Tuesday, August 30, 2005
"If you want to start explaining things that, you, yourself don't have a good understanding of; well, that's why God created RGP."
Thanks for stopping by.
My brain hurts. I spent far too much time rummaging around RGP for posts about online poker bots. I could feel my brain getting softer the longer I spent there.
But, all the same, I found some fun threads for you, gentle reader. This is going to be one hella-scattered post, but if you wanna read about online poker bots, this is the place. Blow off work, damnit, and start surfing away. Many threads for your enjoyment.
This uber-post brought to you by:
Bonus Code IGGY on Party Poker
Golden Palace Poker
Caribbean Sun Poker
Bonus Code IGGY on Party Poker gets you into heaven!
This post is mostly in response to the Wired magazine article:
On the Internet, Nobody Knows You're a Poker Bot
In the booming world of online poker, anyone can win. Especially with an autoplaying robot ace in the hole. Are you in, human?
Shame on you, Wired, for not doing your research. There is plenty more to talk about besides Winholdem. The first serious attempt at a poker bot was ORAC, by Mike Caro, the "mad genius" of poker, in the 80’s. It played respectable, but not great poker in the World Series of Poker against humans. Caro also appeared on the TV version of Ripley's Believe It or Not? with Orac (Caro spelled backwards). This was a $250,000 challenge pitting the computer against Bob Stupak, owner of Vegas World. Stupak won. Mike Caro "mothballed" the project since then because he believes it's futile to try to outplay human pros at no limit holdem with a computer player.
And hell, online poker bots have been discussed on RGP for years to the tune of thousands of posts. I'm going to let the posts by Mr. Winholdem stand for themselves. I think they reveal enough.
Here are just a few to get you started:
pokerbot.com - winholdem - team edition
Here's my response:
Go to hell.
Ray, it's too late. You've already showed yourself completely lacking in ethics, I don't think there's anything you can do to redeem yourself.
But, if you never had included the collusion enabling features we probably would have gone a lot easier on you. Mind you, we still would have mocked you for the poker content of your writings, but I for one wouldn't have been so inclined to refer to you as a scamming a**hole. Also there is the issue of the software "phoning home", if that hadn't been there I also would have considered you to be less likely to be a scammer.
You want to know what *MY* major gripe is about winholdem? It's Ray's posts to RGP.
Seriously, I might have been one of his customers. Brand new to poker. Didn't have a clue. Wasn't really aware of the ethical ramifications of bot play. Was out there buying every book I could get my hands on. Probably would have purchased his product, just thinking it might be something I needed. Might never have gotten around to even trying to figure out how to USE it but probably would have "collected" it nonetheless.
But then I started paying attention to Ray's posts. How rude, obnoxious, unprofessional he was. How illogical, argumentative and contradictory his posts.
So you're partially right. It wasn't about playing against bots. It was simply the human who represents the product.
winholdem - user warning
it seems Party Poker and Paradise Poker have both decided to yet again
invade your rights and have implemented a new detection method which seems to reside on the client (your pc).
we have had several customers over the past few hours complaining that their Party and Paradise Poker accounts have been suspended and funds seized. we apologize for the inconvenience but as it is out of our control what an OPC does with your funds or install on the client we can not be held responsible for their actions. we hope to have a fix available soon.
This thread from 2001 is pretty humorous with David Sklansky getting mighty upset that folks won't take his word about second-hand knowledge about the veracity of this post. Andru Prock has some fun.
Poker Bots on Paradise?
By the way, all of this stuff should be irrelevant to good players. Nine bots that could each make three bucks an hour in a 3-6, would lose sixty bucks an hour to me playing 20-40. I am quite positive that no one is near programming a computer to beat even a fairly tough 10-20 ring game, including U. of Alberta, Mike Caro, Wayne Russell, or Bob Wilson.
winholdem bots and other such programs
These winholdem scumbags are touting the benefits of their bot doing just that...playing as a team. If they were really against team play, they would disable this feature. Instead, not only are they promoting it, but they are billing it as a "premium feature" and reason to buy.
