MVP Backgammon question
Armin
To start with please excuse my ignorance on the subject of backgammon...
I'm new to the game but started to really enjoy it in the few month since I
have started playing.
I bought a copy of a shareware game called MVP Backgammon and I'm wondering
if anyone has had any experience with it. It seems to me that the rolls are
not very random. For example, when starting the game, white seems to win
the first roll almost all the time. I play against a coworker every lunch
hour and we have both started to wonder about this. I always play white so
this is annoying him more than me. ;)
Armin
Art Buell
> "Armin" <ahas...@awinc.com> wrote in article
<01bbc2e8$f744c0e0$5455...@ahasenko.awinc.co
> I bought a copy of a shareware game called MVP Backgammon and I'm
wondering
> if anyone has had any experience with it. It seems to me that the rolls
are
> not very random. . . .
I bought the same program, and I certainly agree. I know the tendency is
to subjectively blame the dice, especially with computer programs, but I'm
convinced that MVP is definitely biased in its rolls. Multiple high
doubles in a running game or when bearing off, hail mary combo shots off
the bar, leaving me stuck on the bar routinely for 5 or 6 rolls with a
3-point board--all these are routine, especially once the cube is
activated.
The worst feature, and one that's NOT subjective, is that if you do get
the computer into a Crawford situation, say I go up 6-3 in a 7-pointer,
and the computer wins that game, it doubles in the next game BEFORE you
get to see its opening roll!! The roll is made, but it's hidden from view
until you either blindly accept the double or resign. Very bad feature.
I'm ready to give up on this program.
artt
Phill Skelton
Art Buell wrote:
> The worst feature, and one that's NOT subjective, is that if
> you do get the computer into a Crawford situation, say I go up 6-3 in
> a 7-pointer, and the computer wins that game, it doubles in the next
> game BEFORE you get to see its opening roll!! The roll is made, but
> it's hidden from view until you either blindly accept the double or
> resign. Very bad feature. I'm ready to give up on this program.
That's how doubles work. You double BEFORE you roll - it's
always a bit of a risky affair.
Phill
Mark Betz
Hi, Art. I'm the author of MVP Backgammon, and I'd like to respond to your
points. I'll take the rolls issue first.
It is possible that the random algorithm used in MVP Backgammon is less
than ideal. I think it's a pretty good scheme, and I'll describe it in a
moment, but the key point is that is is _flatly impossible_ for my program
to bias its rolls in favor of the computer.
The dice engine is a server to the player objects. It has no knowledge of
whether the player object requesting a roll is a computer or human player.
In fact the distinction between human and computer players is invisible in
90% of this 50k line program.
The dice engine is contained in dtimer.dll, which is located in the program
directory. At startup this module creates two medium speed timers using the
multimedia hardware timer services. These timers "roll each die" in the
background about 18 times per second using a standard Borland RTL
pseudorandom number generator. When a player object (computer or human)
needs a roll it calls a method implemented in dtimer.dll which samples the
"roll stream" to obtain a value for each die.
I'd like to make it perfectly clear: the AI in MVP Backgammon does not
cheat; it has no invisible advantages over the human player, and in fact
the game is carefully architected so that all types of players are
subjected to the same constraints, and operate off of the same information
about the game context. In fact the AI is at a decided disadvantage
compared to the human player, as it knows nothing about the game history,
its opponent's history, or strategy, other than what is evidenced in a
single board position. Anyone caring to verify this independently may
contact me, and we can arrange for code to be examined.
On the issue of the operation of the game when the Crawford rule is enabed:
as Phil has pointed out the nature of doubles is that you offer them prior
to rolling. However, I confess that I am a much better programmer than
backgammon player. I implemented the Crawford and Jacoby rules according to
the best information I could get. My sources included Magriel's
"Backgammon", as well as Marc Ringuette, who worked on the AI with me, and
the beta testers in the MVPSOFT forum on Compuserve. I've relied on users
to verify the behavior, and to date noone has written to say that it is not
working properly. If there is a problem with the implementation of either
rule I hope someone here will be kind enough to make me aware of it. I am
about to release version 1.3 and would like to implement as many fixes as
possible.
Thanks for your thoughts.
Mark Betz
Art Buell <a...@adf.com> wrote in article
<01bbc3b5.fac60940$1a0baace@arthur>...
>
>
> > "Armin" <ahas...@awinc.com> wrote in article
> <01bbc2e8$f744c0e0$5455...@ahasenko.awinc.co
>
> > I bought a copy of a shareware game called MVP Backgammon and I'm
> wondering
> > if anyone has had any experience with it. It seems to me that the rolls
> are
> > not very random. . . .
>
> I bought the same program, and I certainly agree. I know the tendency
is
> to subjectively blame the dice, especially with computer programs, but
I'm
> convinced that MVP is definitely biased in its rolls. Multiple high
> doubles in a running game or when bearing off, hail mary combo shots off
> the bar, leaving me stuck on the bar routinely for 5 or 6 rolls with a
> 3-point board--all these are routine, especially once the cube is
> activated.
