Bot Results - Interesting Site Comparison
BeavisChrist
Like it or not, Bots are used more often than you think online. Over the past 8
months I developed a bot more as a proof of concept than anything else - also to
prove once and for all that solid play over hundreds of hours results in winning
poker. The bot really isn't all that smart, and any player paying attention
should be able to destroy it in a limit game. It does some things very well
though - including never going on tilt and always playing draws only when pot
odds dictate it is correct. Remember, this is more of a research project than
anything else - I already know how the world feels about bots in general,
although I don't really understand why. A bot can never... ever be as good as a
human player. If you had a table full of bots you'd make tons of money.
Back in August I turned the bot loose on several sites in the .25-.50 limit
game. The results are fascinating. I won't mention which sites it has been
playing, but I'll designate them as:
Site 1: Ten handed game, fairly busy site
Site 2: Nine handed game, moderately busy site
Site 3: Nine handed game, site barely has traffic but enough to play one or two
tables
Summary statistics
Site 1: 923 Total Hours - 130,000 hands played - net profit $924.71 (exactly
2BB/hour)
Site 2: 366 Total Hours - 59,000 hands played - net loss $75.35 (losing a little
less than half BB/hour)
Site 3: 33 Total Hours - 5,000 hands played - net loss $141.75 (losing 8.5
BB/hour)
Remember, this is a bot, playing EXACTLY the same at each site. I have logged
all of the hands in Poker Tracker for analysis. This is where things really get
interesting.
Avg Players Per Hand
Site 1: 8.31
Site 2: 8.18
Site 3: 8.24
Flop %
Site 1: 46.67%
Site 2: 37.99%
Site 3: 38.65%
Avg Rake
Site 1: 0.17
Site 2: 0.04
Site 3: 0.13
Play Stats
Site 1: VP$IP: 14.97% - PFR: 5.83% - WTSD: 28.95% - W$SD: 58.27%
Site 2: VP$IP: 14.67% - PFR: 5.84% - WTSD: 26.75% - W$SD: 58.63%
Site 3: VP$IP: 15.14% - PFR: 5.88% - WTSD: 28.04% - W$SD: 50.27%
The play stats are almost identical - the only statistically significant
difference is Site 3 seems to win at showdown less than the other two. In
reviewing hand histories, there are sessions where the bot has gone to the river
as a big favorite 20 times and lost all 20 hands, resulting in a big losing
session. It's bordering on hand manipulation, or a hacked site. Raising with
AKs, getting J5o to call, flopping AK2, J5o calling all the way for runner 3, 4.
This happens frequently at Site 3, and almost never at Site 1.
Even though Site 1 is ten handed, the number of players per hand is almost the
same as the other two sites. The big difference is flop %. Site 1 shows 8% more
than the other two. For a minute let's just throw out the results of site 3
since it hasn't played nearly as much as the first two.
Is it possible that having one less person seeing the flop (i.e., tighter game)
can mean the difference between earning 2 BB/hour and being a losing player?
Also note that Site 1 is charging, on average, more than FOUR TIMES the amount
of rake as site 2. Logically you'd think this would negatively affect win rate
at Site 1 - yet it's still performing at a much more profitable clip.
I can't really explain why Site 2 is showing a loss. Most of the statistics are
identical to Site 1. One thing that stands out a little is that generally it
wins small pots on Site 2 and loses the huge ones. Even after flopping a set of
aces and building a gigantic pot it will lose to a crappy flush or whatever.
This is not a rant about losing money - the money is insignificant. Believe it
or not I'm actually interested in proving/disproving the theories around online
sites being rigged. Clearly a poker site has much to be gained by keeping the
money even amongst the players, not allowing the more skilled players to take
the money from the crappy players. This is especially true of the smaller sites
trying to establish a consistent user base.
A human player could never create statistically meaningfull data - humans tilt,
play differently, go on "feel", etc. This bot will play exactly the same over
thousands of hours.
At some point I'll have a few hundred thousand hands logged for 6 different
sites, and I plan to publish the results.
For those wondering, generally the bot plays strategy straight out of SSH for
loose, passive games. Does anyone else consider this information valuable? Would
it be more valuable at a higher limit, such as 1-2?
_______________________________________________________________
Your Online Poker Community - http://www.recpoker.com
mo_charles
sites are using bots? seems like you could have just as easily lost some
hands to poor/drunk players.
mo_charles
----
looking for a better newsgroup-reader? - www.recgroups.com
BeavisChrist
I haven't concluded that the sites are using bots - but I've definitely spotted
several player created bots. What I'm interested in is the idea that sites will
stack the deck and create more winning situations for people who play badly in
order to keep the money basically "even".
One thing I'd like to analyze eventually is the standard deviation in
wins/losses per site for every player. If a site is "evening out" the losses it
should be obvious - no players will have statistically significant wins or
losses. Everyone will come out basically even after several hundred thousand
hands - which will actually amount to everyone coming out as a slight loser
after the rake is paid :)
As far as losing hands to poor/drunk players - that's definitely happening. The
problem is, playing strong hands aggressively and being able to lay down hands,
in the long run, should win money in a limit game. Going to the flop as a big
favorite the majority of the time should end up being profitable. If the bot
plays 100,000 hands at a given site and doesn't show a profit, but at another
site is showing a 2BB/hour win rate - that raises my suspicion.
