2007 Pokerbot World Championship

[RayBornert]

[ QUOTE ]
[ QUOTE ]
Why are some turning this thread into an ethical debate about bots? Stop being so damn rude, IMO. Save the debates for a different thread, if you must. This guy came here with the permission of mods from 2+2, some kind of legitamacy speaking there, right? He's also been completely understanding and reasonable with everyone's questions and objections, so far. But if you truly still disagree with the ideas/ethics behind this event/person, then disagree elsewhere or do some real research and find some concrete reasons to ram him. Otherwise just stay out of the thread and mind your own business.

[/ QUOTE ]
I have to agree that everybody seems to be giving him a pretty hard time for organizing a competition that does not break any rules/laws, but on the other hand I can understand some of the anamosity toward the fact that he is the author of WinHoldem (aka Ray E. Bornert II).

Juk [img]/images/graemlins/smile.gif[/img]

[/ QUOTE ]

juk,

your comment is understandable. but if you think about it
there isn't any way to have a pbwc unless there are real
pro pokerbotters that use a virile platform. if strong
human pros enter and play then you're gonna need strong
pro botters that are worthy opponents.

it will mean much more to a tv audience if they know
and understand that the bots facing the humans are
very serious virile bots at the professional level.

the audience just simply will not care about sterile
bots and if the pbwc isnt at least as interesting as
any other televised professional poker event then it
probably doesn't have a future beyond the first event.

it will be much more interesting to hear the announcer
say
"xbot has won tournaments as high as $40k online"
rather than
"xbot beat the entire math club last year"

ray

[RayBornert]

deuce,

just wanted to say thank you for
providing some balance here.

ray.

[disjunction]

[ QUOTE ]


Even after reading every paper published by the group it is not quite possible to repeat their work and thus validate their results...



[/ QUOTE ]

Can you elaborate? As you may know, the problem with publishing Computer Science papers is that you're limited to 6 pages, or whatever, and you have to include a bibliography and intro and prior work. And for the longer papers you're limited by reader patience.

Usually code is available upon request but of course theirs probably isn't. I've never really seen a paper where I can replicate what the exact results, although hopefully there's enough there to get the picture?

If a published paper has an algorithm with two nested for loops, for example, and everybody previous had used 3, then I can take the intuitive idea and make use of it, even if I can't replicate the exact program. If not, I agree it's a pretty crappy paper.

[jukofyork]

[ QUOTE ]
Can you elaborate? As you may know, the problem with publishing Computer Science papers is that you're limited to 6 pages, or whatever, and you have to include a bibliography and intro and prior work. And for the longer papers you're limited by reader patience.

[/ QUOTE ]
It's not from lack of space or reader patience I'm afraid. The main two examples (which pretty much cripple any attempt to replicate their experiments):

1) The expert defined betting strategy used in their Poki bot was never published and upon trying to repeat their work it quickly become obvious that this is the core constituent of their system (and small flaws in it makes the rest of their Bayesian weight tables and related hand strength formulas presented completely fail). Using just a slightly modified version of this strategy will cause vastly different results from what they have published.

2) They did exactly the same with their SimBot code and neglected to provided the expert system rules it was using to simulate the action of the other players. This has an even greater effect on the stability of their search algorithm than it did on their weight table methods (a tiny alteration of a single parameter will cause the play to change from hyper aggressive to hyper passive, etc).

I know somebody else tried to replicate the VexBot work, and found that (amongst other missing info) they have used many heuristics and spent a great deal of time making default models (which were not published), as bare-bones VexBot replica took so long to converge it would not play sensible poker for many 100's or 1000's of hands.

As far as I know the SparBot papers are repeatable (I've never tried, but can't remember seeing any problems after reading their papers on it) with several coder's on the Poker Academy forums creating similar bots (one using fictitious play) and another recent attempt to create a NL HU bot using the same ideas.

[ QUOTE ]
Usually code is available upon request but of course theirs probably isn't.

[/ QUOTE ]
All code (including the elusive expert systems and VexBot heuristics) went into the commercial Poker Academy application.

[ QUOTE ]
I've never really seen a paper where I can replicate what the exact results, although hopefully there's enough there to get the picture?

[/ QUOTE ]
Agreed, but since the expert systems and heuristics used are MUCH more important than the rest of their system combined (perhaps as much as 95% of the work), I wouldn't expect to have to spend 5+ years tuning my own attempt to replicate their system.

