer. If you are really
interested, there is a book called "Can a fish taste twice as good" which
goes into games between unequal players.
There is also a nice article
http://www.gammonvillage.com/news/article_display.cfm?resourceid=934 by Jake
Jacobs which addresses an example of the type of thinking required. It
includes a response by Neil Kazaross. I would guess that this is an example
of the current "state of the art". Who knows whether the future will show it
is correct.
> There were other questions of mine that had gone unanswered
> but let's leave it it this for today... Come on, Daniel, you guys
> know it all and you have all them world class gurus behind you...
> Is it too much to ask from you to answer the above questions
> before you argue any more on this issue on this forum...?
It is a lot to ask that someone doesn't contribute to the newsgroup unless
they answer YOUR questions. Particularly since YOU haven't answered my
questions in other posts.
Regards,
Ian
Craig Campbell
the topic of match play with opponents of differing strengths quite well. Trice
has a companion program MEG that generates Match Equity Tables for opponents
of different skill levels. I find this very useful in my analysis. For instance
using Snowie's evaluations and substituting these METs you will see that if you
are the better player it is frequently correct to play an opening 2-1 as 13/11,
6/5.
CTC
--
Posted from spider-wc074.proxy.aol.com [205.188.193.54]
via Mailgate.ORG Server - http://www.Mailgate.ORG
Ian Shaw
"Craig Campbell" <ccamp...@aol.com> wrote in message
news:04a37712f8ab17922fc...@mygate.mailgate.org...
>For instance
> using Snowie's evaluations and substituting these METs you will see that
if you
> are the better player it is frequently correct to play an opening 2-1 as
13/11,
> 6/5.
>
stronger player. Interesting to see the theory supported by numbers.
Do you have any more changes for opening rolls?
When I'm the weaker player I usually run with 64, sometimes with 63 but
never with 62.
When stronger I usually pla 64 as 24/18 13/9, but I'm experimenting with 8/2
6/2.
I bet 43 is interesting, because there's lots of good candidates anyway. I
sometimes fo for 13/9 13/10 when stronger, but usually I try one of the
splits. The double split intruges me but I don't go for it.
41 & 51 slotting plays have fallen into disfavour. Could they ever be right
if you are stronger?
Ian