Know the Psychology of Your Opponent
Understanding human psychology helps exponentially with the pursuit of success in any field, especially trading. In addition to understanding how your own emotions and behaviors impact position management, being able to identify how they affect the markets as a whole is imperative. This starts to get murky in today’s electronic market landscape, where you’re dealing with a vast, varied group of invisible participants, all implementing a multitude of strategies and technologies simultaneously.
Though you will never be able to ascertain with complete certainty exactly what the masses are thinking and feeling at any given time, educating yourself on the latest trading technologies, trends, and the psychology of those building and using them can help you exponentially.
Current State of the Market: Who and What You’re Trading and Competing with
The technology used to facilitate trading has been evolving since the inception of ancient bartering systems. With each incremental advance, changes occur in the way human participants interact with one another. Take for example the open outcry system, which dominated 20th century trading and allowed buyers and sellers to physically see one another while they verbally communicated their orders. As electronic trading began phasing in through the second half of the 20th century, it eventually replaced open outcry almost entirely in the 21st century. With this evolution came the loss of the trader’s ability to see the faces of, and communicate with, those they were transacting against.
Losing this physical, visual element, combined with a new dominant breed of algorithmic traders, has had a massive impact on the psychological aspects of trading and the markets today.
Today’s Algorithmic Trading
Though it’s difficult to quantify the exact percent of volume transacted algorithmically today, Reuter’s reports that automated trading accounts for 75% of all financial market volume. This is the number most market makers and research groups also estimate (some even surmise higher). Marko Kolanovic, global head of quantitative and derivatives research at J.P.Morgan, reported that only 10% of volume in stocks comes from fundamental discretionary human investors.
A survey conducted by J.P.Morgan on FX e-trading in 2017 collected responses from nearly 200 institutional FX traders to get an idea of where automated trading was headed. They found 71% of volume is done through e-trading, with traders planning to increase their use within the next year. Their biggest reason for trading electronically was competitive pricing, followed by depth of liquidity, and platform user friendliness. 38% of traders planned to increase their use of algorithms. Below are the surveyed institutional trader’s current five most popular algo strategies. For clarity, I added my own short explanations of what the algos generally seek to achieve, but since each is built and used differently, it doesn’t mean that’s how the traders used them.
Liquidity seeking/limit based: Seeking out pools of liquidity around the market to help ensure that as much of the order is priced at the limit as possible.
Market trading/pegged: Maintain a purchase price relative to the national best offer (NBO) or a sale price relative to the national best bid (NBB).
Time/schedule based: Likely time-weighted average price (TWAP) based algos.
Participation based: Generally used to minimize market impact by keeping position size at a defined percent of the overall volume of the asset.
Implementation shortfall: Difference between the price of the asset upon first execution to that of the order’s final execution price, including all fees, etc. (the lower the better).
Minds Behind the Technology: Who Builds the Algos that Keep the Markets Moving?
Based on the estimates, automated trading accounts for at least 75% of market volume. Since this number is expected to continue to grow, it’s imperative traders have a good understanding of how the minds of these algorithms work. It also helps to understand how the people that built them operate as well. Since institutional trading accounts for roughly 90% of the volume in most asset classes (95% in FX, less in options), having a rough idea of the general software development lifecycle (SDLC) for electronic trading products is helpful in understanding how they work. This might help non-algo traders get a better sense of who is behind the machines they’re trading against.
Right now in the institutional space, there’s a knowledge gap between traders and software developers, but it is starting to close and will continue to. Many professional traders who are market experts and veterans are just that – traders. They are not software developers. They’re very talented and skilled in the profit generating practices of trading the financial instruments they specialize in and coming up with innovative trading strategies, but they’re not always extremely technical. The inverse is often true for developers; some of the most talented, specialized computer scientists I’ve encountered are developers, not traders. There are of course exceptions, but this is what I commonly observed.
Bridging the Knowledge and Psychology Gaps
Within investment companies, these knowledge gaps are often overcome through old-fashioned collaboration and, after experiencing this process many times first-hand, I found it to actually be a very positive attribute. As a business analyst, I observed their personality differences, what drove and inspired them, and the unique psychological traits captured and replicated within the trading technology they created. The collaboration made the final product well rounded and robust, as diverse groups of minds had been infused into the development process.
A simplified example of the differences would be a trader who enthusiastically comes up with a very creative, complex trading strategy they want automated. By the time it gets to the developers to be coded, all aspects of the functionality will have been scrutinized for optimization of performance and elimination of weak spots that could pose as trading risks, negative edge cases explored, and vetted for expert feedback. The final product then contains both the trader’s creative strategy and the technologist’s thorough performance analysis.