I have read plenty of papers in my time, but what they have presented is almost like saying "I have this great algorithm which takes all of these parameters... You can have most of the algorithm, but none of the parameters we have found to be useful. Here are our results...". IMHO, that is worthless and shouldn't get published. What is the point in publishing your research if it doesn't actually advance anything for the scientific community in general, basically leaving others to start afresh?

Juk [img]/images/graemlins/smile.gif[/img]

[JackStrap]

Hi jukofyork ,

first betting algorithm for poki was realased in previous paper, it's was in the Papp paper. i think i have a really close poki clone based on my experiment on poki academy 2.5, but you probably know that rule-based bot are far far away from world class play.

[ QUOTE ]

I have read plenty of papers in my time, but what they have presented is almost like saying "I have this great algorithm which takes all of these parameters... You can have most of the algorithm, but none of the parameters we have found to be useful. Here are our results...". IMHO, that is worthless and shouldn't get published. What is the point in publishing your research if it doesn't actually advance anything for the scientific community in general, basically leaving others to start afresh?


[/ QUOTE ]

well i agree that there are a lots of informations hided for the reader in CPRG papers, maybe they keep it private for winning competion and/or for privates interests like poki academy wich i agree are against the usual scientific community methodology

When it's come to poker ai, research seem very "closed" wich make a lot of individual work and a lack of finding in my opinion.... unlike what happen in backgammon with TD-GAMMON or GNUBG for exemple. Maybe everyone thinks that they will be rich when they will find world class algorithm!

But to return to the CPRG have you looked this new paper?
Algorithms and Assessment in Computer Poker.
pretty nice phd dissertation from one of the best thinker in the subject Darse Billing

[jukofyork]

[ QUOTE ]
first betting algorithm for poki was realased in previous paper, it's was in the Papp paper. i think i have a really close poki clone based on my experiment on poki academy 2.5, but you probably know that rule-based bot are far far away from world class play.

[/ QUOTE ]
If I remember correctly they give a very cut down version of thier expert system(s) in each paper (they improve it from Loki to Poki), but it's not what they themselves use (ie: the one defined by Darse Billing).

I too have tried to clone both a Poki style bot using weight tables and also a SimBot style bot. Both were pretty easy to code up, but the really hard part is creating a model of the opponent strategy (bad expert system = bad overall system). My guess is tuning the expert system is 95%-98% of the work involved.

[ QUOTE ]
[ QUOTE ]

I have read plenty of papers in my time, but what they have presented is almost like saying "I have this great algorithm which takes all of these parameters... You can have most of the algorithm, but none of the parameters we have found to be useful. Here are our results...". IMHO, that is worthless and shouldn't get published. What is the point in publishing your research if it doesn't actually advance anything for the scientific community in general, basically leaving others to start afresh?


[/ QUOTE ]

well i agree that there are a lots of informations hided for the reader in CPRG papers, maybe they keep it private for winning competion and/or for privates interests like poki academy wich i agree are against the usual scientific community methodology

When it's come to poker ai, research seem very "closed" wich make a lot of individual work and a lack of finding in my opinion.... unlike what happen in backgammon with TD-GAMMON or GNUBG for exemple. Maybe everyone thinks that they will be rich when they will find world class algorithm!

[/ QUOTE ]
If people want to work on creating their own poker AI then that's fine (and I agree, most are interested in the "get rich quick" aspect, rather than scientific interest), but I don't think it's right to publish academic work with crucial information missing.

Juk [img]/images/graemlins/smile.gif[/img]

[RayBornert]

fyi,

I wrote an article that discusses the 6 different levels of
bots as defined by pro botters. i refer you to
specifically to the definitions for levels 4-6.

given that most if not all pbwc entrants will be
meeting for the first time in 2007, it's safe to
say that no entrant would have any strong profile
info for any given opponent (for a moment, allow
me to ignore the case where two of the entrants
happen to know the playing habits of the other
very well).

so for this first year, the best that any entrant
could do out of the gate is level 4 and then over
time - level 5 - as the current session progresses.

any bot that was heavily dependent on level 5 or 6
profile info might be at a disadvantage if it did
not also have a strong level 4 foundation. level 4
is the best you can initially do against a table of
strangers.

