We talk a lot about how machines are being used more and more in finance. This is especially important in High-Frequency Trading (HFT) and Algorithmic Trading or algo-trading (AT). There is simply no way for humans to compete on these levels, as a few milliseconds means the difference between making money and losing it. These timescales are shorter than it takes to speak a whole word, and hence it is no place for screaming brokers.

Well, there isn’t really a place for screaming brokers anymore, because not only do computers dominate the super short trading scene, they appeared in the human-directed trading scene long before. Let’s look at a very brief history of the shift from brokers to computers, then let’s look at the consequences. Some are nice, some not so nice.

You can track HFT, AT, and AI on our personalised news platform, CityFALCON, here.

History of the Shift

Once upon a time, brokers crowded the NYSE floor to make trades with hand signals and loud calling. They also used phones to speak to each other and buy and sell. What an antiquated idea, no?

The first stirrings of electronic trading were in 1971 when Nasdaq was set up. It wasn’t a full-fledged system, and it used an electronic bulletin board, not computer commands. The CME set up its Globex system, a much more widespread system, in 1987 and fully implemented the platform in 1992. Commodities, treasuries, and FX were among the asset classes traded on it.

As the usual driver of innovation and technology adoption, lower prices and faster execution times drove more exchanges to become electronic. Today, there is no major exchange that uses pits and open outcry to trade. This electronification has made it much cheaper and faster to trade, which has allowed all of us (i.e., people without institutional connections) to trade.

Not to be left behind, the big banks and hedge funds all use electronic trading, too. However, there are some firms that use lightning fast trades to conquer the market. These are the HFTs and algo-traders. They are programmers and mathematicians, and they rarely do the grunt work of actually trading.


What is HFT? How Frequent is it? And why are Algorithms so important?

The fundamental characteristic of HFT is very short holding times (hence high-frequency and not simply high-speed). Trades are entered and exited quickly, even going as low as microsecond holdings. This is not speculation, but a number reported directly from the Bank of England. Whether holding an asset for one one-thousandth of a second is really considered owning it is a philosophical issue. For business purposes, it doesn’t matter. You bought it at Price 1, sold it at Price 2, and the difference is your profit or loss that you must record.

This is not really possible to execute via a human hand pushing buttons. Neurons simply do not move signals around your body fast enough to make these trades, let alone process incoming information and make a decision. Hence, algo-trading and HFT generally go hand-in-hand. Algo-trading (AT), in the most basic sense, could also be applied to any kind of trading that uses an algorithm (whether human or computer processed is irrelevant). This includes long-term investment and days-long decisions. However, HFT is always done with algorithms due to time scales.

While sequential processing (step one, step two, step three) launched the AT industry, neural networks and parallel computing are likely to take the future. Machine learning uses many inputs, processes them the same way through many nodes, and gives some outputs. This kind of network is absolutely suited to parallel processing, so it seems future AT is likely to be parallel instead of sequential.

Another recent development is Big Data. The more data that can be collected, the more information there is to decide whether a trade is a good idea or not. With more data, more efficient markets might arise. Nothing is more logical and efficient than computers making decisions based on large amounts of data.

With this in mind, it seems that HFT and AT are ripe for AI and automation. In fact, it already is mostly automated, because it is not possible for humans to execute trades so fast. So, trading computers must be largely autonomous already. There are some firms that are fully automated, though, which is quite a feat: systems that make decisions by reading news and data and they have no human element.

So, how far can prices move in these tiny fractions of seconds? Well, fractions of cents (or pence, or whatever the base unit is in the traded currency). Often high-frequency traders (HFTRs) are looking for movements on the order of tenths of cents because they only need a few cents on each trade. If you make an average of €0.005 on a trade, but you make 100,000 trades in one day, you can still turn a tidy profit fast enough to even consider it a profit.

In two words, algo-trading is “computers trading”. As with anything else, there are advantages and disadvantages.

Investing 101 – Let’s get you started in the stock market.

The benefits and drawbacks of HFT

The first benefit is for the HFTRs themselves. They make money, and they do it off the noise in the market. The disadvantage is large, runaway losses due to bugs in the algorithm. Since you are probably not a HFTR, though, let’s look at the consequences for the retail investor.