On Nov 22 2005 10:28 AM, mo_charles wrote:
> an interesting post to say the least. why have you concluded that the
> sites are using bots? seems like you could have just as easily lost some
> hands to poor/drunk players.
>
> mo_charles
_______________________________________________________________
The Largest Online Poker Community - http://www.recpoker.com
Aardvark
> Like it or not, Bots are used more often than you think online. Over the past 8
> months I developed a bot more as a proof of concept than anything else - also to
> prove once and for all that solid play over hundreds of hours results in winning
> poker. The bot really isn't all that smart, and any player paying attention
> should be able to destroy it in a limit game. It does some things very well
> though - including never going on tilt and always playing draws only when pot
> odds dictate it is correct. Remember, this is more of a research project than
> anything else - I already know how the world feels about bots in general,
> although I don't really understand why. A bot can never... ever be as good as a
> human player.
Absolutely... and a computer can never... ever be as good a chess
player as a human player. At least that's what people used to say.
> If you had a table full of bots you'd make tons of money.
No, you would lose money since the site is taking a rake and your bots
are simply transferring their money around to one another.
<<<SNIP>>>
> I can't really explain why Site 2 is showing a loss. Most of the statistics are
> identical to Site 1. One thing that stands out a little is that generally it
> wins small pots on Site 2 and loses the huge ones. Even after flopping a set of
> aces and building a gigantic pot it will lose to a crappy flush or whatever.
Did you actually go through all 59,000 hands? I'm a little wary of
anyone who claims that something like this "generally" happens. I would
be willing to believe that you found that pot sizes tended to be small
wins/big losses since you could easily query the data to determine
this. Adding in that it loses with a set of aces against a "crappy
flush" and implying that this is the cause wreaks of bias.
-Tom.
Chris in Texas
On Nov 22 2005 12:12 PM, BeavisChrist wrote:
>
>
> <u>Summary statistics</u>
> Site 1: 923 Total Hours - 130,000 hands played - net profit $924.71 (exactly
> 2BB/hour)
> Site 2: 366 Total Hours - 59,000 hands played - net loss $75.35 (losing a
> little
> less than half BB/hour)
> Site 3: 33 Total Hours - 5,000 hands played - net loss $141.75 (losing 8.5
> BB/hour)
>
If I'm interpreting these stats incorrectly please let me know, but to me it
seems you're playing at 8-handed tables and playing ~150 hands per hour on each
site? Seems a bit high for almost full tables.
Chris
_______________________________________________________________
The Largest Online Poker Community - http://www.recpoker.com
BeavisChrist
You misunderstood. I meant if you are the only human player at a table full of
bots, YOU'D make tons of money because the bot patterns and play would be easy
to figure out. Also note that in general, poker is considered a vastly more
complex game than Chess. One developer working part time to create a bot doesn't
quite compare to hundreds of engineers contributing to Deep Blue. But I'm
flattered at the comparison ;)
I have gone through *most* of the hands. Note that it's not necessary to go
through 59,000 hands, only through the hands that have won or lost money, along
with targeting specific trouble hands in limit such as A9s. I'm not prepared to
post the results of all hands, the one I used is just an example. Of course one
single beat on one hand is meaningless.
You say it wreaks of bias - how about a theory as to why a computer controlled
player, using the exact same decision logic, playing exactly the same way would
result in such drastically different results from one site to the next? How is
that possible? It's not like there are a bunch of sharks playing .25-.50 on one
site, but not on another. The gameplay stats are showing the play to be very
similar from site to site.
Forget individual hands, individual sessions, or even individual weeks. Let's
talk LONG TERM. How do we account for the difference?
On Nov 22 2005 10:49 AM, Aardvark wrote:
> BeavisChrist wrote:
> > Like it or not, Bots are used more often than you think online. Over the
> > past 8
> > months I developed a bot more as a proof of concept than anything else -
> > also to
> > prove once and for all that solid play over hundreds of hours results in
> > winning
> > poker. The bot really isn't all that smart, and any player paying attention
> > should be able to destroy it in a limit game. It does some things very well
> > though - including never going on tilt and always playing draws only when
> > pot
> > odds dictate it is correct. Remember, this is more of a research project
> > than
> > anything else - I already know how the world feels about bots in general,
> > although I don't really understand why. A bot can never... ever be as good
> > as a
> > human player.
>
> Absolutely... and a computer can never... ever be as good a chess
> player as a human player. At least that's what people used to say.
>
> > If you had a table full of bots you'd make tons of money.
>
> No, you would lose money since the site is taking a rake and your bots
> are simply transferring their money around to one another.
>
> <<>>
>
> > I can't really explain why Site 2 is showing a loss. Most of the statistics
> > are
> > identical to Site 1. One thing that stands out a little is that generally it
> > wins small pots on Site 2 and loses the huge ones. Even after flopping a set
> > of
> > aces and building a gigantic pot it will lose to a crappy flush or whatever.