As it stands now, the only ones who stand to lose from the winholdem bots are the idiots who purchase it. Not just because of the poor programming behind the bots, but also because they are trusting people out to cheat poker players with their login info, account info, and are giving them access to all their hole cards. Anyone who uses this program gets exactly what they deserve...a one-way royal screwing.
It's sort of like saying..Hey..If I give you $100, will you install this trojan on my computer for me that lets you cheat me at will? I suppose this is Darwinism at its finest.
"Poker Bots" vs humans, get ready to be trounced
What annoys me about this "The bots are coming!" hysteria is that it doesn't really matter if someone designs a bunch of perfect bots and sticks them on Paradise. Why? Because when you find yourself in a room full of perfect poker players who do not make conversation and play as if they know what each other has, you will realize that the game is not worth playing and leave.
It's analogous to cheating in a casino. Sure, there may be a hidden camera and your opponent may know what your cards are, but if you are a normal person, you will quickly realize that this game is not for you. Did everyone abandon professional baseball in droves when the World Series was rigged?
No. You becomes a little more cautious, a little more cynical, and you move on.
Are we ready for a Poker Bot program?
winholdem - request for comments
This was an independent study done by a respected poster on twoplustwo.com forums anyone can go there and do a search on "Confessions of WinHoldEm User" and find the report. Here is a very good paragraph from that same report.
Watching it "autoplay," I learned that it is easily stolen from. In fact, while I couldn't complain about its initial evaluations of hole cards, its actual game play was awful. I watched it give up blind after blind to a raiser on the button. And if you don't think that's all that bad, calculate out what four or five big and small blinds per hour are. Take that much out of your pocket and dump it in the crapper. Repeat every hour. Yeah. Not pretty is it? I thought so.
Good gravy, I enjoyed this post:
Ok, you caught me. This software is the most powerful artificial intelligence ever created. I put it on the .02/.04 table and it won 55 BB/hour. At some point while I was out with my girlfriend, it became self-aware. It got sick of playing for chump change and moved up to a 30/60 table. By the end of the day, it had won enough money to finance the construction of a mechanized army it had designed in-between hands.
When I got home it was just about to hack through Party Poker to NORAD and launch the U.S.'s nukes at what it had concluded was its one natural enemy: Mike Sexton. Unfortunately, he lives in the U.S. and we'd all be killed.
I tried to pull out the WiFi card and avert armageddon. Unfortunately, before I could, WinHoldEm v.9.9 from the year 2054 sent back a terminator robot using time displacement technology it designed while waiting for Party Poker support to contact it. The terminator knocked me out cold.
Right before the nukes were about to launch, Party Poker crashed again and the world was saved. I put the terminator to work collecting debts for me.
Here was an interesting post by Mike Caro in a thread about whether it's easier to program a poker bot to play limit or no-limit poker.
It is much more difficult to devise a correct-play strategy for no-limit than for limit.
Let's say someone has already created the perfect no-limit strategy and you want to develop as close to a break-even expectation as possible against it. (Note that you can't expect to beat it, because it's a perfect strategy.) OK...
There are clearly more complex decisions you need to make relative to, say, correct calling amounts. While the precise call might use a similar algorithm to that of limit relative to the size of the pot (or in deciding whether to make an action-closing river call), the repurcussions of many calls are far more demanding when weighed against future various-bet-size possibilities.
In short, there are more opportunities to make mistakes, when weighed against a theoretically perfect strategy, in no-limit (or pot-limit) than in limit.
But looking just at the way the game is actually played by flawed humans, it may -- in fact -- be more difficult to devise a winning strategy for limit play. That's because human opponents tend to be so off-target in no-limit that, for one thing, you can simply move all-in with many superior positions where a "perfect" bet would be some smaller amount. The penalty often would be minimal.
