You know, ugly was probably the first one to point out to me this argument that computers will soon make human traders obsolete. But, lately I have been thinking that there’s no reason to be concerned.
Ever since I read The Poker Face of Wall Street, I’ve found it enlightening to look at all the various markets as gambling arenas. The book is very compelling in this regard. After all, if the stock market were really primarily for facilitating investment at fair prices, they could just collect all the buy and sell orders and process them once or twice a day at the fair market price (and that’s just one example from the book).
You know, they have a computer program that’s solved checkers (it plays perfectly). Would you bet money that you could beat that program? Neither would I. Would you buy a computer chess game if it didn’t have a crippled mode that you can actually compete with? Neither would I. You don’t hustle someone at a pool table by crushing them right away. Do you see where I’m going with this?
Any gambling establishment will tell you that they need a steady stream of amateurs coming through to keep the pros (and the house) fed. The book has a demographic breakdown of a poker house in one chapter, and it looks exactly like what I imagine the trading demographic is. A large number of people take money from their jobs and mostly lose it at the tables part time. Exponentially smaller groups play professionally at a subsistence level, and a tiny group gets wealthy. A system like that just doesn’t work without a steady stream of amateurs who think they have a chance of winning.
So, that tells me that the big players probably like the game just fine the way it is. They don’t want to collude to crush the amateurs in an obvious and spectacular fashion (and thankfully, I think there are enough big players that they would pretty much have to collude)… they would much rather bleed them dry slowly over time. And, they sure as hell don’t want to speed up/gap everything until it resembles the fabled “efficient” market, because no one can beat that game.
To me, this means that the trading game will always be easy enough for most amateurs to win sometimes. Specifically, they’ll be able to win often enough to keep them hooked. And that means I will always be able to make consistent money by making sure I’m better at it than most amateurs. That doesn’t mean that the game won’t change dramatically going forward. That also doesn’t mean that most amateurs won’t be trading with super-fast, artificially intelligent computer agents one day… but I have a feeling I’ll still be one step ahead of them, when it happens. :-)
November 2nd, 2007 at 10:26 pm
I think it will just become more difficult for the pros as the amateurs get better and better with technology. And eventually they will all have to get AI trading machines and human traders won’t have a chance. I think it’s already beginning to happen.
I’m trying to stay one step ahead of them, too - by building an ATS.
I think as programmers we have a pretty good chance.
November 2nd, 2007 at 10:27 pm
Computers taking over trading…. never going to happen. Some big hedge funds lost money last quarter on their computer trading system. Computer trading can be programmed to take trades and compute statistical probability outcomes but there is still one thing they can never have.
Emotions.
While a lot of the time trading just boils down to supply and demand but when you add emotions such as fear that can result from news or in general, these attributes can create panic and change the market into acting rational to the irrational. When you throw in the market manipulation aimed at the individual retail trader and the intelligent trader who takes advantage of trading programs, the computer trader doesn’t stand a chance.
November 2nd, 2007 at 11:03 pm
Ugly, maybe so, that’s an interesting idea that the playing field will get more level instead of less level. Yeah, I think our chances are excellent, as well.
Really, KC? I feel computers will continue to increase their role in the trading world. But, what I’ve realized is that they won’t make the markets untradeable. That’s all I care about. From the gambling economy perspective, if the common folk were excluded, the markets wouldn’t be able to serve their function anymore… and no doubt new trading instruments would be introduced to fill the void.
November 3rd, 2007 at 12:37 am
I can see how computers are essential in arbs and the pairs traders are probably in trouble (essentially statistical arbitrage) but apart from them, I don’t understand what it is that you think computers could do anyway. Markets aren’t like checkers, it isn’t a game with right and wrong answers, there’s not a “right” price for a stock. If computers buy breakouts then we frontrun or join them; if they fade breakouts, we fade along side. If they trade randomly, they go broke like anyone else. We small traders just join moves and computers or no, there will be predictable moves. As long as there’s greed and uncertainty about the future, there will be human tradable moves.
So answer me this: what would be different about the markets if the big players decided to crush the amateurs, as you say? How would this differ in any way from the markets we have today?
November 3rd, 2007 at 4:41 am
Basically it comes down to this: computers are getting smarter and smarter. Not just in a brute-force way - they are getting more intuitive and creative. They are really becoming smarter, with pattern-recognition capabilities that are beginning to rival those of humans. Eventually they will be the best traders.
Trading is a lot like poker, and look at the progress of poker-playing AI. I think it is important for traders to watch this progress in poker. When computers dominate poker, as they now do chess, it will be more clear as to how they will also dominate the market.