>
> Did you actually go through all 59,000 hands? I'm a little wary of
> anyone who claims that something like this "generally" happens. I would
> be willing to believe that you found that pot sizes tended to be small
> wins/big losses since you could easily query the data to determine
> this. Adding in that it loses with a set of aces against a "crappy
> flush" and implying that this is the cause wreaks of bias.
>
> -Tom.
_______________________________________________________________
Block Lists, Favorites, and more - http://www.recpoker.com
BeavisChrist
hands per hour.
On Nov 22 2005 10:56 AM, Chris in Texas wrote:
>
>
> On Nov 22 2005 12:12 PM, BeavisChrist wrote:
>
> >
> >
> > Summary statistics
> > Site 1: 923 Total Hours - 130,000 hands played - net profit $924.71 (exactly
> > 2BB/hour)
> > Site 2: 366 Total Hours - 59,000 hands played - net loss $75.35 (losing a
> > little
> > less than half BB/hour)
> > Site 3: 33 Total Hours - 5,000 hands played - net loss $141.75 (losing 8.5
> > BB/hour)
> >
>
> If I'm interpreting these stats incorrectly please let me know, but to me it
> seems you're playing at 8-handed tables and playing ~150 hands per hour on
> each
> site? Seems a bit high for almost full tables.
>
> Chris
_______________________________________________________________
Your Online Poker Community - http://www.recpoker.com
Canadian Monkey
Very interesting post. I am at that stage in life where I am not willing to
discount that anything is possible and will be very curious to see your follow
up studies. Post note - Until the bots get better then the programmed players on
Wilsons simulator programs, I will stay more worried about collusion then
anythng else.
_______________________________________________________________
* New Release: RecPoker.com v2.2 - http://www.recpoker.com
Aardvark
> You misunderstood. I meant if you are the only human player at a table full of
> bots, YOU'D make tons of money because the bot patterns and play would be easy
> to figure out.
Sorry for the misunderstanding.
> Also note that in general, poker is considered a vastly more
> complex game than Chess. One developer working part time to create a bot doesn't
> quite compare to hundreds of engineers contributing to Deep Blue. But I'm
> flattered at the comparison ;)
My point is that at one time people thought that it was inconceivable
that a computer could ever beat a really good human player at chess.
People once thought that a 10MB hard drive would be more than anyone
would ever need. People once thought that a human being could never run
a 4-minute mile. You have to be pretty careful when using the term
"never".
> I have gone through *most* of the hands. Note that it's not necessary to go
> through 59,000 hands, only through the hands that have won or lost money, along
> with targeting specific trouble hands in limit such as A9s. I'm not prepared to
> post the results of all hands, the one I used is just an example. Of course one
> single beat on one hand is meaningless.
Ok, that makes sense. Once you narrow those hands to hands played out
etc. I can see where it would become a reasonable number of hands to go
through. My thought however, is that if the number of hands is small
enough for a human to go through then it's probably not large enough to
be statistically significant.
> You say it wreaks of bias - how about a theory as to why a computer controlled
> player, using the exact same decision logic, playing exactly the same way would
> result in such drastically different results from one site to the next? How is
> that possible? It's not like there are a bunch of sharks playing .25-.50 on one
> site, but not on another. The gameplay stats are showing the play to be very
> similar from site to site.
I can't really say. I'm not trying to claim that your conclusion is
wrong (or hypothesis if it's not yet a conclusion). It just sounded an
aweful lot like those players who can't find a reason for their
continued losses except that they're "unlucky" and they pull out one
hand to "prove" that they're unlucky. I've acquired an automatic
response mechanism to those at this point. :-)
> Forget individual hands, individual sessions, or even individual weeks. Let's
> talk LONG TERM. How do we account for the difference?
I'd say get a bit more data, which you seem to be doing. Without seeing
the data it would be hard for me to say what the cause is, but I'll
give some thought to the possibilities.
-Tom.
Aodhan
> complex game than Chess.
By who??
Chess is incredibly more complex than poker.
Aodhan
______________________________________________________________________
RecGroups : the community-oriented newsreader : www.recgroups.com
Mrs. LHE
On Nov 22 2005 2:32 PM, Canadian Monkey wrote:
>
> Very interesting post. I am at that stage in life where I am not willing to
> discount that anything is possible and will be very curious to see your follow
> up studies. Post note - Until the bots get better then the programmed players
> on
> Wilsons simulator programs, I will stay more worried about collusion then
> anythng else.
>
From what I understand, a bot can only be as good as the human who programs it.
Right, Beavis?
In any case, I'm surprised you've been able to do this without detection from
the sites your bot is playing on. There was a discussion here last year about
PartyPoker scanning players' computer screens while they're playing to make sure
a bot program isn't running (at the time, WinHoldEm was the bot program in
question). A few players allegedly got their accounts closed (and I believe
their money confiscated) by Party because of this. Oh, and also a site will
become suspect if someone plays for long periods of time without sitting out.