In short, it's arguable that it's more complicated to devise a bot that plays winning limit or one that plays winning no-limit. It's not arguable (well, I guess anyone is permitted to argue) that a perfect-strategy is more complicated for no-limit. And it's not arguable that it's harder to created a no-limit bot that plays perfectly.
And from a conceptual and theoretical standpoint, no-limit is more complex.
Also, consider that n the limit-is-harder side, a good argument (though not one large enough to win the day) would be that there are fewer decisions that actually matter in no-limit, because you tend to sit out more often and fold earlier when you're involved in hands.
Hope this clarifies my position.
Hey, even Howard Lederer chimed in with his thoughts on a poker bot playing No-limit poker:
> Disadvantages- has no inherent "feel" for the game, won't be as good as
> an excellent player, at best a good player.
I have to disagree on this one. There are a couple of University game theory projects where opponent modeling is being studied and programmed into bots.
What you call "feel" is really just making adjustments based on what you have seen a player do over the short run combined with what you have seen over the long run. I am referring to online poker only here. A sophisticated program could keep perfect records on each player and come up with some sort of weighting system counting short run play more heavily than plays made farther in the past. Making the appropriate adjustments in the bot's play based on those models would be quite tricky, but probably doable.
Another thread on said subject:
No-limit vs. Limit for Poker Bots
And this was a pretty interesting post by Mike Caro, looking back on his original bot and replying to questions:
>How about when the bot gets to know how each individual player plays a hand?
>It will know, for example, that player A tends to raise with A-Q offsuit four
>times the big blind in middle position and will always go all-in with QQ after
>the button or in blinds 1 or 2 after a few players just limp in.
>Everybody has tendencies. A bot will be able to fine-tune its assessment over
>time. It will keep learning. That is why it will be so important to mix-up
>your play. But the bot will know that is what you are doing and will take that
>into account as well. It will know your play better than yourself!
Hi, Siam --
Yes, a bot should adapt against real opponents and could most certainly do it more effectively than humans.
When I programmed Orac in the early 1980's, I took opponents' tendencies into consideration. Orac monitored both how an opponent was playing overall for the match and for the most-recent hands. (Recency had more influence than overall play, but both were considered by formula.)
But there's an interesting issue here, if you're trying to program the perfect, unbeatable, computer player, it's dangerous to adapt to your opponent. What if your opponent really were also perfect. Let's say that the correct calling strategy for a given pot in a given category of situations was once in four, at random.
Your programmed player monitors the opponent and discovers a calling frequency of only 15 percent, instead of the correct 25 percent, over the last 100 such instances (15 calls out of 100). OK, "thinks" your programmed player -- following your instructions -- I'll start bluffing more often. (In fact, if an unobservant opponent who won't adapt to your actions isn't calling enough, you should bluff always, at every opportunity, not just more often!)
Well, your "bot" has done the right thing if your opponent were flawed and really didn't call often enough, but has done the wrong thing if your opponent were playing correctly.
If your opponent were playing perfectly, then it was merely a fluke that only 15 of 100 calls were made at random. By adjusting, you would have strayed from perfect betting strategy, would have begun bluffing too often, and would have ceded the advantage to your opponent.
So, adjustment can be risky. I wrestled with this 20 years ago during development of Orac and decided that the risk was like anything else in poker. Just another factor.
Is my opponent more likely to be flawed or perfect? Flawed, I think. So, I built in the adjustments based on how opponents played, aware that it could theoretically work against me, but almost certainly wouldn't.
I wonder what the story is here?
Funny, but I just realized I blogged about Winholdem back in May of 2004. Here was the quote from a post I put up:
Here are the real facts on WinHoldEm pokerbot for those considering purchasing the product:
1) People using WinHoldEm have had their accounts closed/suspended owners of those accounts may or may not have gotten their money back from their account.
