Algo trading (AT) refers to using computer programs for trading. These programs can buy and sell stocks or other instruments, on their own, based on rules specified before. AT drives a significant percentage of the trading in global stock markets. Algo trading can help you buy and sell several securities at the same time in multiple markets. And without any common human errors. The new algo trading software uses machine learning and AI. Despite these advantages, algo trading has a few downsides. But given the development of big data, cloud computing and advanced machine learning algorithms, algo trading is all set to reach new heights.
Algo trading (AT) refers to using algorithms or computer programs for buying and selling securities. These programs can trade stocks or other instruments, on their own, based on previously specified rules. And algo trading drives a significant percentage of the trading that takes place in major global stock markets. Using algorithms, you can buy and sell a wide variety of assets in several markets at the same time. And take advantage of the price differences in different markets. The biggest advantage of automated trading is the minimization of human errors.
Algo trading dominates most stock markets in the world. For example, algorithms drive about 80% of the trades in the NYSE. And about 60% in the London Stock Exchange. In India, algo trading accounts for almost 40% of the transactions in the NSE. Algo trading, also known as high frequency trading (HFT), uses advanced machine learning and AI to analyze market data. This data includes company reports, financial statements and even social media. It also helps in executing trades without any human intervention. And this eliminates order entry errors and emotional trading mistakes.
But with this growing technological convenience, come a few downsides as well. There have been flash crashes in several markets. Plus, execution glitches and frauds. Regulators are playing catch up and increasing surveillance to take quick corrective action when required.
Advantages of algo trading
Algo trading has significant advantages over manual trading. Let’s look at each of them.
Risk diversification on multiple parameters
The trading software can handle trading in multiple exchanges, with different trading accounts. And the programs can trade in a range of assets and even execute different strategies in different markets. These algorithms do all this at the same time and with split-second accuracy! As a human trader, you will never be able to pull off a super-diversification strategy like this! The software does all this inexpensively because there are no intermediaries for order placement and execution.
Even if you are an experienced trader, there are times when you may lose your cool. And commit an error while placing an order and incur a huge loss. But trading software can perform according to plan at all times. And the software can place planned orders in milliseconds.
Development of super-efficient trading models
With the availability of advanced deep learning methods, you can make use of a lot of information such as quantitative variables that affect prices and other market indicators. And you can use market reports, news and even social media comments. Once you train the model, you can even test the models on new, hypothetical data to see if it works if the market reality changes dramatically. Also, your automated systems can even copy the strategies of top traders and benefit from that.
Disadvantages of algorithmic trading
There are a few disadvantages of algo trading that we should be aware of.
Flash crashes and risk amplification
Simultaneous execution of trades on a global scale can cause sudden market crashes. For example, in May 2010, the US markets crashed by 5%. And the Dow Jones fell about 1000 points in a day. And there were similar crashes in 2015 and 2019 as well. Though, it was not conclusively proven that high frequency algo trading caused these crashes, many market experts pointed fingers at automated program trading.
Over-optimization of trading models
Imagine that you are preparing for a very difficult academic test. And as a preparation strategy, you take all the previous test papers and learn the answers to those questions thoroughly. Now may be able to recall the answers to those questions easily because you have learnt them by rote and never really bothered to understand the principles behind the questions. If the actual test has completely differently-worded questions, you will not be able to perform well.
This is what happens when we train a trading model based on past data. It tends to get over-trained on past data. And we call it over-fitting. When presented with completely new data, these overfitted models don’t perform well. This can lead to huge trading losses, especially when the markets change course from the past.
Need for regular updating of models
During changing market conditions, you need to keep re-training the trading models with new data regularly. Also, once a lot of traders copy successful trading models, they become ineffective.
The need for constant monitoring of execution of trades
If you think that you can sit back and relax and watch your automated trading system making you a lot of money, you are going to be disappointed. Unfortunately, these systems need regular monitoring. This is because these systems can malfunction due to software bugs or power losses during trade execution. Also, the exchange software can malfunction and prevent proper execution.
For example, during the IPO of Facebook in 2012, failure of software led to a serious glitch and Nasdaq had to stop trading. And a leading Wall Street trading firm, Knight Capital lost almost US$ 440 million in a single trading session because of a trading software glitch.
Lack of explainability
Most trading programs will learn and adjust their strategies based on the data it is fed in real time. And most of the time, these models work like a black box and offer practically no explanation for a strategy- i.e., why a program did what it did. AI researchers are working on what we call ‘XAI’ or explainable AI. But we have a long way to go.
What are the well-known trading platforms?
There are several algo trading software platforms that offer software that you can use to trade a variety of assets across multiple stock exchanges. Most of them are quite easy to use. And you can back-test your strategies using real market data.
For example, Streak AI Technologies, helps you create custom strategies across stocks, commodities and currency futures. QuantRocket is another platform that offers you a wealth of data and takes the grunt work out of data wrangling that you normally have to perform to develop your trading model.
Yet another platform, Alpaca helps you with an API to trade for free. Blueshift is a free platform that brings institutional-grade algorithmic software to common investors, with the facility to back-test strategies.
And if you are a crypto-trader, CryptoHero, helps you trade in popular currencies such as Bitcoin and Ether using bots.
What is the future of algo trading?
Discretionary trading will be a thing of the past. With growing amount of big data and advances in cloud computing, AI is going to dominate trading. And with the rise of cryptocurrencies as a new digital asset class, the field is going to get even more complex.
Another impetus for competitive model-building is the crowdsourcing of best-performing trading algorithms by companies such as Numerai and Quantiacs. This will dramatically improve the quality of trading algorithms. So, very soon, institutional-grade algorithms may become available to even individual investors.
At the micro-level, trading algorithms seem to reduce the uncertainties of the market. But a massive number of computers trading against one another on a global scale, may result in higher volatility at the macro level.
And there is another important fact that doesn’t get discussed enough. Though the front-end, that is the trading systems, are getting more and more advanced, the backend software that handles settlements at several exchanges, is still quite primitive. Researchers are working on developing settlement systems based on blockchain technology to improve error-free settlements. Also, regulatory surveillance needs to catch up with the sophistication of the trading algorithms. So, our conclusion? The future of trading is AI-driven. And may the best robot win!