I can't imagine Party is the only poker site out there that actively searches
for bot programs, but apparently they just may be.
- Mrs. E
_______________________________________________________________
Posted using RecPoker.com v2.2 - http://www.recpoker.com
FellKnight
> > to figure out. Also note that in general, poker is considered a vastly more
> > complex game than Chess.
>
> By who??
>
> Chess is incredibly more complex than poker.
>
> Aodhan
By programmers ;)
Fell
--
Visit http://www.fellknight.com for strategy, blog, reviews and more!
(STILL IN BETA MODE)
_______________________________________________________________________
BeavisChrist
It's true that many sites have "bot detection" logic. For these reasons I've
developed a bot that has no footprint on the machine it's running on. Basically
running as a hidden windows service with no UI. Taking a screenshot would reveal
nothing. I also limit the bot playing to random periods that are reasonable
amounts of time. Playing three tables for 4 hours at a time is way below the
"bot radar" on the sites. Humans are playing 8 tables for 10 hours (read some of
the posts here on RGP).
I'd say the bot I've developed is right up there with the Wilson simulated
players - likely even more intelligent. It actually adjusts in real time to
changing table conditions (over aggressive play, etc.). It might be an
interesting exercise to pit it against Wilson to see how it does against the
best simulated players. One thing it doesn't do at this point is player
profiling - I'm not sure if Wilson does that or not.
_______________________________________________________________
The Largest Online Poker Community - http://www.recpoker.com
BeavisChrist
Hehe, nice. And true. Programming a reasonable chess game is much easier than a
decent poker player. Chess has a much (and I mean MUCH) smaller set of
"possible" moves - and each move can be played out considering what your
opponent's counter will likely be. Poker (especially no limit) is infinitely
more complex.
I can't remember the web site, but I once read an article written by a chess
grand master who decided to take up poker. After a few years he came to the
conclusion that poker is more complex than chess. Of course that was just his
opinion (and now heresay from me).
But yes, I can say from a programming perspective, poker is very tough. High
school students can program a chess game.
On Nov 22 2005 11:59 AM, FellKnight wrote:
> On Nov 22 2005 11:49 AM, Aodhan wrote:
>
> > > to figure out. Also note that in general, poker is considered a vastly
> > > more
> > > complex game than Chess.
> >
> > By who??
> >
> > Chess is incredibly more complex than poker.
> >
> > Aodhan
>
> By programmers ;)
>
> Fell
> --
> Visit http://www.fellknight.com/ for strategy, blog, reviews and more!
> (STILL IN BETA MODE)
_______________________________________________________________
Watch Lists, Block Lists, Favorites - http://www.recpoker.com
Aardvark
> > to figure out. Also note that in general, poker is considered a vastly more
> > complex game than Chess.
>
> By who??
>
> Chess is incredibly more complex than poker.
Chess is a game of complete information (both players know where all of
the pieces are at all times). Since poker is a game of incomplete
information (you don't know what your opponent's hand is) it brings its
own set of complexities when trying to model it in a computer
application.
Also, with poker your play has to take into account your opponent's
play. Whether or not you want to reraise may depend on whether or not
your opponent is a very tight player or a maniac. With chess you don't
really have this problem. Given any setup of the pieces there is
typically a theoretical "best" move. Of course, if you are against a
weak player you might want to try to set up simple traps that you
wouldn't against a grand master who is sure to avoid them and against a
player who is not so good at complex positions you might want to play
for those kinds of positions, but the effect of this is much less than
the same effect in poker of playing your specific opponent.
Of course, chess has its own complexities of course - early in a game
the choice between two moves has to take into account things such as
control of the center, mobility of pieces, etc. which are also hard to
model in a way in which they can be compared accurately. For example,
just look at a typical gambit... a grand master can make a good guess
at whether or not the sacrifice of the piece is worth some intangible
(for lack of a better word) advantage in position. How do you quantify
this so that a computer can make the same decision correctly?
-Tom.
jacksup
be ahead. Look at your overall results for all three sites. You're
winning 1.07 big bets/hour. Right where you should be.
Matt
Shrike
overall losing streak on a site. Numerous possible reasons for this.
One, a winning player can easily be down in cash after 60k hands. It
sounds huge, but 60k STILL isn't that large a sample. Two, as you
said, if a player suspects you're running a bot it would be relatively
simple for him to defeat it if he could figure out the algorithm. It
would only take the loss of a few big pots to seriously skew the data.
Finally, and no offense intended, but the bot may not be as good as you
think. This is not a slight of your coding ability, as I'm fully aware
of the complexities involved in programming one of these buggers. But
playing straight pot odds, even adjusting for table conditions, is
unlikely to net much more than 1-2 BB/100. As Jacksup said,
assimilating all of your results shows a 1BB/100 profit, which is about
what I'd expect for this strategy.
I commend your efforts here though, and I have a great deal of respect
for people that are able to generate programs like this. It is an
impressive feat to be able to code something like this that shows a
steady profit. You're right that it defiintely shows that solid
consistent poker is a winner. I'd love to see more data from your
future trials.
uglykidjim
You should try to hit every major site on the Internet for at least 500,000
hands (a million if that's feasible). While you've got the bots running, you
should spend time getting ahold of some poker magazines, and probably 2+2
Publishing, to see if anyone's interested in printing your work. You may even
want to consider some non-poker publications.