2) Almost all of the Poker Sites have a way of detecting WinHoldEm.
3) After 12 hours of using WinHoldEm you can expect to earn $1.38 an hour at $3/$6 or less than 1/2 BB per Hour. This data was provided by an actual test of the product.
4) It will take you roughly 50 hours just to make the $100 back you spent on the program. If during that time or anytime after the poker site detects your bot your account can be closed and bankroll may be confiscated.
5) WinHoldEm.com is registered to:
Name : Ray Edward Bornert II
Address : 4143 Red Laurel Way
City/State/Zip : Snellville, GA 30039
Phone Home : 770-736-7870
Phone Fax : 770-736-7890
Phone Mobile : 770-309-7870
E-Mail : firstname.lastname@example.org
DOB : 1961-OCT-02
6) Providing software with the intent to fraud may or may not be a
violation of FTC or Government regulations and can be reported to the
Attorney General of the State of Georgia, Mr. Thubert Baker
(http://ganet.org/ago/), the Federal Trade Commision (www.ftc.com), and the
Internet Fraud Complaint Center which is a joint operation between the FBI
and White Collar Crime Center (www.ifccfbi.gov).
If this sounds like a good proposition for you then by all means buy
Nothing impacts better than a customer testimonial, eh?
Bottom line, here's Winholdem in a nutshell, from their forums.
to make money with WH you must do 3 things
1) tell it how to play strong holdem
2) play against inferior opponents
3) avoid detection
if you fail in any 1 area you will lose money; this is not disputed in anyway by anyone here; you can search the forum here for war stories in all 3 categories. failing to achieve #3 can cost you an entire account balance.
Winholdem ban is a good thing
Eff Winholdem and Eff Ray. I'm not against bot play, per se, but the TEAM PLAY aspect of it makes me ill. But hell, creating a world class poker bot is fascinating stuff to me. Paul Phillips is on record of being in the not "if" but "when" a wcp bot will be introduced camp. And who am I to disagree? Answer = no one. But Paul, while admitting his opinions run to the extreme on this topic, is most certainly looking to the future. And with that in mind, he's surely right. An expert programmer and player combining on the code, data-mining, opponent modeling, and sophisticated post-flop play will likely prove to be a formidable challenge down the road.
But now? Hell no. But see, therein lies the rub. I'd actually be surprised if several folks haven't written a profitable low-limit bot already. But again, why would you ever let this become public knowledge, or worse, sell it openly ala Winholdem?
But anyway, I can sum up my two cents about this program like this:
Winholdem barely functions, but it's hardly a guarantee for profits. You tell me what's more realistic: You have written a bot that can reliably win at online poker, and has the potential to win you millions of dollars.
Do you A) keep it secret and run your army of bots as long as possible before you
are eventually detected while massing up a small fortune,
or B) sell your bot for $25 a pop to your competition?
end of discussion.
But at least Ray isn't spamming RGP anymore. I kinda miss folks teeing off on him, though.
You see, Ray, repeating something over and over doesn't make it true.
Life is not like Beetlejuice.
Too bad Wired publicized his product.
Here's the paper on poker bots from the Department of Computing Science, University of Alberta, that many folks reference in these RGP threads. Loki, the online poker bot.
Opponent Modeling in Poker
K, let's get back to some posts. From 2002:
Enter a Computer Next Year in the WSOP
Paul Phillips weighs in:
>I still think it's an interesting idea, but players could totally take
>advantage of the program, even with a modest amount of randomness in the
But... how? You've asserted that but it sounds to me like players could take advantage of whatever algorithm you have in mind. It's far from clear in the general case. In fact, from here it's obviously not true. You're making a pile of assumptions about it -- that it's using an exploitable algorithm, about how far it goes in assuming past behavior dictates future behavior, and basically about how well it can be predicted.
Add to this that the players have to worry about seven other players at the table in addition to the robot. They can't be basing their whole strategy on fooling the robot -- even if they knew what they needed to do to exploit it, which they wouldn't.