> The worst feature, and one that's NOT subjective, is that if you do get
> the computer into a Crawford situation, say I go up 6-3 in a 7-pointer,
> and the computer wins that game, it doubles in the next game BEFORE you
> get to see its opening roll!! The roll is made, but it's hidden from
view
> until you either blindly accept the double or resign. Very bad feature.
> I'm ready to give up on this program.
> artt
>
Mark Betz
Hi, Armin. I have responded to Art in a separate post which I hope you will
read. As the author of the game I am very sensitive to the idea that MVP
Backgammon does not give the human player a fair shake, and I believe in my
reply to Art I make it clear that architecturally the game cannot bias in
favor of the computer.
Whether or not a series of random numbers are really random is incredibly
subjective. What is important to realize is that "random series" does not
mean a homogenous mix of values, such that no sequence like 6, 6, 6, 6, 6,
6 will ever occur. Sequences of like values will certainly occur. That is
the nature of true randomness (if it even exists). The important metric is
the frequency distribution of values within the range. That is how a
pseudorandom algorithm is judged, and they are judged using very long
sequences indeed.
No computer is capable of true randomness, and most random number
generators leave much to be desired when it comes to simulating dice rolls
(which are not random at all, actually, but are simply the result of too
many forces for us to comprehend and measure). In my reply to Art I
described the steps I took within the program architecture to try for a
more realistic result.
I hope this answers your questions. Thanks for posting your thoughts.
Mark Betz
Armin <ahas...@awinc.com> wrote in article
<01bbc2e8$f744c0e0$5455...@ahasenko.awinc.com>...
> To start with please excuse my ignorance on the subject of backgammon...
> I'm new to the game but started to really enjoy it in the few month since
I
> have started playing.
>
> I bought a copy of a shareware game called MVP Backgammon and I'm
wondering
> if anyone has had any experience with it. It seems to me that the rolls
are
John B. Clements
In article <327490F1...@sun.leeds.ac.uk>,
Phill Skelton <ph...@sun.leeds.ac.uk> wrote:
>Art Buell wrote:
>
>> The worst feature, and one that's NOT subjective, is that if
>> you do get the computer into a Crawford situation, say I go up 6-3 in
>> a 7-pointer, and the computer wins that game, it doubles in the next
>> game BEFORE you get to see its opening roll!!
>
>always a bit of a risky affair.
>
Surely not before the _opening_ roll?
john
Mark Betz
One other quick follow-up. An alert reader pointed out that I had missed
the distinction of "opening roll" in the description of the problem
reported by Art. If the computer player doubled before the opening roll of
the game that is a bug. I don't think it can happen, but then that's true
of almost every bug I've fixed. If it can happen it will be corrected in
version 1.3, due to release shortly. The computer player can, obviously,
double immediately before it's first roll.
--Mark Betz
Cane Garden
Art Buell
>
> On the issue of the operation of the game when the Crawford rule is
enabed:
> as Phil has pointed out the nature of doubles is that you offer them
prior
> to rolling.
> Mark Betz
>
Mark--Thank you for your clear and well-reasoned reply to my posting. It
seems likely that my view of MVP's dice rolls were colored by
subjectivity--not the first time I've fallen prey to that. :-) So I
apologize for criticizing your program unfairly, and I'll shut up about
that and take the dice that are given.
However, the problem with the Crawford rule implementation still does
exist. Maybe I didn't explain it well enough in my posting. Yes, doubles
are offered before rolling, but what MVP does is offer the post-Crawford
double before the opening roll of the game is displayed (the roll
determining who makes the initial move in the game). I'm quite sure that
this initial post-Crawford double can only be offered AFTER each player
rolls one die to determine who takes the opening move and after this move
has been taken--that's how it works on FIBS and that's how I interpret the
rules as I've read them.
What you lose with MVP is the option of dropping if 1) your opponent wins
the opening move with a good roll, or 2) you win the opening move with a
poor roll. This is not a major flaw by any means, but it is a minor
annoyance.
Again, let me apologize for succumbing to the "the dice hate me"
syndrome. I purchased your program because I was impressed with your
software, and except for the above-mentioned Crawford problem and my own
subjectivity I still think you did a fine job with MVP. I said in my
earlier posting that "I'm ready to give up on this program," but to be
honest I've felt that way about backgammon in general many times--FIBS,
"real" backgammon, computer games, etc. But I'm still playing the damn
game, and I'm still using MVP--in fact I won 3 games in a row last night!
:-)
artt
Mark Betz
Hi. Backgammon is a frustrating game, sometimes. In any case, I've cursed
out a number of programs, and never felt the need to be personally
apologetic to the authors <g>.
The Crawford rule problem is simply a bug. Thanks for pointing it out. It
will be fixed in the next release.
--Mark
Chuck Dapoz
In your next update of MVP Backgammon, why don't you allow external
input of rolls (i.e., option to disable the dice engine so players can
toss physical dice or use another source)? Programming should be a snap.
It would end debate about capabilities of the program and whether its
rolls are random.
Chuck Dapoz