When you've finally got your data sorted and your hypotheses written out, you
should try to get it all out there in the public RGP isn't a big enough forum.
Whether your results point to rigging or not (or any other phenomena), they will
be a significant study, and they may even have an effect on online cardrooms'
business.
Porsche_Dan
Try your bots and SNGs, play chips versus real money. My results indicate
identical behavior which would mean the play chip games are rigged as well.
_______________________________________________________________
New Feature: Mark All As Read! - http://www.recpoker.com
David Nicoson
> From what I understand, a bot can only be as good as the human who programs it.
> Right, Beavis?
I don't think that's true. It's entirely possible for a bot to do the
math better than a human. (It would seem its advantage could be
greater at stud than hold'em.) It's also entirely possible for the bot
to keep track of the past moves of the other players better. Plus the
ability to avoid tilt.
> In any case, I'm surprised you've been able to do this without detection from
> the sites your bot is playing on. There was a discussion here last year about
> PartyPoker scanning players' computer screens while they're playing to make sure
> a bot program isn't running (at the time, WinHoldEm was the bot program in
> question).
I suspect the easiest thing to do is to run the bot on a different
computer, which just accesses the UI on the client.
> A few players allegedly got their accounts closed (and I believe
> their money confiscated) by Party because of this. Oh, and also a site will
> become suspect if someone plays for long periods of time without sitting out.
That seems easy enough to work around. The bot can sit out. Another
bot can run second shift, with some random noise.
Michael Harrington
can be programmed to randomly alter its play to prevent
anyone getting a handle on it!
A delux bot would continually collect and colate opponents'
play and store that data for future use. One of the unsung
advantages of a "good" bot it that it never makes bone headed
blunders. Was that a straight I just folded?
As far as prevention goes, I know some sites alter the
pixellation on the screen to make "screen scrapers" life
a misery.
"BeavisChrist" <4308...@recpoker.com> wrote in message
news:1132683173$674...@recpoker.com...
mo_charles
> advantages of a "good" bot it that it never makes bone headed
> blunders. Was that a straight I just folded?
this goes both ways. a human becomes aware of the mistake, but a bug in
the program will repeat mistakes forever.
i built a black box trading program a few years ago and passed it to
another trader to manage. he didn't pay attention to what it was doing,
so when he made some change in the stocks he wanted it to trade, forgot an
important update in the code. the result was an illegal short position of
over 100k shares - $2.5 million - which cost me about $50k to unwind.
mo_charles
----
* kill-files, watch-lists, favorites, and more.. www.recgroups.com
Teddy Silvetti
tyslothrop
Kinnipak
> shut up beavis.
someone had to have the guts to say that..
-------
Mrs. LHE
On Nov 22 2005 4:33 PM, tyslothrop wrote:
> shut up beavis.
>
>
You said up. Heh heh heh hehheh heh hehheh hehheh.
- Mrs. E
Michael Sullivan
> Of course, chess has its own complexities of course - early in a game
> the choice between two moves has to take into account things such as
> control of the center, mobility of pieces, etc. which are also hard to
> model in a way in which they can be compared accurately. For example,
> just look at a typical gambit... a grand master can make a good guess
> at whether or not the sacrifice of the piece is worth some intangible
> (for lack of a better word) advantage in position.
Typical gambits are a bit more straightforward than that. Standard
opening gambits sacrifice a pawn for a tempo, meaning that an extra
piece is able to be deployed early. Reasonable trade. You can get that
into a computer's evaluation fairly simply by making a deployed piece
worth just barely more than 1 pawn of material.
Of course, exactly how the piece is deployed makes a small difference
that is hard to quantify but you don't need a perfect evaluation
function when you can look ahead much broader and deeper than your
opponent, just a decent one.
More spectacular gambits that lose whole pieces, if they are a good
idea, almost always lead to some set of forced sequences which either
mate or recapture the material. A human player may only know such an
outcome is very likely based on an overwhelming concentration of force
on a weak point, but a computer that looks deeply may well be able to
work out the line all the way to a capture.
With the exception of pawn structure, positional advantages tend to last
only as long as you maintain the initiative. If, having sacrificed a
piece, you fail to recapture your material (or better) before giving up
the initiative, you have generally lost in the exchange, because it's
rare that a long-term positional advantage is worth much more than a
pawn. Occasionally you trap an important piece for more than two turns
which may give you multiple tempos -- this just means that your forced
lines can contain an initiative giving move or two and still hold up.
I'm pretty sure a brute force alpha beta search which considers only
material strength and a weighted function for the value of squares under
attack in the evaluation, and running on a typical modern PC would be
stronger than the majority of tournament players and probably be rated
in Class A.
An equivalently simple algorithm for playing poker is likely to get
slaughtered by almost any winning online or b&m players, just like the
equivalently simple algorithms for playing go couldn't even beat
intelligent beginners.