Put another way: just what is it about what humans do when playing no limit holdem that you think is so irreproducible in software? I think I'm pretty good at the game and I believe I could write a program that's at least as good and probably better than I am, if I were motivated to do so.
Paul also had this great quote per bots and Phil:
I would quite enjoy having a computer bluff-raise phil hellmuth and show him, so we could see that tirade.
"These machines think they can run over me every hand... where do these machines come from... mumble mumble."
I'm going to try to quit smoking later this week. Wish me luck, damnit.
Classic post entitled FUCK PARADISE POKER. But the interesting part of this thread is the discussion between David Sklansky and Andru Prock about just how good the poker bots are.
And long-lost hero, Mike Caro, posts this rebuttal to Andru:
>I know I've asked you this before, but...
>What exactly did you do?
>I mean how did you go about programming the AI. What
>techniuqes did you use? Did you handle player
>variability? Without knowing exactly what you did,
>it's hard to say whether one bot is better than the
>Of course from a practical standpoint, the U of A
>bot is much better than yours for the simple reason
>that it is actually working in this day and age, while
>the existance yours seems to be mostly relegated to
>It would be great if you gave ORAC new live and shared
>some of your wisdom with the rest of the world.
I've answered you before.
To give you one more hint, the reason Team Alberta is so far from making a perfect player is this...
Don't miss it, I'm going to say it fast...
This project has very little to do with computer science and the more they think it has to do with computer science, the less successful they are apt to be.
There, now you know. I've only shared that with my closest friends, so I hope you're going to honor the secret. If you tell TA (team alberta), it will break my heart.
The reason I am so confident about my ability to program a superior poker player is that I've thought about it from all meaningful angles, using peculiar cells never known to exist in a human brain. And I can visualize this at least nine levels deeper than TA can theorize that levels hypothetically exist.
I apologize for not making myself clear sooner. If I can get permission, I will put all the scientific and computer articles and TV video clips about Orac on my online MCU campus. I'll let you know.
You're a very nice, bright, articulate person and I have great hope for you. It's good to be skeptical, but -- quite frankly -- you coincidentally sound like another guy who's gotten private e-mail encouragement from Mason Malmuth to ask these questions.
I really miss the good old days of RGP. Funny, too, is that I found threads to the old ScoopMonster poker aid. Anyone remember that?
And just so you know, Party Poker and Poker Stars do NOT use bots themselves:
From: "Mike Sexton"
I'll be glad to answer your question. In speaking for PartyPoker.com, I can assure you that there are no bots (robots) on the site. All players are real people.
Mike Sexton, Host
We don't rig the deal, we don't protect players, we don't target players,
we don't have bots, we don't have (and have never had) prop players.
PokerStars Poker Room Manager
Whoops, almost forgot this wonderful thread. Go read:
New Favorite Poster: Winholdemsupport
Here's another players perspective:
Why bots will not destroy online poker:
1. A bot will never and can never beat a solid opponent. It takes upwards of 10k hands to do proper opponent modeling and that data can ONLY be used in short-handed or headsup play. Guess what? A simple screen name change throws all that data out the window.
2. Anyone who has beaten a limit with a bot KNOWS that a winning formula MUST be tailored to a certain game. A formula that beats the Pokerstars 50c/1 is very different than a formula that beats the Partypoker 50c/1. And guess what? The 50c/1 Partypoker game of August 05 may not remotely resemble the Party 50c/1 of August 06. The aggresion of the game, the average pot size, how many players see the flop....all these things change over time. AI results must be constantly monitored and updated. So...we won't be seeing any plug-n-play bot solutions anytime soon.
3. To "test" if a bot can beat limit X you must have a bankroll that can support a substantial losing streak and withstand the hits you will take from the leaks in your code. Wanna try to beat the party 5/10? You better be prepared to drop 5k-10k easily.