I don’t think even the best unassisted human brain will be enough to compete with this AI in 10 years. I think it will be like playing Deep Blue. Over time it will just be better and take your money. Why is it so hard for traders to make money in the markets today? Because it is competitive, we all can’t win, and there are really smart, good traders to compete with. Smart AI is currently and will continue to be a growing part of this competition.
November 3rd, 2007 at 4:46 am
[...] Richard says he is not worried about computers that can trade. Others have told me they aren’t worried either - they believe there will always be ways for a human to make money in the markets. They could be right, but I don’t think so. I think it is only a matter of time before even the best human traders (without the help of AI) cannot make consistent profits in the market. Here is a comment I just added to his post: Basically it comes down to this: computers are getting smarter and smarter. Not just in a brute-force way - they are getting more intuitive and creative. They are really becoming smarter, with pattern-recognition capabilities that are beginning to rival those of humans. Eventually they will be the best traders. Trading is a lot like poker, and look at the progress of poker-playing AI. I think it is important for traders to watch this progress in poker. When computers dominate poker, as they now do chess, it will be more clear as to how they will also dominate the market. I don’t think even the best unassisted human brain will be enough to compete with this AI in 10 years. I think it will be like playing Deep Blue. Over time it will just be better and take your money. Why is it so hard for traders to make money in the markets today? Because it is competitive, we all can’t win, and there are really smart, good traders to compete with. Smart AI is currently and will continue to be a growing part of this competition. [...]
November 3rd, 2007 at 9:07 am
Tyro, think about it. They are faster, with quicker access to information, and deeper pockets… if they made it their goal to freeze out the small players, they could do it without much trouble I think, even today. Look at different ways players can collude at poker, for examples. A common example is where they raise and re-raise until the mark is playing at higher stakes than they are comfortable with… on a chart I think this would usually look like a broadening wedge, which you may have noticed by now is a royal pain in the ass to try to trade through.
Contrary to what your comment said, I think an easy method would actually be to inject more random noise into the markets. As long as the amplitude of the noise is smaller than their threshold for pain but larger than yours, you can’t win. Yes, they would take losses in this process, but since you will be forced to walk away and they won’t, they will make it all back (minus their practically $0 transaction fees). The money stays “in the family” so to speak. There are people that say the markets are already like this… so close to random on small timeframes that daytraders are just fooling themselves. I heard someone making this claim about FX trading just yesterday.
Luckily, the line of thinking behind my post says that they would never intentionally do this, because they want people to keep bringing their money to the markets. They want to promote it as a safe way to build your wealth, etc. etc.
Ugly’s right that AI traders could become good enough that no human can beat them in the long term (like a “perfect” poker player). But, at that point, wouldn’t there suddenly be a glut of these ubertrading computers in the market? And if that happened, I’m not sure the markets would stay “easy” enough to keep people coming in. Changes would have to be made… The “Small Investor Protection Act” of 2012 will be passed, or something.
November 3rd, 2007 at 10:56 am
Tyro, I enjoyed you post.
Ugly brought up poker. Most traders do also like and play poker. Trading and playing poker is both methodical and deals with high probabilities. If you follow professional poker, then you know who Phil Laak is, AKA “The Unabomber”. On NBC’s Poker After Dark, Phil talked about how he was paired up with a computer to see how it would fair against a pro. Laak, knowing that the computers play would play high percentage odds used that information to his advantage and altered his actions to take advantage. I believe the same thing can happen in trading. It’s no secret that as soon as higher highs start to be made, buying programs are triggered. Smart hedge funds simple cause the market they are in to consolidate by buying at their lower price and shorting at higher prices, knocking price back down to where they can continue to further accumulate, Then begin to cover their shorts while the market still consolidates. Once their shorts are nearly covered, they finish covering and become the catalyst of the box pattern break, sending the market high and “using” those buying systems to take the markets up for them and buy from them.
Computers getting smarter and smarter are in result of human program traders, yet human traders will always be able to out manipulate a computer.
This little computer versus human discussion reminds me of two movies, the Terminator with Arnold and War Games with Matthew Broderick.
November 3rd, 2007 at 12:50 pm
Ugly & Richard,
I respect you guys and I know you’re both tech savvy, but I’m not sure you’ve worked this through, or at least I’m not able to follow your thought process. In chess and especially checkers the question is a computational one. There are fixed rules and with speed and storage space, you can walk down each tree through brute force. When Schaefer talks about solving checkers, that’s exactly what he’s done - calculated every decision path. Chess is more complex and they use heuristics to prone different trees, but it’s easy to see how better computers and better heuristics will just exaggerate this.