That said, I don't think that the size of the search tree, or how easily
computers beat a game is the right measure for how deep the game is.
You can design games that are completely impossible to brute force
search, but that are quite simple (and boring) to play or solve.
Michael
Shrike
programs it.
Right, Beavis?"
I wrote an Othello program that can kick my ass. Sure, I have to know
the algorithms for good play, but the computer can just crunch through
thousands of positions many moves ahead. I can't hope to compete. I
imagine the same is true for a poker bot that operates on pure math.
Luckily poker is more than just math.
BeavisChrist
On Nov 22 2005 1:26 PM, Teddy Silvetti wrote:
> HOw can i contact u via email?
>
>
_______________________________________________________________
* New Release: RecPoker.com v2.2 - http://www.recpoker.com
BeavisChrist
The bot does indeed have *some* amount of randomness built in. Check raises
sometimes, but not always, etc.
Player modeling is a whole other thing that is a pipe dream at this point -
enough hours in the day are spent creating rudimentary decision logic (which I'm
planning to completely rewrite already).
As far as "pixelating" the screen - this is something I haven't seen yet. Of
course the bot relies on screen scrape, and I've configured at least 10 sites
for use... none of which have been a problem (and I've hit all the major ones).
> > Your Online Poker Community - /
_______________________________________________________________
The Largest Online Poker Community - http://www.recpoker.com
BeavisChrist
A few points:
- To me, 60K hands is enough to show trends. After 60K hands, Site 1 showed a
net of nearly $500. That's almost a $600 difference. At .25-.50, that is much
more than just a few big pots. The avg pot at that limit is around $3.50.
- It's very unlikely that any player has pegged the bot. From the play, these
folks barely pay attention to their own cards, much less spend any time figuring
out opponent's play. Even with 130,000 hands no single opponent has been at the
same table with the bot for more than 3000.
- I didn't mean to claim that the bot was a great poker player. At best, it
should be moderately profitable. 1BB/100 hands is fine results. It does play
decently though, and should be profitable at the low limit fishy games. Long
term, I want to measure it's effectiveness at each site for comparison
Eventually I'll have a web site dedicated to publishing the results - I'll make
sure and keep RGP posted on progress.
_______________________________________________________________
torx
i have no respect at all for people who have it make money for them tho.
-Alexander Knopf
https://www.allinpoker.com/aip/registration/register.jsp?refCode=AKF040D
https://secure.pokerchamps.com/pokerpublic/arequest?acode=TORX
_______________________________________________________________________
Mrs. LHE
On Nov 22 2005 10:13 PM, torx wrote:
> i also have great respect for people who can write a bot like that.
> i have no respect at all for people who have it make money for them tho.
Why not? I don't see the big deal. (I'm really interested in your answer, by
the way, as many people seem to feel the way you do about it.)
- Mrs. E
>
>
>
>
> On Nov 22 2005 10:43 PM, Shrike wrote:
>
> > I'm not sure why you suspect bias or rigging just because you have an
> > overall losing streak on a site. Numerous possible reasons for this.
> > One, a winning player can easily be down in cash after 60k hands. It
> > sounds huge, but 60k STILL isn't that large a sample. Two, as you
> > said, if a player suspects you're running a bot it would be relatively
> > simple for him to defeat it if he could figure out the algorithm. It
> > would only take the loss of a few big pots to seriously skew the data.
> > Finally, and no offense intended, but the bot may not be as good as you
> > think. This is not a slight of your coding ability, as I'm fully aware
> > of the complexities involved in programming one of these buggers. But
> > playing straight pot odds, even adjusting for table conditions, is
> > unlikely to net much more than 1-2 BB/100. As Jacksup said,
> > assimilating all of your results shows a 1BB/100 profit, which is about
> > what I'd expect for this strategy.
> >
> > I commend your efforts here though, and I have a great deal of respect
> > for people that are able to generate programs like this. It is an
> > impressive feat to be able to code something like this that shows a
> > steady profit. You're right that it defiintely shows that solid
> > consistent poker is a winner. I'd love to see more data from your
> > future trials.
>
>
> -Alexander Knopf
> <a href="https://www.allinpoker.com/aip/registration/register.jsp?refCode=AKF040D" target="_blank">https://www.allinpoker.com/aip/registration/register.jsp?refCode=AKF040D</a>
> <a href="https://secure.pokerchamps.com/pokerpublic/arequest?acode=TORX" target="_blank">https://secure.pokerchamps.com/pokerpublic/arequest?acode=TORX</a>
Raccoon
BeavisChrist wrote:
> session. It's bordering on hand manipulation, or a hacked site. Raising with
> AKs, getting J5o to call, flopping AK2, J5o calling all the way for runner 3, 4.
> This happens frequently at Site 3, and almost never at Site 1.
Nah ... last night here at Site 4, one hand I 3-bet with AKo, 2
callers. Flop K32. Bet and 1 caller. Turn Q. Bet and call. River rag.
Bet/raise(!)/call. Other guy shows JTo. This is not, in my experience,
unusually bad low limit play; I only remember it because it happened
last night.