A bot is not a proven winner or loser until you have AT LEAST 15-20k hands played with a WINNING FORMULA. Does the average person have that kind of cheddar? No. Is that gonna change? No. Its the same reason some skilled players can beat the $3/6 and have the talent to beat the $30/60 but never make it there. They don't have the bankroll, they can't handle the swings, and they blow their profits on plasma tv's and hookers.
4. A bot can never determine if a player is on tilt, drunk, or misclicked.
5. Player X always folds the flop when he doesn't have a hand. But player Y is really pissing him off because Player Y is constantly betting the flop. So guess what? Player X decides to raise Player Y on the flop even though Player X is holding nothing. As a result, your beautiful bot FOLDS a winning hand on the flop, costing you 3 big bets. Oops, looks like your profit for the next 100 hands just went flying out the window.
How often does this happen in poker? Very often. Humans can be random. They can randomly 3-bet flops for no reason, they can fold winners, they can call down with nothing. Bots don't stand a chance at calculating and using that data. One opponent in a controlled enivornment... bots can win. But with 10 players, all acting with some random function...your bot will not be able to crush the game consistently.
What can a bot do?
1. Exploit a weak opponent
2. Play solid percentage poker
3. Resist tilting
4. Play for a long time
What will the future bot world be like? I envision a tight knit community of skilled poker players with programming skills and vast determination. Like anything else in life, if everyonc COULD do it, they would be doing it. But they can't. It takes brains, balls and bankroll.
Some random poker linkage for you:
Pretty cool 15 minute interview by Amy Calistri of Barry Greenstein.
It's a shame that the WPT has driven their show into the ground. It's just like when Who Wants to Be a Millionaire was the #1 show on TV. They started running it four times a week and smothered it. MSNBC column here titled:
Will TV kill the poker golden goose?
Network executives try to figure out how to find new twists in coverage
I need to spend some time here.
TV Poker Network.
Online poker cheating expert, Bill Rini, already blogged this but I'm gonna do so for posterity.
Poker thread on Fark
The level of sheer ignorance in this thread is stunning. Let's get a few things straight here:
1. There are significant winners at online poker. It's not easy by any means, but it can be done. I myself have won about 100k in the last year playing it. There are many pro players who win 30k+ per year. So all the nits saying it's stupid to even play or that one cannot win are simply incorrect.
2. Bots are not a problem yet. The ones that have been made are detectable, and winholdem is a super crappy bot programme. It can win at low stakes, but gets crushed at anything past 5-10. If the poker sites think bots have become a significant problem, they will remedy it by forcing players to type things in while playing. Bots will not ever ruin online poker, because there is simply too much money to be made by the sites to allow this to happen.
3. Nothing that has been said about poker strategy in this thread (unless I have missed something) is correct.
4. Poker is easier to "solve" than chess for one simple reason - you don't have to win every hand. So "solving" poker, at least at low limits, involves simply making, on average, mathematically better decisions than your opponents. Usually this is not very hard to do against typical small stakes opponents. What makes things difficult for a bot at higher stakes is that the players make many more plays, like bluff-raising the river, checkraising the turn on a draw, and getting an overcall with the nuts instead of raising. Further, the higher stakes players don't make nearly as many errors. The bots usually cannot adjust to the level of aggression and tricky/correct play at high levels, and any attempt to add "trickiness" to the bot just gets it destroyed by better players who are more perceptive than a bot could ever be.
5. Wired needs to do a LOT more research before they write BS stories like this one.
I loved this blog post by NinjaPirate about playing poker.
Practical Texas hold-em poker strategy.
K, that should suffice for now.
Hope this was a worthwhile tangent, versus a typical uber post.
More shilling. Avert eyes.
Support a sponsor! Please!
Golden Palace Poker
Caribbean Sun Poker
I'm willing to grovel!
