High Tech Pool
John Black
http://www.break.com/index/technology-meets-billiards.html
John Black
Jim White
Jim
"John Black" <jbl...@texas.net> wrote in message
news:MPG.252488547...@news.eternal-september.org...
Dan White
rack.
dwhite
"John Black" <jbl...@texas.net> wrote in message
news:MPG.252488547...@news.eternal-september.org...
John Black
says...
> Hmm. Coolest thing to me was how precisely it could place the balls in a
> rack.
Yeah, if it was faster, we wouldn't need racks or rackers anymore.
John Black
bk4...@hotmail.com
That was truly incredible.
Both the rig and the lights tracks showing ball paths.
That could be some kind of high tech (but expensive) training tool.
Thanks John.
Bob Keller
Fast Larry
Oh geeze Louise Bob, I have been doing this shit for years in my
lessons, using hand held lasers to teach my pol aim system, which is
super secret. I only teach this to smart people, no bozos, ding dongs
or rsb'ers may apply.
Or you can buy my $80 38 minutes fuckin DVD with $10 S&H, and I'll
throw in a shamu towel and a crack pipe with RSB on the side. B the
first in yo bar to light up with it.
Operators are standing by to take your order at 404-260-1318
snooze & U lose, so hubba hubba chop chop right now, U savy, go get
yours bayboose.
dave y.
wrote:
>Enjoy.
>
>http://www.break.com/index/technology-meets-billiards.html
>
>John Black
I found a few relevant links:
http://www.deepgreenrobot.org/
http://rcvlab.ece.queensu.ca/~deepgreen/
dave y.
John Black
says...
Thanks Dave,
The first link did not work but I found the project website:
http://rcvlab.ece.queensu.ca/~deepgreen/DeepGreen/index.html
There are lots more videos of it there. They hope that at some point the
robot will be the best player in the world. Right now its high run is only
5 so even I can beat it!!
John Black
bk4...@hotmail.com
> In article <avunb513v27hch832kebgqde6haevjc...@4ax.com>, nos...@myhouse.com
Those other videos are interesting. I had wondered how well it could
break. The answer is...well....like a girl! lol They are going to
have to generate a hell of a lot more cue speed for break shots. But
you would think the robot ought to be able to hit the head ball
exactly full every single time, huh?
I'd like to see it draw the cue ball straight back too. I think this
tool will be very illuminating regarding what type of cue movement (as
we often discuss) is needed to accomplish different things. The human
brain unconsciously accounts for SO MANY variables when we play! It'll
be cool to keep up with these guys and see how the robot improves.
Bob Keller
JakartaDean
> I found a few relevant links:
>
> http://www.deepgreenrobot.org/
>
> http://rcvlab.ece.queensu.ca/~deepgreen/
>
> dave y.
The project page FAQ, under the Q "Why robotic pool?" says the
following: "Another reason is that many people think that it can't be
done. Their argument is that pool is too difficult a problem, and that
playing pool is a skill that is uniquely human. We agree that pool is a
very challenging problem, and that it will take a lot of sweat and
effort to create a system that plays with a better-than-amateur level of
proficiency. We also believe, however, that the problem is solvable, and
that with enough dedicated effort, robots will be able to compete
against the best human players."
So, what do you think? I'm doubtful, I must admit, but it's never wise
to bet against technological advances. I think the logical challenge of
establishing an algorithmic approach to 8-ball strategy could outlast
the interest / funding of anyone taking it on. Still, it should be able
to learn 3 and 4 rail kicks out of a safety. Will it be able to learn
aggressive and defensive play, and when to use each? Will it be know to
hit out of a safety and leave a safety returned?
I guess these are solvable problems, and I'm sure it will soon be better
than me at breaking clusters and other geometric problems. Does anyone
here *cough* Ron *cough* know enough about modern computing to tell us
how to approach it? Can the computer simply learn by doing, storing
statistical data on what approach it tried and the result? Could this
statistical approach be layered on top of some basic strategy that we
could collectively list in, say, 200 points? I'd love to know the
answers. If a few people post their thoughts, perhaps we can exchange a
few thoughts with their team?
Dean
JakartaDean
> I guess these are solvable problems, and I'm sure it will soon be better
> than me at breaking clusters and other geometric problems. Does anyone
> here know enough about modern computing to tell us
> how to approach it? Can the computer simply learn by doing, storing
> statistical data on what approach it tried and the result? Could this
> statistical approach be layered on top of some basic strategy that we
> could collectively list in, say, 200 points? I'd love to know the
> answers. If a few people post their thoughts, perhaps we can exchange a
> few thoughts with their team?