What you haven’t done is to demonstrate that the stock market shares any of these characteristics. What exactly is the problem which needs solving? Richard, you say computers have deeper pockets and bigger info and so could freeze out humans - how exactly could they do this, and what does info and buying power have to do with anything? Ugly, you should know that the only edges a computer can exploit better than a human are those where execution speed or intensive calculations are paramount, and that is only in arbitrage. So yes, pairs traders and spread traders will be hurt but who else? Neither of you have answered that.
You’re acting like a fully computerized trading environment would have zero volatility because the computers would magically “know” what the price “should” be and even more magically, all agree. As if that wasn’t ludicrous enough, you act as if the computers buying and selling aren’t acting on behalf of humans who are trying to move money around for their own reasons. That’s why I asked you to be specific in saying exactly how they could lock out humans. I hoped that by trying to come up with details, you would see for yourself that computers aren’t an enemy, but actually another edge that can be exploited.
If you’re just going to wave your hands and argue, in essence, that computers are just better and therefore humans will just lose, how can we discuss anything?
KC brought up a good point, that even poker bots who can bluff well will, at their best, act like a human! This isn’t computers replacing humans, it’s computers joining. Now make the poker game with not just 10 people but 10,000 without any hand history and no hole cards. Imagine two trading bots head to head: if they are going opposite directions, you have a narrow consolidation and when one breaks you have a breakout and NR7/dummy/breakout traders clean up; if they go in the same direction, then human trend traders are printing money. Computers just don’t change the field that much, except by providing better liquidity and reducing slippage.
November 3rd, 2007 at 1:36 pm
Well, since we are talking about the future it has to be speculative. If I knew exactly how to build a trading-bot that could decimate all competition, I’d be building it rather than writing in my blog. The post wasn’t even really about that… there’s plenty about institutional trading that I know nothing about. I was only thinking that, from the point of view of a trading market as a gambling arena, computerized uber-traders would not freeze out the amateur group, because they are a necessary part of the ecosystem.
In my last reply I tried to give two examples about how deep pockets could collude to freeze out small traders, if they made it their goal to do so. Maybe you didn’t like my examples… it’s not really my area of expertise. In my original post I specifically said that no one would voluntarily work toward the zero-volatility situation, because no one can beat an efficient market. So, I’m not sure who you are arguing with, but I can’t see how it can be me. I mean, your final sentence says that computers won’t change the field that much, and my whole post is about how there will always be room for the amateur trader…. you see?? I am AGREEING with you.
But, I do think that computers will be excellent traders, and human traders will need more and more computing power to stay in the game. Ok, let’s say computers merely join the game by doing what humans do. Maybe this is too hand-wavey for you, but… If I ask myself, “would I be a better trader if I could consider more options faster, in more simultaneous markets, with more capital at my disposal?” the answer would be an unqualified yes. How could it not be yes?
People said a computer could do arithmetic but never calculus. oops! People said it could compete at checkers but never chess. oops! Now people are saying chess and poker have limited number of moves, but markets are waaay more complicated. You know what word comes next? oops!
November 3rd, 2007 at 1:43 pm
Tyro: Yes, chess and checkers aren’t as much like the stock market as poker - because as you say, they are more about computation and brute force. I’m not trying to say the stock market is like this completely. I was talking about AI and automated trading programs that learn. You say “Ugly, you should know that the only edges a computer can exploit better than a human are those where execution speed or intensive calculations are paramount, and that is only in arbitrage.” That’s not true. A computer has an edge that it never tires and it does not have emotions to get in the way. You yourself post about how difficult trading is emotionally and how hard that makes it. Imagine the edge a computer has with this. But there are many more advantages a computer has. It’s speed and intense computation can also be applied to learning.
I simply see the exponential growth of information technology and AI compared to human intelligence and I see a point in the not-so-distant future when AI moves past us. At that point, yes, there still may be ways to exploit the market and make money - but I think the computers will find them first, more often, and with a consistency that we won’t be able to achieve.
I can’t be specific in saying exactly how they could lock out humans. I don’t know the answer to that.
But, think about this: if everyone knows that a NR7/dummy/breakout system works and everyone tries it - it no longer works. Right? We can’t all make money all the time. So I think computers will be able to take advantage of these systems and adapt to the changes in the markets much better than us.