Over the years, a smaller subset of hybrid traders has been growing. This group includes individuals equipped with a deep understanding of trading and the markets, as well as the technical skills necessary to actually build and implement their trading strategies. Today, most firms refer to individuals with this dual skillset as quantitative traders, or “quants,” and they’re highly in demand. This term is a bit ambiguous as “quant” used in an automated trading capacity describes various types of algo traders with technology skills (as opposed to a quantitative analyst who builds models and isn’t involved with trade execution).
The rise in popularity of this breed of trader and software engineer is a result of demand for flexible, agile traders who can best compete in sophisticated algo-dominated markets. Over the past decade, there’s been a mass exodus of the traditional, non-tech savvy traders losing their jobs to machines. For example, the MIT Technology Review reported that in 2000, Goldman Sachs’ US cash equities desk had 600 traders buying and selling for institutional clients. By 2017, there were just two human traders left. Automated trading programs have replaced the need for the majority of a trader’s manual work and today, roughly a third of Goldman’s staff is made up of computer engineers. This has had a massive impact on former practices of institutional price negotiation and relationship based trading.
Coalition, a U.K. research firm, reports that today around 45% of revenue from cash equities comes from electronic trading. They explain that algo trading is now permeating less transparent markets as well, such as FX and credit, with algorithms being designed to emulate as closely as possible what a human trader would do.
This is one of the best parts of algo trading psychology (and also a central component to AI); the machines are being built to mimic human decision-making. However, they’re doing it in a far more efficient manner, such as remaining unhindered by human emotions, sticking rigidly to their systems rules, and being able to accomplish vast amount’s of human work in a fraction of the time.
At larger, more regulated, and risk averse institutions, there will usually be a team of specialized trading and engineering professionals that’s more collaborative in nature. Some smaller, more aggressive trading firms seek the hybrid quant traders. An example would be development of ultra high frequency trading (UHFT) algos, which leverage advanced technology heavily and execute rapidly without human intervention. Quants also operate independently, trading their own account or starting small funds. Some smaller full-time solo traders are also getting involved in quant trading as well.
At this time, creation of the best possible algorithms that excel in both trading strategy conceptualization and tech implementation is still best achieved via collaborative development team. Right now, it’s a lot for one person to be both the best possible computer scientist who is expert level at Python, Java, C++, C#, Perl, .NET, R, MATLAB, etc. and also a full on trading expert with years of experience who knows everything about the asset classes and strategies the algo is meant to succeed in. Though an algo’s requirements generally come from a trader and are built by an engineer, most traders today do posses a solid understanding of technology.
Psychology of Trading Algorithms
One of the advantages of algorithms is their ability to follow their system’s rules exactly, unhindered by human emotion while making decisions and executions. Most investment psychology literature focuses on controlling the emotions that can result in negative trade performance. Emotions such as greed, fear, hope, regret, complacency, and over confidence can all be detrimental if they cause you to make irrational decisions, deviate from your strategy, and break your rules.
I’ve always understood technology very well and machines make perfect sense to me. But when it comes to people, I can never fully ascertain what is going on. Some people are very irrational and emotional and do things impulsively that don’t make much sense. I don’t care how well you think you know someone, every single person has their hand of cards held tightly to their chest and short of advanced fMRI brain scans, there’s nothing you can do to ever really know what they’re truly thinking or feeling.
This is why I find understanding how human brain chemicals work really helps me in better understanding people (in a good way). Certain things (food, sex, gambling, etc.) elicit the release of various chemicals and neurotransmitters (dopamine, norepinephrine, serotonin, oxytocin, etc.). In turn, these drive people to do things and exhibit different behaviors.
These chemicals are also all catalysts to what someone might think, feel, and experience while trading. Examples would be the euphoric feeling you get from a dopamine hit after a trade win, or norepinephrine’s fight or flight response in times of panic when the market moves against you.
When trading algorithms are built, they are not programmed to emote. They do not release any neurotransmitters that would cause them to ignore their exit signals and chase after the high of a previously winning trade to re-experience it. They also don’t have emotional biases, such as following Whole Foods Market’s ticker today instead of Perdue because they love healthy grocery chains and don’t believe in factory farming even though the latter would be a better trading choice.
Though the algorithms themselves are not swayed by human emotion, there can still be a human trader who intervenes in some way with the algo while it’s executing.
Tight automated risk controls can mimic human fear and panic in exiting positions during volatile climates. This is demonstrated with the snowball effect seen when a small drop hits a common stop loss level (whole number prices, for example), triggering the stops and further contributing to a rapid decline that turns into a fast and, for those in long positions, nasty plunge.