I've noticed that many of my low limited opponents just love to raise
the turn with an inside straight draw! And bet the river, what the
hell! Sometimes they make it (let's hope the WSOP-TV guys keep saying
"so-and-so picked up a straight draw" when what they mean is a gutshot
or worse).
Besides, your J5 holder had two count'em two ways to make a straight! ;)
Teddy Silvetti
that email didnt work ill try again but heres my message:Hello my name is
teddy I go to st. xavier university in chicago. Im doing a research paper on
online poker , human vs. bot. ive read that uve created a bot , how can i ubtain
one? Please help me out here sir. I apprciate it greatly.
On Nov 22 2005 6:28 PM, BeavisChrist wrote:
> My anon email (recpoker.com) should be in the header
>
> On Nov 22 2005 1:26 PM, Teddy Silvetti wrote:
>
> > HOw can i contact u via email?
> >
> >
>
>
_______________________________________________________________
F8thless
My thoughs:
I calculate the BB/100 for all the hands to be .72 which seems reasonable for a
bot.
I don't think 5000 hands (presumably about 900 Flops seen given the vol.%) is
enough to say that site 3 is rigged.
I calculated an average of 3.9 players seing the flop at 1, 3.1 at 2, and 3.2 at
3. Given that the number of players dealth into a hand is only slightly higher
at the ten handed site (1) vs. the other two nine handed sites, I think this
difference of about 20% fewer bets in pre-flop and then less chance for people
to be in on later rounds would have a significant effect on total profitability,
because your bot would have a lot of pot equity over those extra pf callers on
site 1.
> Is it possible that having one less person seeing the flop (i.e., tighter
> game)
> can mean the difference between earning 2 BB/hour and being a losing player?
I'm sure that is probably the biggest reason for the discrepancy.
I agree that you can't look at someone catching runner-runner to lose a big pot
as being evidence that perhaps that player knew the cards were comming. If
everyone played well, no one would win (but the rake of course)
I think the rake at these limits is usually going to be 5% of the pot, which
will have a huge effect on profitability, and only an extremely loose game, can
be beaten
Do you have data on total rake taken out of pots you won at each site?
I estimate it to be $3,800 on 1, $1,600 on 2, and $125 on 3, am I close?
I don't think the players are at all aware of your bot, but it may have some
holes in it's game (are all the hands it vol. put's $ into the pot w/ +EV?) do
you have a way of calculating the EV of calls in certain situations? etc.
Finally, the VP$IP and PFR% are nearly identical to the way I play (I play a
little bit higher limit HE but basically use the S&M strategy, although have
made a few modifications) I only have 221 hands of LHE HH b/c I just started
using PokerTracker, but Have been playing the same way for a few months over
which I have done better than 2BB/hr, but I only play if the game is very
loose. For the 221 hands I Have, the PF numbers are very close to the bots, but
in those 4 sessions, I think I have had a bit of bad luck w/ flops (in those
hands, no AA or KK, 0/3 w/ AK, 0/1 w/ QQ, and 1/2 w/ JJ) and have a Went to
showdown % of 36% and won money at showdown % of 44% - both of which are
historically higher, but I don't have the exact number. For a total BB/100
hands of -3.14
It's possible that the bot should go to more showdowns, even if it lowers the
win%, the dollar amount may increase. The bot will on average have the best
hand at the table, so it's possible it is not getting enough in with better
hands and not going to showdown with enough of them.
I would be happy to look at some of your summary data (I'm sure you have plenty
more than you posted) and am Happy to share some of my poker tracker stats when
I build more.
I think this is a very interesting project, from a statistical analysis
perspective, and to me suggests that the sites are not rigged.
> session. It's bordering on hand manipulation, or a hacked site. Raising with
> AKs, getting J5o to call, flopping AK2, J5o calling all the way for runner 3,
> 4.
> This happens frequently at Site 3, and almost never at Site 1.
>
_______________________________________________________________
BeavisChrist
I've wondered the same thing. Why are people so against playing against a bot?
At this point the bot isn't making any real money (see original post).
Even if it was, a decent limit player rips it apart without a problem.
> > > > href="https://www.allinpoker.com/aip/registration/register.jsp?refC
> > ode=AKF040D" target="_blank">> >
> > target="_blank">https://www.allinpoker.com/aip/registration/regist
> > er.jsp?refCode=AKF040D
> > /
_______________________________________________________________
Your Online Poker Community - http://www.recpoker.com
BeavisChrist
The only way to obtain a bot is to buy a commercially available one(Winholdem),
or create your own. I'd estimate I've put about 1000 man hours into the one I
developed and it still has a lot of work before it would be something I consider
"complete". I've already turned down legitimate 5 figure offers for the code, so
if $200 for Winholdem is too much then this is definitely out of your price
range.
Someday this may be open source once I have enough results to publish... but not
today :)
_______________________________________________________________
cinnamon sky
perspective, and to me suggests that the sites are not rigged."
