“Simon is an extraordinary trader in terms of his process and performance,” says Vadantia board advisor, Blair McPherson, who is helping to institutionalize the organizational structure and help launch a Cayman fund. The strategy has profited in every calendar year from 2017 to 2023, with annual returns ranging from circa 20% to 50%, and leverage averaging 6 times. Very few currency strategies have generated consistent long-term alpha, at least over the past decade, but Vadantia – a name adapted from Abundantia, the ancient Roman goddess of abundance, money flow, prosperity, fortune, valuables and success – is an extraordinary exception. “It is all built on our systematic process based on patterns of price movements, and even the discretionary parts are very process-driven, disciplined and procedural,” says Cotterill.

Vadantia trades various combinations of G10 currencies, against the USD and other pairs, as well as gold and silver. The strongest two years have been 2018 and 2022, when advanced pattern recognition worked especially well at identifying repeatable entry and exit levels. Market volatility in 2018 and 2022 was not the primary driver, and the opportunistic strategy itself has no fixed volatility target.

The strategy has had 3 double digit percentage losing streaks of 4-5 months: April-July 2020, January-May 2021, and January-April 2023, all of which were rapidly recovered from. “These were all within our range of statistical expectations. Emotionally, while these stretches were challenging, my response was measured and disciplined, reflecting my ethos to not react hastily but rather adjust systematically,” recalls Cotterill.

Losses are one reason to pause trading. “Most traders just love trading and pushing buttons, but we are disciplined not only about markets but also about our frame of mind. If we are tired, or have had an argument, it may be best not to trade. Psychological factors, mental capacity and a positive psychological state are crucial for bolstering confidence and enhancing cognitive clarity to sustain long term performance,” says Cotterill, who meditates as part of his daily routine, to regulate emotions and have a focused pause before work. Journalling trades consistently records the psychology behind them.

The strategy trades on average 3.8 times per day but there can be days – or even a month – without any trading. Intraday, there can be hiatuses around economic data to await reactions.

A structured multi-stage morning routine of analyzing trade timeframes, machine learning and statistical integration, helps to determine if there is an actionable trade. Discretion follows 65% and rejects 35% of model signals. “We do a lot of back testing to track missed opportunities and work out a repeatable system. The discretionary part has been valuable over longer periods to ensure a steady progression of smaller gains,” says Cotterill.

Though the average holding period is just a day, longer term trends such as the Yen’s multi-year decline inform strategic decision-making and trading across all time frames. “Longer term macroeconomic analysis complements shorter-term real-time market data, whereas AI-enhanced pattern recognition can pick up patterns over hours or days,” says Cotterill.

15 years of trading

Before the Vadantia track record, Cotterill had great success trading his personal account and developing his trading personality between 2009 and 2014. His trading was further refined at Triton between 2014 to 2016 and former LIFFE options trader, Jason Sen, was an important mentor: “Jason gave me my first start, supporting me through every step, and guiding my trading journey,” says Cotterill.

Prior to trading, Cotterill ran his own digital technology business (not connected with trading), which provided useful experience in managing cashflows, suppliers, clients, decisions and pressure. “I doubt if I would have succeeded at trading earlier. Entrepreneurship also taught me the importance of adaptability and innovation, which have been helpful in developing new trading strategies and markets. I anticipate market conditions rather than just reacting to them,” says Cotterill.

Eclectic influences

Cotterill draws on influences from physics, especially chaos theory and fractal geometry: “Structures and patterns that might appear chaotic actually have deterministic elements”. He is also something of a polymath, who heeds psychology and Bayesian analysis, behavioural finance, systems, programming, research and risk perspectives, and the research process is constantly evolving. “In developing my own trader personality, I have worked out what my strengths are and what I feel comfortable with, step by step. I read everything and cherry pick the elements that I believe work, adapt them, make them my own and plug them in,” he explains.

For example, Cotterill has adapted World Series of Poker champ Annie Duke’s decision-making framework to trading. “She rejects the cognitive biases of both ex-post and ex-ante ‘resulting’, which judges decisions only on actual or expected outcomes, in favour of a sound decision-making process. This should be based on logic, available information, probabilities and feedback loops, rather than outcomes,” says Cotterill. Though Duke was a poker player, Cotterill characterizes himself as more of a Blackjack player, making decisions under uncertainty, based on risk management, balancing risk and reward, probabilities, patterns and emotional discipline. Cotterill has also adopted Clare Flynn Levy’s concepts of Behavioural Alpha, using psychological and behavioural analysis. Multiple inputs have been synthesized into his EmoQuant process, which combines emotions and quant to improve self-awareness and promote continuous learning.

Psychology and behavioural finance

Trading psychology complements this data with human behavioural patterns including emotional influences, biases and collective behaviour. Insights from behavioural finance are based on cognitive biases, investor sentiment and decision-making processes. “They include Prospect Theory, which looks at how loss aversion leads investors to value gains and losses differently. Herding behaviour picked up in price and volume analysis can lead to bubbles and crashes. Overconfidence leads traders to overestimate their knowledge or predictive ability, increasing volumes and volatility. Anchoring bias leads to investors using reference points such as a previous high or low price, and can inform entry, stop loss and take profit levels,” says Cotterill.

