Pokerbabe/market wizards

[Warren Whitmore]

I took your advice on reading market wizards & the new market wizards. Very interesting material and I believe useful both for trading and poker. I found it particularly interesting that some traders had run up several thousand dollars to several hundred million. Most of those who did were pure technical traders or pure fundamentalists. It reminded me alot of poker where some (Mike Caro/Al schoomaker) have stated that they believe computer programs will be able to eventually beat full games, Bobby Baldwin who felt that the best way to go about poker was to be tutored by the best before playing in thier first game, and, others who believe that intelligence mixed with experiance cant be beat. My question: Do you believe a computer will eventually hold its own or beat a line up of the 8 greatest poker players? How about a computer vs the 8 greatest traders on wall street?
Thanks in advance.

[mdlm]

You may also be interested in knowing that the majority of the traders in Market Wizards blew themselves up after the book was published. Market Wizards are drawn from a population of tens of millions of traders. Some of them are bound to get lucky and turn pennies into millions.

You ask: "Do you believe a computer will eventually hold its own or beat a line up of the 8 greatest poker players?"

Of course. Within 100 years there will be nothing that humans can do that computers can't do better, except looking better than the Babe while they're beating you.

You ask: "How about a computer vs the 8 greatest traders on wall street?"

That has already happened.

[PokerBabe(aka)]

Hi Warren- glad you enjoyed the books and learned from them. I think that good discipline, knowledge and solid decision making abilities are required in both poker and trading. A computer can make "correct" decisions about poker and trading based on "technical" data, but I doubt if a machine can "read" players or get a "feel" for trading. /forums/images/icons/ooo.gif For example, Stu Ungar was known for his excellent "reads". This skill is crucial at the highest poker limits. Can a computer duplicate Stuey's decision making process in this area? I really don't know. Perhaps the computer will "modify" it's play based on patterns it picks up from opponents, rather than simply playing "by the book". Perhaps the computer can pick off a bluff after observing certain patterns of play. I would like to get a computer expert's opinion of whether a machine can be programmed to "read" (and beat) the best poker players. Al knows of a lady who is working on this right now, and he might provide more explanation. I'm sure that Mason has an opinion on this as well.


As for the market, there are already numerous programs and technical systems for trading. These simply run off what is happening (and HAS happened) in the market but are notoriously poor at predicting the future! /forums/images/icons/crazy.gif I had this discussion just the other day with Barry T. about how technical analysis is ex post facto. As an example, consider the current market. Every prior bear market was different than the one we are in now. They started for different reasons, lasted for different amounts of time and were different in severity. Sure, we can look back and say ....YUP...there is the bottom. We can look back at the July 23 lows of this year and say....ok...now we are "retesting" them. This "should" be bullish from a "technical perspective", etc. BUT those support levels mean nothing when a very HUMAN event occurs and emotion runs the market. You can throw the computer right out the window when you get a currency collapse, an unexpected interest rate hike, a presidential assasination, a terrorist attack, etc. If you ever watch CNBC, you are familiar with a guy named Art Cashin. If you listen to Art he will often tell you about how the floor "feels". He will mention how he "thinks" the market will act as the day progresses. He will talk about the "trader activity" (how fast they are walking around, for example). He mentions "taking the pulse" of the market, etc. After many years on the floor, he has a "sense" of what is going on. Can a computer duplicate Art's "touchy-feeley" knowledge of the market? I really don't think so. /forums/images/icons/tongue.gif . A friend of mine who worked on the CBOE floor once told me that when someone is vomiting in the restroom, go in the direction opposite the market! Can a computer do THAT type of analysis? Hmmmm.....

If I was pressed to give a simple yes or no answer, I would say that a computer may someday beat any one great individual poker player heads up, but not 8 of them at a time. Similarly, I think a computer can trade via various technical systems, but could never ever duplicate Art Cashin's "touchy-feely" analysis of the stock market action or my pal's "vomiting index"!. /forums/images/icons/tongue.gif LGPG, Babe

[Warren Whitmore]

Much appreciated, thank you

[eMarkM]

Excellent response, Babe. I'm a computer programmer, though no expert in artificial intelligence. I'm also like to play the markets, again not at any expert level. Let me just add that something about markets called "everchanging cycles" and how that might apply to a sophisticated poker bot.