if the pbwc gains popularity such that official
qualifiers begin to occur on a world wide basis,
then the community as a whole would be slowly
accumulating a statistical impression of everyone
that had ever played in any official pbwc related
event.

so the point is that during the early years the
level 6 bots would be starved for stats but over
time they'd have an ever growing history to sift
through.

fast forward a few years to a tv event where the
announcers are well versed in poker tracker style
statistics and are providing color commentary very
much like a major league baseball announcer might
discuss the stats of a batter.

ray

[piranha]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
The OP of this thread's intentions might be questionable, but rather than be the judge I will move it from B&M to Software where it rightfully belongs. There is a possibility the mods in software lock this thread, as poker bots tend to be a sore subject for any people for very obvious reasons.

TT [img]/images/graemlins/club.gif[/img]

[/ QUOTE ]

tt,

the reason bots are a sore subject with some is because
of the expectation that the online game should be quite
identical to the live game in every way. the wisest of
us understand that's not possible. the children among
us continue to argue that it is.

i love the live game, always have always will.
and that's why the pbwc is a live digital event
and not an online event.

the live game is much more respected than the online game.

ray

[/ QUOTE ]

Ray:

advice - you don't need to be a salesman, nobody trusts salesmen when they push the issue too far. I told you in my prior post that I am not judging your intentions, I know there are legitimate reasons for poker bots, however from your posts I don't think your reasons are 100% legitimate.

I should add that if the US eventually permits poker online and starts legitimate regulation, I will be first in line to lobby to make it a criminal offense to use third party tools that may be used to cheat through the placement of Trojans, or via tools that make automatic decisions for the player ala a bot. In short, I hope that someday your actions become a felony criminal act but as it stands today I won't judge you for hosting a poker bot conference/contest.

TT [img]/images/graemlins/club.gif[/img]

[/ QUOTE ]

Amen

[RayBornert]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
The OP of this thread's intentions might be questionable, but rather than be the judge I will move it from B&M to Software where it rightfully belongs. There is a possibility the mods in software lock this thread, as poker bots tend to be a sore subject for any people for very obvious reasons.

TT [img]/images/graemlins/club.gif[/img]

[/ QUOTE ]

tt,

the reason bots are a sore subject with some is because
of the expectation that the online game should be quite
identical to the live game in every way. the wisest of
us understand that's not possible. the children among
us continue to argue that it is.

i love the live game, always have always will.
and that's why the pbwc is a live digital event
and not an online event.

the live game is much more respected than the online game.

ray

[/ QUOTE ]

Ray:

advice - you don't need to be a salesman, nobody trusts salesmen when they push the issue too far. I told you in my prior post that I am not judging your intentions, I know there are legitimate reasons for poker bots, however from your posts I don't think your reasons are 100% legitimate.

I should add that if the US eventually permits poker online and starts legitimate regulation, I will be first in line to lobby to make it a criminal offense to use third party tools that may be used to cheat through the placement of Trojans, or via tools that make automatic decisions for the player ala a bot. In short, I hope that someday your actions become a felony criminal act but as it stands today I won't judge you for hosting a poker bot conference/contest.

TT [img]/images/graemlins/club.gif[/img]

[/ QUOTE ]

Amen

[/ QUOTE ]

pirahna,

online holdem is a different game than live holdem.
the two games should be treated differently.
i love both games equally well.
but i have different expectations for each.
and you should too.

i have a very difficult time with those who
demand that they be the same. tt obviously
does since he wants it to be a felony for the
online game to be different. i assume you do
too since you blessed his comment.

so i'm going to take this opporuntity to school
you both in exactly how the internet forced
changes to classic holdem:

1) the muck is no longer guaranteed to be forgotten

in the live game where good dealers know how to
reconstitute the deck at the end of a hand, the
cards remain face down all through the shuffle
thus making it impossible for any one entity
to know the entire history of the hand. each
player may know their part of the muck but
nobody knows all of it.

the online game pretty much guarantees that the
game server will have permanent knowledge of the
entire muck. there is no way that a game server
can prove that the remembered muck stays private.
sites could solve this issue by forcing it to be
public but they dont.

2) sites do not prove their deck selection is random
- they could do this with player determined shuffling
but only if they publish the muck at the end of each
hand. therefore we have to just trust them.