November 3rd, 2007 at 1:52 pm
[...] nice discussion about AI and Trading has developed from Richard’s post and my reply. I agree 100% with this comment Richard just made in reply to Tyro: People said a [...]
November 3rd, 2007 at 2:27 pm
Ugly, yes, computers are able to replicate many behaviors of humans such as buying NR7 breakouts, but what exactly can they do better? Neither speed nor computational power are significant, and on these small timescales liquidity is an issue so buying power certainly isn’t an advantage (or the computer will become the market, creating opportunities). The bot becomes another player, neither better nor worse than any other and we all adjust just as we always adjust to the faceless traders every other day. Bots will not qualitatively change anything in this scenario.
Richard - many (most) traders earn their living by joining along institutional buyers, so I can’t imagine how they could become our enemy, even if they wanted to. How could a billion dollar fund act without creating opportunities for small traders? (They could leave the public exchanges to NYFIX Millennium or other dark books, but if their price is different, then arbs will make Millennium the same as any other ECN.)
Anyway, I think I’ve made my case and without details on the speculation, we’re not going to make any headway.
November 3rd, 2007 at 3:58 pm
So, lemme get this straight… at the point of breakout, speed is not an issue, but liquidity is? Dude, put those two ideas together. The early bird is getting that worm, end of story. You can go on thinking “neither better nor worse” all you like, but unless you keep up with your competition, at some point you may realize you have brought a knife to a gunfight, so to speak.
Are you seriously saying your list of positive trader qualities would not include speed and pattern matching ability? Emotional control? Are you seriously saying that edges don’t disappear? You seem so sure of yourself, but last I checked you aren’t actually making any money. Consider the possibility that you don’t have all of this completely figured out. And, if you do, then why not take your own advice and make a living joining along institutional buyers, since it’s unimaginable that they could act without creating opportunities for you.
November 3rd, 2007 at 4:12 pm
i’m not a programmer, but i know one thing….as long as computers are programmed by humans, there will be different ideas on the correct path…as proof i offer up tyro, ugly, and richard
as long as prices move, humans will have an opportunity to make and lose money on their own
btw, news events trump technicals and any trading bot…as postitions are taken on and off as people (and i guess trading bots) process the new information, there will be opportunity
November 3rd, 2007 at 4:21 pm
I miss ‘R’ … does anyone want to tell me ‘R’ is pointless, and possibly then set their penis on fire?
November 3rd, 2007 at 4:24 pm
i have found that most proponets of “R” are the same people that will tell u size doesn’t matter…if that makes u feel good about yourself, continue to believe the lie (about “R” that is)
November 3rd, 2007 at 9:57 pm
Richard - Breakouts aren’t instantaneous and we already have stop and market orders so no, I don’t see how more speed will chalge things. Help me out and explain it to me. And yes, there is a limited number of shares that you can take without becoming the market. If you think that this means that computers with huge buying power have an advantage, presumably imagining that they’ll grab all the shares, you’re wrong. They’ll become the market and create their own tradable pattern of frontrunning them and selling into their move. If they wait for others and limit their size, they’re no different than trading against other traders (many of whom have far more leverage than they use). Huge buying power here is no different than someone placing a stop buy order for 50k shares which may take shares from some, but will probably screw the big trader when they can’t unload. Where’s the advantage which is meant to squeeze out the little guys?
Yes, some edges and patterns come and go, some will remain. Large computer systems may change some things but where one pattern stops, another must open. It’s the nature of the market as traders of different size, different interpretations, and different scales all interact. No different than today.
Sure, emotional control and pattern matching are elements of trading, of course. So? Put that together with something and tell me how you imagine that helps your case, ’cause I don’t know where you’re going.
As for your attacks on me and my trading, if that makes you feel better, go for it. I’m an easy target. But attacking me doesn’t change what I say, nor does it change the fact that your argument rests entirely on hand waving and innuendo. It’s yet another logical fallacy called an ad hominem and is generally the last resort of someone without evidence or argument.
November 3rd, 2007 at 11:27 pm
You are right, Tyro. Thanks for setting me straight.
November 4th, 2007 at 1:30 pm
[...] know, when I wrote Friday’s article, I wasn’t planning at all to get into a discussion about artificial intelligence in trading. [...]
November 4th, 2007 at 3:33 pm
Remember that Chess and Checkers have a finite number of moves. It could be in the millions. But take; stocks, options, forex, commodities & futures and then try creating a finite combination. Good luck.
So in some senses there will always be opportunities. And if there aren’t any, someone will create a new marketplace with new rules.