Oh really? You would feel comfortable playing at site#3, at which Beavis
states there were sessions (plural) where he went to showdown 20 times as a
big favorite and lost ALL 20 of them? The odds against this happening are
astronomical. If he were, say, a 2-to-1 favorite all 20 times, the odds
against him losing all 20 showdowns are 3,486,784,400-to-1! I don't know
about you, but I don't NEED to see any more hand histories for that site.
And I can't WAIT to find out their name.
Way to go, Beavis!
"F8thless" <4308...@recpoker.com> wrote in message
news:1132721496$674...@recpoker.com...
F8thless
I see what your saying,
I'll admit that it's possible 3 could be rigged, but I would have to see more
data and no more about the bot before you could really be sure. He does have a
lot more observations for the others.
Maybe you could do a moving average of say 50 sessions at a time and calculate
the bb/hr or bb/100 hands on a rolling basis to see if there is similar
variation in other sites.
He also says it has little traffic, so It's probably not one of the major sites
but like I said, I would love to see more data, that might explain the
differences.
> > * New Release: RecPoker.com v2.2 - /
Ian Stuart
> Even though Site 1 is ten handed, the number of players per hand is almost
the
> same as the other two sites. The big difference is flop %. Site 1 shows 8%
more
> than the other two. For a minute let's just throw out the results of site 3
> since it hasn't played nearly as much as the first two.
>
> Is it possible that having one less person seeing the flop (i.e., tighter
game)
> can mean the difference between earning 2 BB/hour and being a losing player?
That 1 extra person can definitely make a big difference - its an 11%
increase in the number of players so it would be a mistake to assume that
it has no impact on the game. One thing to consider is that in limit
Hold'Em the big money generally goes in on the turn and river. You should
therefore compile stats for the average number of players still in the
game on those streets (where there is any betting action), rather than
relying exclusively on the number of players seeing the flop as an
indication of the looseness of the game.Stats for the average pot as BBs
would also be helpful.
________________________________________________________________________
Chris in Texas
On Nov 22 2005 5:36 PM, Shrike wrote:
>
> I wrote an Othello program that can kick my ass. Sure, I have to know
> the algorithms for good play, but the computer can just crunch through
> thousands of positions many moves ahead. I can't hope to compete. I
> imagine the same is true for a poker bot that operates on pure math.
> Luckily poker is more than just math.
In high school I wrote a tic-tac-toe program that'd never lose. It would tie
against those that new the few rules needed to master the game, and beat those
that didn't.
Not that that means anything. Your post made me recall how much I thought I
knew about computers because I could write a program in BASIC language on a
Radio Shack TRS-80 LOL.
Chris
Mrs. LHE
On Nov 23 2005 11:03 AM, Chris in Texas wrote:
>
>
> On Nov 22 2005 5:36 PM, Shrike wrote:
>
> >
> > I wrote an Othello program that can kick my ass. Sure, I have to know
> > the algorithms for good play, but the computer can just crunch through
> > thousands of positions many moves ahead. I can't hope to compete. I
> > imagine the same is true for a poker bot that operates on pure math.
> > Luckily poker is more than just math.
>
> In high school I wrote a tic-tac-toe program that'd never lose. It would tie
> against those that new the few rules needed to master the game, and beat those
> that didn't.
>
> Not that that means anything. Your post made me recall how much I thought I
> knew about computers because I could write a program in BASIC language on a
> Radio Shack TRS-80 LOL.
>
> Chris
I used to write programs all the time on my Commodore VIC-20, complete with
cassette tape drive. (My family didn't have enough money to buy me the
Commodore 64 which I believe had a drive for floppy disks.) Thought I was The
Shit back then. LOL
- Mrs. E
Chris in Texas
On Nov 23 2005 10:09 AM, Mrs. LHE wrote:
>
> I used to write programs all the time on my Commodore VIC-20, complete with
> cassette tape drive. (My family didn't have enough money to buy me the
> Commodore 64 which I believe had a drive for floppy disks.) Thought I was The
> Shit back then. LOL
Cassette tape drives....LOL. I got so good at mine (no $$ for floppy either) I
could tell just by the change in pitch and time elapsed that I was close to the
spot where my wanted program was (i.e., I don't need no stinkin' counter)
Chris
Teddy Silvetti
and play yourself an dmake more. WTF ...5 digit figure su kidding me all i need
it some kinda bot not that winholdem rip off crap.
_______________________________________________________________
* New Release: RecPoker.com v2.2 - http://www.recpoker.com
PokerPleb
address? I'm not having any luck with it...
Tommy S.
wrote:
>haha 1000 hours damn your freaking loser. You could take those 1000 man hours
>and play yourself an dmake more. WTF ...5 digit figure su kidding me all i need
>it some kinda bot not that winholdem rip off crap.
Don't throw stones when you live in a house of rare and very delicate
crystal.
--
Tommy S.
PokerPleb
BeavisChrist, I'd like to send an email to you. I haven't been able to do it via
your recpoker.com address. Would you mind sending an email to me so that I can
reply? My email address is in the header of this post.
Thanks,
Jim S.
_______________________________________________________________
The Largest Online Poker Community - http://www.recpoker.com