c3po

Artificial Intelligence in Stock Trading: Future Trends

Opportunities in AI-powered Trading for Start-Ups
1. Start Your Own Private AI-Powered Hedge Fund
2. Build Algorithms and Sell Them to Other Hedge Funds
3. Sell AI Algorithms Designed for Basic Investing Practices
Eugene Fama’s Theory
The History of AI-powered Hedge Funds
The Medallion Fund
Bridgewater
Sentient
AI Stock Trading vs. Human Stock Trading
Your Algorithm Has The Advantage of No Emotion
Your App Needs a Fast Kill Switch
The Dangers of the Automated Trading Ecosystem
Should You Market Your Algorithms Towards Large Businesses or Average Investors?
Individuals Can Almost Never Beat the Market – But They Have Other Needs
The System Is Evolving. Stay Alert!

Opportunities in AI-powered Trading for Start-Ups

I see three major opportunities for start-ups in the AI-powered stock trading space for the next few years. These are essentially three different paths starting from the same fork in the road: You’re building an algorithm to trade with. Who is it for, and how will you monetize it? Here are the three paths:

1. Start Your Own Private AI-Powered Hedge Fund

If you can build the best algorithm for stock trading, it makes sense to take the gains for yourself. That’s why starting a private hedge fund with a proprietary algorithm is a huge idea.

This isn’t for the faint of heart, even by entrepreneurial standards. You’ll need to gain access to hundreds of millions of dollars to have the kind of leverage that it takes to earn immense stock gains. You’ll also need to outperform 99% of the other funds on the market, some of whom are already working with the leading developers in this field.

Remember all that info about the Medallion fund, who have earned 30+% for 28 years? That is the competition, and they’ve got around 100 employees working on their algorithm in all its various forms. You’ll need Elon Musk levels of fortitude and endurance to make a dent with your own fund – but you’ll reap Musk-level rewards if you do. Good luck.

2. Build Algorithms and Sell Them to Other Hedge Funds

This is the much more reasonable version of the above idea. You don’t need to get a fund started and work with hundreds of millions of dollars of other peoples’ money. Instead, you simply target the people who run hedge funds and convince them that your algorithm can improve their returns.

blog-banner-download-project-specs-2

AI-powered stock trading is an increasingly hot topic. The New York Times, Wall Street Journal, wired.com, Forbes, and many other major publications have written about it in the last few years. People are interested, they’re hungry to get in on the action – but access is limited.

As banks continue to update themselves and enter the modern era, there is a big need for profitable and reliable (i.e. disaster-proof) algorithms that can earn better returns than humans can do alone. If you can build it and get the right information to the right people, you can sell your services to the banks.

3. Sell AI Algorithms Designed for Basic Investing Practices

There are only a few companies right now offering automated investing for the average person. If somebody’s got tens of thousands of dollars to invest, they don’t need advanced hedge fund technology – they just need some software that can allocate their funds into a simple and smart investment portfolio.

You could build a brand and sell your services directly to consumers, ala Wealthfront or Betterment. Alternatively, and I think this is the real winning idea, you could sell these services to banks who then offer them to their customers. Companies like Schwab Bank and Ally are prime examples of businesses that could get a lot of benefit out of consumer-oriented AI trading algorithms. It isn’t as exciting as high-stakes hedge fund trading, but it may be enough to earn you a serious payday.

walter-white-money

Eugene Fama’s Theory

When Eugene Fama won the Nobel Prize for economics in 2013, the investing world was in an uproar. How could the nobel committee legitimize his heretical views?

Eugene’s efficient-market hypothesis cast doubt on the entire idea of actively managed funds. His claim was simple: The market is so efficient that no single entity can realize significant long-term gains by trading stocks. He argued that the market is the sum of all available financial information. He saw stock prices as the financial equivalent of wikipedia pages, drawn from thousands of different sources of information and updating in real time.

The implication of this information was clear: Actively managed funds are a waste of time. Yet, a few powerful and secretive hedge funds begged to differ. They had already harnessed the power of data scientists to build automated trading systems, generating double digit returns for nearly 30 years.

The question is: Who will build the next generation of AI-powered trading algorithms? There is a lot of room for new start-ups in this field.