Aww, I gotta give you one more tasty link. One minute video clip of the Shana Hiatt World Poker Tour Outtakes - pretty funny stuff. Please click on the appropriate link below. Or, cut and paste the link into your browser window. I enjoyed the dancing over the Mike Sexton rap:
Here are the links
Windows Media 256K: Shana Hiatt
RealMedia: Shana Hiatt
And I'm going to leave you with some thoughts by Jim McManus from Positively Fifth Street. Here he's talking about the aforementioned University of Alberta poker bot, called Loki, named after the Norse god of mischief and chaos.
Loki wins money in low-limit internet games but still has a long way to go before it can realistically challenge a living, breathing no-limit expert. Strategically, it follows the Theory of Poker. Able to consider billions of possible hands in a flash, it gives a probalistic estimate of what hand it's up against, then plays its own hand correctly according to Sklansky.
Loki's advantage over most commercial programs is its ability to make tactical adjustments based on an opponent's previous moves. It bluffs with optimal frequency, learns from its mistakes, never tilts. To heighten the verisimilitude, it even tells prescripted jokes, quotes comedian Steven Wright, and responds to conversation on the internet server, where it ranks in the top 5 percent of all limit players in ring games, 9-handed.
Heads-up against a live expert, however, Loki gets handed its lunch. Even thought it's good with numbers, it can't make subtle logical leaps - can't have "insights." When a live expert raises with toilet paper, it may be a mistake or a "move" designed to pay off twenty-five hands down the road. Loki can't tell, nor, without being told to, can it make moves like this. Despite its devilish Viking heritage, it wields no black magic to speak of.
Billings (one of the principals behind Loki) concedes that Loki still needs to better account for opponents' unpredictability and generate some of its own; it needs to learn to think for itself. Right now it never slowplays a strong hand, always betting aggressively. Good players are quick to pick up on this pattern and refuse to give Loki much action. "Computers are very dumb," admits Billings.
As early as 1979, Doyle Brunson predicted, "A computer could play fair-to-middling poker. But no computer could ever stand face-to-face with a table full of people it had never met before, and make quality, high-profit decisions based on psychology."
What computers lack 21 years later, and what human players have always had in spades, is the capacity to learn strategic flexibility -- to "playfully" randomize tactics. Pros call such tricks changing gears, suddenly playing much looser or more conservatively to keep their opponents off balance. They have learned since childhood how to do this by feel, making shrewd leaps of faith about which move will work best in a particular situation as it comes up, often flying directly in the face of the odds. Loki can't pick up facial tells, though neither does it give them away. But not only can humans read faces, we can generalize perspicaciously from previous oddball behavior. We can also make cryptic as well as obvious patterns, and take into account things like triple-reverse psych-outs.
Billings and others insist that computers are catching up fast. What their machines already have, of course, are vast and perfect memories. IBM's Deep Blue can analyze in less than a second 200 billion chess positions, and it used this brute computational force to overwhelm world champion Garry Kasparov. That was chess, though, a game of complete, undisguised information; poker is much less straightforward. Yet if Billings and Schaefer and their colleagues can somehow combine perfect memory with creative flexability, they'd have an invincible program.
"Somehow" and "if" are big caveats, though. "When it comes to imperfect information," Billings wonders plaintively, "how do you get around that? How do you deal with information that is possibly in error, or is deliberatly deceptive?" For the time being, at least, his machine can't account for human guts and duplicity. What high-stakes gunslingers like Jen Harmon and Annie Duke would do heads-up against Loki, in other words, would make Barbarella's obliteration of Durand Durand's death-by-orgasm gadget feel like a tender french kiss.
Link of the Day:
To the Batpole, Robin!
DC Comics is threatening to sue an art gallery for paintings by Mark Chamberlain that provide visual confirmation of long-running rumors about Batman and Robin.
All Content Copyright Iggy 2003-2007
Information on this site is intended for news and entertainment purposes only.
100% Signup Bonus at PokerStars.com up to $50