Yes, computer programs have been in use in trading for many years. Some of them have been remarkably successful. Long Term Capital Management used computer models to great effect in the late 90s, basically finding arbitrage opportunities. They had a remarkably consistent record in their early years and the models held up fantastically. Well, until they blew up that is. The problem with computer model--or any money marking market method really--is that once it's discovered that your "system" makes money, others also discover it, copy it and at some point dilute the results so much that the method no longer works. The "cycle" changes and you have to rediscover another method. I know of no computer model that takes into account ever-changing cycles. A human must intervene to re-program or come up with new method.

The same may be applied to a "Deep Blue" poker super-bot. You could program it play perfect cards, but to mix up play with game theory so it's not predictable. But to be successful I would think it would HAVE to adjust its style to the way the humans play. Just like any great player, knowing your opponents is the biggest advantage.

But I would think the great poker players would be able to use the concept of ever-changing cycles to confound the program. The program would keep a "book" on a player, but a good player can change his "book", e.g., loosen up if the bot is taking advantage of his tight play. In the end I think the humans would be able to outplay the most sophisticated bot.

The reason a chess program can beat a human is because chess is pure number crunching. Poker is a lot of number crunching, but it's so much more. I'm not sure how much psychology you can program into a bot and I think most great players could adjust their game to it. At least with today's technology.

[eMarkM]

Here's a link to an article about a bot that can beat 90% of the players it faces. Interesting article about the differences between a poker bot and a chess program. One has complete information, the other does not.

Poki (http://www.expressnews.ualberta.ca/expressnews/articles/news.cfm?p_ID=2765)

[Al Schoonmaker]

I don't see the link. I clicked on "poki," and nothing happened. I'd like to know how to get to that article or any others on this general subject.

A few months ago I argued in Poker Digest that computers would soon beat champions at limit poker. Most of my friends disagree, but chess players said the same about computers and chess. Poker players insist that poker is somehow different and more complicated than chess or other activities at which computers have demonstrated their superiority, but they are kidding themselves. Poker is NOT that different.

The incomplete information issue is nonsense. Computers destroy people at Rock, Scissor, Paper, which is also a game of incomplete information. In fact, the only thing that counts is reading the other player. Computers can read you MUCH better than you can read them.

I'd appreciate links to any articles on this subject.

Thanks,

Al

Poki (software developed by Darse Billings and friends at U. Alberta) may be a good start, but unless it's advanced a lot in the last year or so, it really plays very weakly. They say it has a winning record, but I must assume that's in play money games against humans who either play terribly, or who aren't trying at all. If you go on their site and play against it a little, you'll see that it's pretty weak.

That said, I do believe computers will one day be able to play very well, if not quite as well as top experts. That day could be here now, I think, but for the fact that there hasn't so far been enough interest to fund large scale projects to create such software.

The link worked fine for me. Odd.

Being a chess master (albeit a weak one) and having followed with great interest the evolution of chess playing computers this is a fascinating subject for me. I think this is a question that may never find an answer. I believe that computers will be eventually able to play top level poker but who's to say if they play a little better or a little worse than top humans? One tournament (or even 10) is not enough to prove anything.

I disagree with Al on one thing though - poker IS way different from chess. eMarkM makes an interesting point with respect to ever changing cycles and this is a good example of just how different poker and chess are. In chess the best move in a given position will still be the best move 100 years later. In poker the best "move" is in most cases open to debate and opinions may change from hand to hand depending on the situation at the table.

If computers are proven to be better than humans I believe it's more likely to happen at limit poker which is more calculating and less feel.

Al can you provide any links regarding computers playing Rock, Scissors, Paper. I would be very interested.

[eMarkM]

Link is ok for me, but here it is.

http://www.expressnews.ualberta.ca/expressnews/articles/news.cfm?p_ID=2765

This includes other links at the bottom to their web site and even has some Java source code for us programming weenies.