If you are an entrepreneur or developer that works at the intersection of AI and stock trading, the rest of this article is for you. You need to understand how large of an impact artificial intelligence is having on the world of finance. I’m going to show you how AI is a powerful new technology that you can sell to big banks and to average investors alike.

Whether you’re a developer, a data scientist, or an investor, this post will give you context and show you the future trends of AI stock trading. Let’s get started.

man-at-financial-trading-computers

The History of AI-powered Hedge Funds

Hedge funds powered by computer models are more prevalent than you might think. Cade Metz of Wired points out that “some 1,360 hedge funds make a majority of their trades with help from computer models—roughly 9 percent of all funds—and they manage about $197 billion in total.” This has led to massive profits as well as some total disasters – like Knight Capital’s loss of $400 million due to a computer glitch, which I’ll talk about later.

The only way to properly explain the past 30 years of computer-powered, and eventually AI-powered trading, is to look at a few key hedge funds and financial executives who have dominated this area of investing. The most famous and reclusive of these by far is the Medallion fund.

The Medallion Fund

Jim Simon founded Renaissance Technologies in 1982 after a long and successful career as a cold war code-breaker. While working to decipher encoded enemy messages in the war, he developed a keen sense for finding “signals” of valuable information amongst the “noise” of random gibberish.

This is not as easy as it sounds. Nate Silver, one of our generation’s most prominent statisticians and founder of FiveThirtyEight.com, explains: “I think the challenge people don’t realize is that when you have more and more data, in some ways that makes it harder.” In the world of encryption, this makes sense: Your enemy is trying to fool you by giving you so much information that you don’t even know where to start.

In the world of finance, everybody wants to earn the profits. Sadly, it’s a zero-sum game. For every extra percent of returns that one investor earns, another investor must lose a percent. Jim Simon’s cryptography experience helped turn him into the data-parsing equivalent of a Mike Tyson or Muhammed Ali. When pitted against other investors in the world of finance, he was in a higher weight class of knowledge and would readily dispatch them.

The medallion fund, which opened in 1988, has not been open to outside investors since 2005. It’s been referred to as the “Manhattan Project of Finance” by Andrew Lo, the director of MIT’s Laboratory For Financial Engineering. With $55 billion of profits so far, it has earned an average return of more than 30% per year for 28 years. How’s that for an efficient market?

Joel Weber of Bloomberg Markets suggests that the medallion fund’s success is in part because “these people aren’t wall street people – they’re data scientists.” Rumor has it that a wall street resumé is certain to get you rejected from the firm. Instead, they employ the world’s best scientists. It’s not hard to attract the top talent when you consider that this fund is exclusively available to employees of the company.

Little is known about the actual computer programs behind the success, due to the intense secrecy of the group. However, it is known that they have pioneered the use of AI to capture immense value in a market where quantitive edges are getting smaller each year.

The question moving forward is simple: As more funds dig into AI-powered technology, will Renaissance Technology be able to maintain its edge? If so, they will have to maintain secrecy and stay one step ahead of the technological curve.

Bridgewater

Ray Dalio founded Bridgewater Associates in 1975 right in the middle of Manhattan. They have since relocated to southern Connecticut, just a stone’s throw from the old headquarters. Ray has been a hugely influential voice in the world of finance for many decades now, with a net worth of more than $16 billion and having been featured as one of Time Magazine’s 100 most influential people in the world.

Ray is a true pioneer is the field of artificial intelligence as it relates to finance and stock trading. He’s a great guy to watch when looking for future trends because he’s only 67 years old. In the world of mega-finance-billionaires, that’s pretty young, and it means he’s able to take a longer term view than others. In contrast, Jim Simon from Renaissance Technologies is 78 and mostly retired.

While Bridgewater does seem to be dabbling in AI-powered trading (public information is scarce), they’ve been much more vocal about their goal of building AI-powered business management.

You heard me right. Ray Dalio wants to let artificial intelligence run his business. If the singularity is imminent, this might be where it starts. The Wall Street Journal gives some context for what this looks like in practice: “The role of many remaining humans at the firm wouldn’t be to make individual choices but to design the criteria by which the system makes decisions, intervening when something isn’t working,”

This puts humans at the micro and macro levels of management – leaving robots to handle everything in-between. When a big decision needs to be made, all of the opposing viewpoints and arguments will be fed to the machine and a decision will be rendered by the AI. The humans will observe this process and tinker with the inner workings of the AI to improve the results.