For us, the main advantage is increased liquidity and decreased spreads. With so much trading happening, there is a lot of liquidity for retail investors. The HFTRs are exiting trades fast enough to let retail investors quickly get enough to fully execute their trade. It also causes the bid-ask spread to be lower. As a rule, liquidity and spreads have an inverse relationship.

If all the algorithms are correct, it can also increase market efficiency, at least in a technical analysis sense. With so much buying and selling, the price is more likely to approach the market demanded by the overall market. Problems arises, however, when those algorithms have issues.

One of the biggest drawbacks is the possibility of rapid or large swings in the stock price. A great example is the May 6, 2010 Flash Crash. After an investigation, it was found that HFT was not the initial cause, but it exacerbated the problem.

Runaway selling led to withdrawal from the market by many computerized systems (to avoid further losses), and this caused the spread to massively widen. It was all over in 36 minutes. Over one trillion USD were lost and regained in that time frame just on the DJIA. That surely does not reflect the underlyings, and hence is indicative of another problem: decoupling of the stock price from the company it represents. That is for another article, though.

Another disadvantage for retail investors is simply the inaccessibility of the technology. It is not possible for retail investors to trade like the HFTRs, because people don’t have the infrastructure at their houses. This is not really much different from most people being locked out of trading before the advent of online brokers, but it does show that the playing field was only leveled for a little while. Specialized firms (and of course Wall Street) have made sure to take advantage of HFT to the detriment of retail investors.

One very interesting problem is essentially deception. The HFTRs send many thousands of orders to buy or sell an asset. Since trading is largely decoupled from the underlying asset, psychological finance takes over. The flood of orders seems to indicate a change in sentiment or direction, and algorithmic trades, which can adjust quickly, can process the changes faster than humans. By the time humans understand the market dynamic, the HFTRs are already out of the trade with their predictable (i.e., created) profits. See quote stuffing, spoofing, and layering for more information (there’s plenty on the internet, but it is outside the scope of this article). These are all illegal now.

The publication that shone the spotlight on these practices was Flash Boys: A Wall Street Revolt by the same author that wrote The Big Short. The book is a great introduction to how HFT has forever altered the landscape of Wall Street, and it details the deception that can occur when prevailing market trends change faster than humans can think.


How it affects retail investors and traders

Clearly, something like quote stuffing will affect all of us, but it hits shorter-term retail traders especially hard. Even long-term investors, though, may incur losses if they have stops set that they don’t check often. Since HFT focuses on the noise in the market, long-term strategies are somewhat insulated from the effects. However, it could affect people who have set stops or are leveraged, because it could trigger margin calls and sell orders. As Keynes said, “Markets can stay irrational longer than you can stay solvent”. Even for people looking at multi-year investments, HFT might still cause some losses.

This only concerns the deceptive practices mentioned earlier. What other ways do HFT and algo-trading affect retail investors?

One is the inability of humans, especially single traders, to process information as quickly as a computer. This is true in all areas in which computers have become dominant. Its consequence in finance is that humans cannot enter a trade based on news or trends earlier than a computer. This can be earnings news, disaster announcements, bankruptcies, or management changes. It also means algo-traders (ATRs) can enter trades based on news that impacts the long-term before retail traders ever could.

This problem spills over into front- running. The most common example is traders buying up shares of a company that will soon be added to an index. Index funds need to track the index, which means they must buy the stocks that appear on it. But the announcement usually comes before the actual addition, and HFTRs can spot the announcement before anyone else can. By the time the index fund can update, the price has already appreciated. This does imply efficient markets, but is it fair? Since many non-engaged investors (i.e., your ordinary man or woman investing for his/her future) use index funds, this affects much more people than just those actively investing.

This practice is completely legal (ethical is another question) because all the information is public. Should it be? Does it keep the markets fair if ATRs can swoop in and push up the price before human retail investors can even read the first word of the news?

Other forms of front-running, like brokers buying shares before entering a client’s large order, is a problem regardless of HFT and AT. These blatantly unethical practices, however, are already illegal.

Since speed is so important, many HFT firms will buy data at a premium to have it early (or even gather it themselves). But many retail investors cannot afford their own personal Bloomberg Terminal, and hence are locked out of crucial information. If you are a long-term investor, you could miss out on the only gains your stock will see for a while. The gap in speed and information is growing, too, as AI becomes more involved in finance.