>
> Dean
Never mind, I should finish scouring the links already posted. They use
the "Monte Carlo" method, which I need to learn about. One paper is
here: http://rcvlab.ece.queensu.ca/~greensm/papers/cig07.pdf That
article says they have held competitions among different algorithms and
weighting methods, but doesn't give the details on weighting or the
scoring of intermediate outcomes. Unsurprisingly, the algorithms which
looked two shots did better than ones looking one shot ahead, and both
outperformed a "greedy" player who just took the easiest shot every
time. The competitions mentioned did not allow banks, kicks,
combinations or safety play, although they say their programming does
include these. They also deliberately make the machine *less* precise
for some of these competitions, to simulate duffer and talented players.
I think they're years away from beating a pro or gifted amateur, unless
their routines are capable of going much deeper than two shots ahead in
real time and can identify safety opportunities. It'll be interesting
to watch this going forward, though.
BTW, this isn't really a new project, has anyone heard of it before?
I'm really tempted to see it next time I'm in Toronto, as it's only an
hour away. (The fading memories of some wild road trips in my
university days to the school might also be a factor.)
Dean
John Black
698c64...@g23g2000vbr.googlegroups.com>, bk4...@hotmail.com says...
> Those other videos are interesting. I had wondered how well it could
> break. The answer is...well....like a girl!
If similar to the girls in the WPBA, then probably good enough.
> I'd like to see it draw the cue ball straight back too. I think this
> tool will be very illuminating regarding what type of cue movement (as
> we often discuss) is needed to accomplish different things.
It would be hard to teach it to put the "wrist action" necessary for power
draw. Just kidding, as I know many don't beleive there is any wrist action
needed. So this robot should be able to get pretty sporty draw by just
hitting low and level and following through.
John Black
John Black
@forteinc.com says...
> dave y. wrote:
> > I found a few relevant links:
> >
> > http://www.deepgreenrobot.org/
> >
> > http://rcvlab.ece.queensu.ca/~deepgreen/
> >
> > dave y.
>
> The project page FAQ, under the Q "Why robotic pool?" says the
> following: "Another reason is that many people think that it can't be
> done. Their argument is that pool is too difficult a problem, and that
> playing pool is a skill that is uniquely human. We agree that pool is a
> very challenging problem, and that it will take a lot of sweat and
> effort to create a system that plays with a better-than-amateur level of
> proficiency. We also believe, however, that the problem is solvable, and
> that with enough dedicated effort, robots will be able to compete
> against the best human players."
>
> So, what do you think? I'm doubtful, I must admit, but it's never wise
> to bet against technological advances. I think the logical challenge of
> establishing an algorithmic approach to 8-ball strategy could outlast
> the interest / funding of anyone taking it on.
If I had to bet, I'd say this thing will never beat any really good players.
> Still, it should be able
> to learn 3 and 4 rail kicks out of a safety. Will it be able to learn
> aggressive and defensive play, and when to use each? Will it be know to
> hit out of a safety and leave a safety returned?
There are certain things that computers naturally do much better (well
faster) than people. Data searches, math, etc. Don't try to out compute a
computer -- they can do more very precise calcuations in a fraction of a
second than you could do in a lifetime and they won't make a single mistake
in any of those. However, there are many things our brains do better than
computers and faster and faster calculating ability has never been able to
make up for whatever they are lacking that we have. There are some limited
areas where sheer speed of evaluating millions of scenarios has allowed
computers to play games like chess pretty well. I don't think that would
translate as well to pool though. Each move in chess can be very simply
described -- move this piece from this square to that square. The speed of
moving the piece does not matter. It doesn't matter if you grab the piece
from the center, top, bottom, left or right or an infinite combination of
those. There is no physics to take into account, etc. Pool is a game of
touch, of finese.
Garbage in garbage out applies too. The computers that have played the best
chess have had extremely good chess players on the programming team. How
good are the programmers here at pool?
> I guess these are solvable problems, and I'm sure it will soon be better
> than me at breaking clusters and other geometric problems. Does anyone
> here *cough* Ron *cough* know enough about modern computing to tell us
> how to approach it? Can the computer simply learn by doing, storing
> statistical data on what approach it tried and the result?
A computer that could really learn by doing has always been sort of right
around the corner for many decades now. My prediction is that in several
more decades we will not be much closer to that goal than we are today.
Artificial Intelligence, neural networks, etc. were all the rage about 15 or
20 years ago. You hear much less about them now. I think we have basically
been humbled by the challenge of trying to create intelligence on a machine.
John Black
bk4...@hotmail.com
> Garbage in garbage out applies too. The computers that have played the best
> chess have had extremely good chess players on the programming team. How
> good are the programmers here at pool?
You made a very important point, John. I think they'll eventually
need to get someone on the team who knows the game. Right now it
seems like they are just trying to pocket balls with center ball and
aren't attempting any variation in cue tip placement.
Bob Keller
Dan White
news:wq%um.65087$3D4....@en-nntp-05.dc1.easynews.com...
> Unsurprisingly, the algorithms which looked two shots [ahead] did better
> than ones looking one shot ahead, and both outperformed a "greedy" player
> who just took the easiest shot every time.
We'll know that the computer is playing at a top level when the easiest shot
is also the correct one!
dwhite