The Guardian points out that this is no small idea: “These tools are early applications of PriOS, the over-arching management software that Dalio wants to make three-quarters of all management decisions within five years.” (https://www.theguardian.com/technology/2016/dec/22/bridgewater-associates-ai-artificial-intelligence-management) This has shades of Pareto’s 80/20 rule, where robots can take over all of the most annoying and mundane decisions so humans can handle the truly significant ones.

The concept of stock trading via automated management seems more likely to have long-lasting impacts than anything else in this article. As I’ll discuss later, it is highly unlikely that AI-powered trades can ever reach mass appeal because if everybody has it, nobody has an edge. When it comes to management, however, everybody can share the same AI tools and human capabilities will still determine who ultimately wins the zero-sum game of investing.

Sentient

Sentient is one of the newest businesses to enter the AI-powered stock trading space. As such, they are a great model to look up to if you are trying to build a business in this space.

Its founder, Babak Hodiat, has serious credentials thanks to his time developing Siri for Apple. After leaving Apple, he “spent nearly a decade—largely in secret—training an AI system that can scour billions of pieces of data, spot trends, adapt as it learns and make money trading stocks.”

Babak offers some useful clues on the inner workings of these AI systems. He describes the methodology as a simulation of genetic evolution. The computers test millions of different trading strategies against historical data, rapidly iterating by discarding the worst ones and improving upon the most effective.

Over the course of trillions (!!!) of simulations, this has lead to a powerful AI machine that can learn from its own mistakes and optimize trading strategies for an ever-evolving market. It sounds great in theory, although there are no public disclosures yet about the actual returns generated by their model. Most of the investments so far have come from a single VC fund with deep pockets.

If Sentient proves to be as profitable as Jim Simon or Ray Dalio’s operations, it will show that the field of AI stock trading still holds value. Should they fail, it may suggest that the ability to gain an edge with AI is diminishing rapidly.

blog-banner-download-project-specs-2

quants-the-alchemists-of-wall-street

AI Stock Trading vs. Human Stock Trading

Another way to identify future trends is to compare the performance of artificial intelligence algorithms as opposed to human traders. After all, these ideas do not exist in a vacuum. If you want to build a profitable AI algorithm for stock trading, you’ll need to outperform the best humans in the market. That’s the only way to sell your services to the biggest players.

If computers can consistently outperform humans in stock trading, they are here to stay. With that said, AI brings different strengths and weaknesses to the table. The next few sections point out some key differences between AI and human trading.

Your Algorithm Has The Advantage of No Emotion

When testing your stock trading algorithms, don’t look towards times of market stability. Test your apps against the most volatile market periods in history and see if the unemotional AI can outperform the humans.

Lack of emotion is an obvious advantage for the AI-powered traders. Humans are notoriously bad at stock trading for the simple reason that their emotions get the better of them. Stocks plummet and they sell out of panic. Or, just as bad, a stock reaches new highs and prompts greedy investors to buy-in at the top.

Dr. Christopher Krauss, chair for Statistics and Econometrics at the School of Business and Economics at Germany’s Friedrich-Alexander-Universität Erlangen-Nürnberg, points out that “Our quantitative algorithms turned out to be particularly effective at such times of high volatility, when emotions dominate the markets.”. Computers are ruthless and they don’t get sad when the markets move in a direction they didn’t predict. Instead, they learn from experience and use that knowledge to better gauge future market shifts.

Your App Needs a Fast Kill Switch

Whenever you build any kind of automated trading software, you must include an instantaneous kill-switch. Clients need to be able to stop your program upon the first sign of any trouble.

This isn’t the flashiest feature, but it is one of the most important. Just ask the team at Knight Capital, whose entire business was ruined in less than an hour due to a coding error in July 2012.

The New York Times explains that “as a torrent of faulty trades spewed Wednesday morning from a Knight Capital Group trading program, no one at the firm managed to stop it for more than a half-hour.” The glitch was identified within a few minutes, but the firm had failed to build a kill switch into their trading software.