There is even competition between HFT firms to build networks that allow for faster quote and execution confirmation. With trades being executed in millisecond intervals and the lightning-fast processing of data, the slower the link to the central exchanges, the more risk present in short term holding.

Since HFTRs can adapt so quickly, a few milliseconds can be the difference between a buy and sell decision. Hence many HFTRs subscribe to “direct feeds” (DF), or lines of communication opened directly with the exchanges. It is required that exchanges provide the data to both the “Securities Information Processor” (SIP), which is probably where you get your “real time” quotes, and to the DF subscribers simultaneously. However, because of distances and bandwidth, some DF subscribers can get the data sooner. Hence the quote is delayed to retail investors. Furthermore, because the SIP will restamp the data, it isn’t detectable that the information is slightly delayed.

While this doesn’t affect retail investors much, as it is largely concerned with market noise and any long-term investor is unlikely to be worried about fraction-of-a-second changes, it demonstrates how important speed is to the HFTRs and ATRs.  It did become a problem during the Flash Crash, though.


AI Hedge Funds

Hedge Funds are often under attack for their practices. The general populace is not very trusting of them, and the lack of trust is understandable. They are often private firms with not much public oversight, but they make headlines for controlling the military assets of sovereign nations (like Argentina). However, they are the ones pioneering some FinTech that may one day affect you. A fully autonomous hedge fund is currently still a rarity, but they do exist.

This is an important development because systems that are entirely dependent on machines might provide a truly efficient market. AI that can rewrite itself will be able to improve its trading strategy over time. This could be great for long-term investors looking to share in the profits of companies over time. Fundamental indicators would be like classical laws of physics, and the AI systems would be like the quantum fluctuations underneath – for us, the classical predictions are good enough. On the other hand, full AI systems are dire news for the value investor, as there won’t be any undervalued stocks anymore. Unless, of course, the AI’s implementation is wrong. Then it could provide huge opportunity as a single contrarian voice could make a massive profit.

For now, though, the same problems arise from AI as from HFT and AT. Humans cannot adjust their strategies fast enough to compete, but those in it for the long-run should not worry too much. If you don’t require immediate profits, you can rest easy your strategy is probably still safe.


The Future

As with many predictions involving AI, there is the real possibility of a bleak outcome for the humans. As AI develops, it could push most traders, and even most financial service providers, out of the market. There is no need for a broker or advisor when computers can do it all faster, more accurately, and more inexpensively.

There are already so-called robo-advisors, and they generate investing strategies entirely without human intervention. There are services that personalise their data feeds to investors without any human intervention. These could threaten the human-curated data services that currently dominate the market.

Investors like Warren Buffett, who are held in high esteem, could be analysed and replicated by AI. There are certainly hidden and confounding variables in Buffett’s strategies, and AI and deep learning algorithms might be able to sniff these out, even if Buffett himself isn’t conscious of it. Does that mean the end of human competitive advantages? If the best of the best can be replicated in seconds with a machine, is there any reason to not expect fully automated markets, strategies, and the evaporation of opportunities for humans to make short-term profits?

Of course, as mentioned earlier, this might simply lead to extremely efficient markets and the rise of the long-term investor. Highly-efficient markets will see long-term strategies become dominant, because companies will still generate revenue to pass on to investors, and the world will continue to develop economically.


In Summary

Is HFT a threat to the retail investor? Probably not. It may actually provide advantages to retail investors because it increases liquidity and reduces spreads. For retail traders, though, it may become more of a problem, because there is no way for the home-based retail trader to compete on millisecond time scales.

Some of the drawbacks, like Flash Crashes, are possible, and because of their big price swings, could affect long-term investors who have placed stops orders or have leveraged positions. For long-term investors who are trading manually, such crashes may provide an adrenalin kick, but it probably won’t affect the return.

Front-running and quote-stuffing, of course, are problematic, but they are illegal because of unfair advantages afforded the HFTRs. Whether you believe the instant processing of large data sets can be construed as front-running is a philosophical question left up to you.

Overall, it seems HFT and AT do not affect retail investors much. As long as markets stay true to fundamental analysis trends and the price of a security does not decouple from the underlying long-term, there is little risk for retail investors from HFT and AT. In fact, some retail investors may benefit from a bit of AT, because it can help sort through the mountains of data available. If you can, why not take advantage of all that data? Everyone else does.

You can track HFT, AT, and AI on our personalised news platform, CityFALCON, here.