Pattern recognition

Pattern recognition is at the core of Vadantia’s process. “I first became interested in pattern recognition while working for Eurostar when the transport police needed to identify somebody in a crowd, before the days of AI and computer power. I had a natural talent for recognizing patterns, and then built statistical analysis to reinforce what I could see. Now the discretionary filter makes it even more robust,” says Cotterill. Vadantia’s pattern recognition uses statistical and visual techniques to identify complex chart patterns for different currency pairs and timeframes, providing the foundation for market analysis. Beyond traditional momentum and mean reversion models, Vadantia has identified complex, non-linear fractal patterns repeated over different timescales.

Machine learning and AI

Machine learning and AI further enhance the strategy, using real time data to be reactive, forecast risk, optimize trade execution, and distinguish noise from signals so that false moves can be faded. “Advanced techniques, including supervised learning and deep learning, crunch through patterns and trends from historical and real time data to predict and anticipate market moves. AI digests the inputs and autonomously makes suggestions,” says Cotterill.

AI and machine learning also monitor changing patterns of correlations and volatility, which feed into an adaptive risk parity system designed to preserve capital and optimize returns. In early 2024, the real time models adapted to the positive correlation between gold and interest rates.

Machine learning statistical analysis and pattern recognition can also pick up and respond to regime shifts such as volatility and market momentum, which may arise from geopolitical events or economic data.

“We primarily utilize supervised learning for its ability to predict future market movements based on labelled historical data, which aligns well with our pattern recognition systems. Additionally, deep reinforcement learning plays a key role in our adaptive trading algorithms, allowing them to learn and optimize performance dynamically based on market feedback,” says Cotterill. In 2023, neural network capabilities were upgraded.

The data input split is 85% technical, 15% fundamentals and sentiment, but this smaller sleeve still sees innovation. New sentiment and news sentiment indicators use natural language processing (NLP) techniques to parse and interpret vast amounts of text in articles, financial statements and social media in real time.

Cotterill’s discretionary judgement still pulls the trigger on trades. He views the combination as greater than the sum of the parts, creating a robust framework to enhance market insights and trading effectiveness, adapting to complex market conditions.

In-house coding and systems

Cotterill is a self-taught Python coder, backed up by analysts, but acknowledges that the team may need a dedicated developer at some stage. The software is constantly evolving.

“Proprietary execution software systems were needed because generic off the shelf systems could not handle the necessary customization to precisely implement and integrate the advanced pattern recognition, statistical modelling and sentiment analysis techniques,” says Cotterill, who also points out that proprietary systems are more secure in terms of intellectual property and client data security.

Risk management

Vadantia has developed proprietary progression and accumulation metrics, which determine optimal take-profit and stop loss points that dynamically adjust to real time conditions and performance.

Trailing stop and take-profit levels were originally discretionary but are now statistically determined based on an algorithm using technical patterns. “We do not want to be watching screens all the time, as it leads to tinkering and emotional worry. We would rather have alarms set on the charting system,” says Cotterill, who can override the stop losses, but very rarely does. There is a special safety valve measure in managed accounts, Capital Guard, which liquidates all positions if a maximum preset drawdown is breached.

Stress tests use both historical and hypothetical scenarios. Markets might occasionally gap through stops and Cotterill has personally traded through gappy market events such as the Swiss Franc de-pegging in 2015 and Brexit in June 2016. Options are also used as a risk management tool.

A scalable strategy

Assets have grown to USD 32 million without any impairment of returns. Cotterill is confident that the technological, statistical and AI infrastructure can accommodate increased trading volumes, data processing, analysis and complexity while maintaining execution speed. Currently six execution venues are used as standard and Vadantia are exploring other possible liquidity providers, while managed account owners can choose their own counterparties.

Risk management will be adapted for larger trades and more instruments. The investment universe might add more currency pairs, and possibly platinum or palladium, but will probably not add emerging market FX. “To add bonds and equities, we would also need another trader. We would only add these asset classes after a lot of testing,” says Cotterill. Cryptocurrencies do not currently meet his criteria for liquidity, regulation and predictability.

Cayman fund

Some clients requested a Cayman fund and some in managed accounts may migrate to it. McPherson, who has spent nearly 30 years in investment consulting and banking at firms including Towers Perrin, RBC, BNP Paribas and Apex Group, has scoped Cayman fund structuring. Auditor, administrator, custodian and other providers have been identified. The strategy will be the same and use the same leverage parameters. Board and compliance will be expanded as assets approach USD 100 million.

Ongoing evolution and readying Gen AI

“Our philosophy of markets as self-organizing complex organisms has stayed 100% the same, but 70-75% of the systems have been developed, just like cells in the body renew every 7 years,” says Cotterill. Generative AI is the next frontier of research.

Vadantia currently uses AI mainly for alpha generation, portfolio construction and trade execution, as well as checking coding errors. Cotterill has already identified several areas where generative AI could be used in the investment and risk management process: “Simulating market scenarios and stress tests to find new opportunities or enhance existing strategies for alpha signal generation; generating synthetic data to test portfolio strategies under different conditions to make portfolio construction more robust and resilient; simulating trading environments to train more effective execution algorithms, reducing slippage and improving trade execution. We plan to assess the potential of generative models like GANs [generative adversarial networks] to enhance our market simulation capabilities”.

It can also help in the back and mid offices: “Streamlining routine tasks such as report generation, auditing financial documents, and detecting anomalies in transaction data, to enhance back office operational efficiency and compliance”.

Cotterill anticipates implementing Gen AI in the coming years but will integrate it cautiously: “We will ensure alignment with strategic goals and compliance with regulatory standards. We anticipate gradually testing and implementing Gen AI techniques as they become more mature and proven in the financial industry,” he says.