What ensued was a 40+ minute scramble as the entire company pored through the innards of their machine, trying to identify and disable the bogus code. By the time the dust cleared, they had lost more than $400 million dollars and were near bankruptcy.

Humans can make mistakes, but not like that. This kind of crisis is only possible when a machine lurches forward, out of control, spewing money with no regard for what it is doing. Build a fast kill switch to protect against the worse-case scenario of an algorithm malfunction.

The Dangers of the Automated Trading Ecosystem

There are other safeguards to consider. When your stock trading algorithm interacts with other algorithms, what will happen? Will it know how to defend itself from irrational or deceitful algorithms elsewhere in the market?

The “flash crash” of 2010 is an example of unexpected emergent properties from AI-powered trading creating havoc in the markets. A single unusual trade from a poorly calibrated machine set off a chain reaction of bizarre computerized trades, sending the market into a tailspin for no real reason.

As the Wall Street Journal puts it, the federal regulators ““pinpointed one trade by a mutual fund company as a key contributing factor to the market’s plunge.” When humans are in control, these kinds of mistakes tend not to spiral out of control. Individual errors happen, but arbitrary systems-level disasters do not. Spencer Greenberg, a big name in AI trading, says: “In the hands of people who don’t know what they’re doing, machine learning can be disastrous.”

bicentennial-man

Should You Market Your Algorithms Towards Large Businesses or Average Investors?

Some AI-powered trading will be hidden within the proprietary software of the legendary hedge funds we discussed earlier. Very few people will be able to access those funds. That means there’s a large market still looking for access to this service.

When you build algorithms for stock trading, you will probably want to target your services at banks. It’s going to be hugely important to network with the people at these institutions who have enough authority to purchase and use your services. They are your true target market.

If you’re looking to build advanced stock trading AI, you won’t be selling it to average consumers. If everybody has the best AI, it’s the same as if nobody had it. It’s like two computers playing each other at chess with one algorithm – it will average out to a tie game.

Ben Carlson summed up this line of thought while speaking to Wired: “It’s really hard to envision a situation where it doesn’t just get arbitraged away.” The more people that have access to AI trading, the less effective it becomes.

This isn’t a Warren Buffet or Peter Lynch kind of situation. If Warren decides to switch his investments to a new fund, he’s still one of a kind. The old fund loses him and the new fund gains him. With AI, once everybody gets access to the best program, it’s like a million Warren Buffets all competing with each other.

blog-banner-download-project-specs-2

Aim to sell your advanced services to the banks. They have deep pockets, so they’re a pretty good customer to sell to anyway.

Individuals Can Almost Never Beat the Market – But They Have Other Needs

It’s important to understand that whatever happens with artificial intelligence in stock trading, most individuals won’t beat the market. Howard Gold, columnist for MarketWatch, points out that ““It’s not just that true stock-picking ability is as rare as, say, being a violin virtuoso or throwing a 95-mile-an-hour fastball; it’s that the profits from such talent are eaten up by trading costs or management fees.” These fees will remain a problem for average investors no matter how smart the technology gets.

Darren Wu of wisebread.com offers a related piece of advice for average investors: “The answer, [Darren] concludes, is to stick to the basics: asset allocation, diversification and rebalancing”. This is relevant to your job as a developer, because it shows a way that AI-powered trading could become available for everybody. It’ll be a good tool for regular people who want to automate their basic investing plans.

It would be quite lucrative to sell a “average investors’ algorithm” to a bank, for example, who would then use it to offer an automated investing service for their clients.

If you want to sell AI-powered trading services to individual consumers, look towards market leaders Wealthfront and Betterment as examples to learn from. Focus on automating simple, straightforward long-term investing.

The System Is Evolving. Stay Alert!

Keep an eye out for new developments in the AI-powered stock trading market. These things are changing every day, and it’s important to stay up to date on your opportunities and on the competition.

Yesterday’s best AI won’t stand a chance against the current model. Remember that investment is a zero-sum game and that you need to create a premium product to attract interest from the real customers: The banks. Use the lessons from this article to help your business succeed in this exciting new field.

Alexey Semeney

Alexey Semeney

Founder and CEO at DevTeamSpace
Alexey Semeney