Future Trends in Algorithmic and Electronic Trading
One of the bigger trends in algorithmic trading is incorporating Artificial Intelligence (AI) and all it’s sub-facets, such as machine learning, into the algo trading process. AI is a broad term used to describe computer programs that leverage data in their environment to mimic actions that would normally require human cognition, such as complex, multi-step decision-making, to successfully achieve a goal.
Machine learning is a type of AI aimed at getting computers to take action without being directly programmed to do so. One way they “learn” is via feedback loops that confirm effectiveness of predictions. Deep learning takes this a step further by building computer programs to achieve success at things that require thought, such as being able to predictively “visualize” the near future based on current and past events.
From a trading psychology standpoint, the concept of neural networks, computer systems modeled on the human brain and nervous system, should give a clue on how to play the mass market psychology of algo dominated markets. Think about how the ideal super trader would react to a pattern or news event, and that is likely what how the AI trader will be programmed to do. AI brains are replicas of human brains.
After listening to a few hours worth of interviews with algo trader Bert Mouler, principal at AI and machine learning algo trading firm Profluent Capital, I definitely see the benefits in applying these technologies to algo trading. Such a vast amount of digital data now exists that being able to harvest and incorporate it into trading strategies could have a profound positive effect on the accuracy of the algo trade decision making processes.
For example, when a ticker’s price doesn’t reflect current news, there is a chance that those with positions in the ticker are not yet aware of the information. Algos that scan the internet and social media move a lot faster and could make price movements do what human psychology dictates they should do.
Though machine learning is being applied to trading, the practice is still in its infancy and not yet mainstream. In August of 2017, Gordon Ritter, adjunct math and finance professor at NYU and senior portfolio manager at GSA capital, produced one of the first academic studies demonstrating the success of machine learning when applied to trading.
How Can Individual Investors Participate and Stay Competitive in an Automated Market? Constant Education and Adaptation
To thrive in today’s technology dominated markets, understanding how algo trading works and the psychology of the minds replicated within them is imperative. Algorithmic trading has many benefits for both institutional and retail traders alike. Algos can harvest and incorporate vast amounts of data at high speeds; compute and trade faster; execute multiple strategies simultaneously; stick to their rules; and be programmed to “learn” as they performs via feedback loops.
But at the end of the day, everything they’re programmed to do came out of a human’s mind. The strategies’ behavior is a reflection and amplification of the mind, human psychology, and what we ideally desire to do. Machine learning demonstrates this very well as one of its prevailing themes is replicating humanistic qualities within the technology in an attempt to be as close to human as possible.
One of the most important pieces of being successful in the trading industry is constant learning and adaptation.
Knowledge is power, curiosity is an integral part of growth, and passion and fascination propel you forward.
Understanding what the algorithmic traders are using to automate their strategies is hugely helpful and can have the added benefit of providing inspiration for new practices to incorporate into your own strategy. In order to help understand the psychology of your algorithmic trading opponents, I’ve listed a few things below that I’ve done and found useful throughout my career in electronic trading.
Continual education: Educate yourself on the current technologies available that are being leveraged by traders.
Get ahead of the trend: Try to predict and get ahead of future technology trends by thinking of ways in which new, non-finance specific technologies could potentially be leveraged and implemented in trading. Believe me, if there is a notable technological advancement you’re hearing about, there is probably someone out there trying to leverage it in his or her trading strategy.
Hands on algo experience: Familiarize yourself with some of the new fintech companies offering front-end technology that allows non-coders to develop trading algorithms. Even if you don’t have an interest in building algos, they contain a lot of educational material on algorithmic trading. QuantStart, Quantopian, and Algoriz are good resources.
Keep up with fintech startups: Check sites like AngelList, TechCrunch, and FinTech Sandbox to see what technologies fintech companies are building and leveraging.
Remember your edge: As a manual trader, you do have some advantages over institutional traders leveraging advanced algos. Institutional traders aren’t as nimble and free, as there are regulations they must adhere to and different constricting corporate rules they must follow; trade risk parameters they must stay within; slowness of implementation and adoption of new technologies; and unfamiliar asset classes they must trade resulting from demand for multi-asset trading technologies. A retail trader putting through smaller orders doesn’t have to deal with the same nuances of being responsible for executing market moving orders.
J.P.Morgan: Institutional FX E-TRADING Trends in 2017
J.P.Morgan: only 10% of volume is from fundamental human traders
MIT Technology Review: Algos replacing human traders
Reuters: Algos responsible for trading 75% of market’s volume
Bank for International Settlements Market’s Group: HFT in the FX market
Gordon Ritter: Machine learning academic paper
Chat with Traders Podcast: Machine learning for algorithmic trading with Bert Mouler
Chat with Traders Podcast: Algo trader using automation to bypass human flaws with Bert Mouler