3) sites cannot physically see their players and
neither can the other players.

this is the single biggest change to the game
because all manner of cool activities become possible
from the privacy of your home that are not possible
in a live game and neither your site or your
opponents can stop you. this is the thing you
guys need to make peace with because it's part
of the game whether you like it or not.

4) there's no way to prove that the server is
not colluding with one of your opponents. the
connection must be encrypted to secure the
card info and therefore it's impossible for
a 3rd party policing organization to put a
sniffer on the line and verify that the server
didn't help any given player. the facts are
that the internet makes it impossible to both
secure the private transmission of card info
while at the same time proving that the server
isn't colluding with a player. this is not a
solvable problem on the internet.

what really disturbs me about your view and attitude
is that you seem to ignore these issues to the point
that you're willing to put people in prison for not
behaving the way you want them to in the privacy of
their own home.

when you lobby to have me put in prison for using
doing things differently than the live game please
remember to also try to get sites put into prison
for also doing things differently. fair is fair.

privacy issues trump all of your needs to have the
online game be what you want it to be. your over
zealous need to police online holdem is an affront
to the very fabric of the u.s. constitution.

pull your heads out of your southern exits and use
some reason and make peace with yourself and the
universe that it is physically !@#$% impossible for
the internet to secure the live game that we all love.

the live game is what it is.
the online game is what it is.
go meditate on a rock somewhere
and make peace with the poker gods
and let it go.

now can we please limit the discussion of this thread
to what can actually be secured with live digital?

ray bornert

[matrix]

[ QUOTE ]

fast forward a few years to a tv event where the
announcers are well versed in poker tracker style
statistics and are providing color commentary very
much like a major league baseball announcer might
discuss the stats of a batter.


[/ QUOTE ]

That is never going to happen unless and until bots gain public acceptance - and online poker players the world over are happy playing against bots.

I'd like to see the bot championship - purely from an intrigued intellectual POV - much the same as I enjoyed the Kasparov Deep Blue contest.

Bots will kill online poker as we all know and love it - you know this to be true.

People are happy to sit in their living rooms and watch Kasparov vs a Computer - they have this illusory idea in their heads than man is better than machine - when realy this depends on what the man/machine is doing - machines are much better at some things.

They are happy to watch this as they think Kasparov is a "super human" - they wouldn't want to try playing Deep Blue themselves - they know they'd lose.

What do you think will happen when 5-10 years down the road when no human poker player will be able to beat bots of that era, technology and s/w advances faster than most other things.

fx: "The over zelous sports announcer amid flashy lights and rock music announces - and welcome to grand finalist x we can't reveal his true identity but x plays on PokerStars and has one the Stars Sunday $billion TWICE this year - and welcome to finalist y - we also can't reveal her identity but y plays at Full Tilt - where she plays $1k/$2k small stakes games (inflation is a [censored] [img]/images/graemlins/wink.gif[/img] ) and made over $4million last month alone - and finally welcome to Phil Ivey yadda yadda yadda...."

What do you know Phil gets wiped out in 8th place and the superbots once again take all 5 top spots.

Who'd want to play in an online arena that's knowingly bot infested when they watch bots take the pros to pieces in such an event regularly?

These days bots are more prevalent than most players imagine, fortunately they're mostly rubbish and even us mere mortals can beat them [img]/images/graemlins/grin.gif[/img] - fast forward 10 years - and this won't be the case.

Please stop trying to kill online poker by morphing it into some kind of playing field where the commercial enterprises you have sunk lots of time and effort into, which are presently disallowed by the Poker sites, become prevalent.

Can't you see that overfishing is bad for everyone?

I am a real person - I want to play poker from the comfort of my own home vs other real people - I have absolutely zero interest in playing against bots.

By all means develop poker bots - invent and refine cool new technologies - have a world championship that pits bots vs humans great idea.

But don't encourage people to break poker sites t&cs - don't bedazzle prospective wh users with EULA eulogies and fancy talk about "fish traps" and all that crap.

Hopefully before too long the US will regulate online gaming - the new sites that start up will have much stronger EULA's like some others I would welcome it if botting (whilst the bot is pretending to be a real human player) was made into a felony.

I am a man not a machine I want to play with other men and women - if I wanted to play bots I'd still be playing videogames.