Automated Trading Strategy 87: Machine Homunculi
Modern-Day Alchemy
Important: There is no guarantee that ATS strategies will have the same performance in the future. Backtests and forward tests are used to compare historical strategy performance. Backtests are based on historical data, not real-time data so the results shared are hypothetical, not real. Forward tests are based on live data, however, they use a simulated account. Any success achieved with live trading is atypical. Trading futures is extremely risky. You should only use risk capital to fund live futures accounts and if you do trade live, be prepared to lose your entire account. It has happened to me several times. Learn from my mistakes and only use ATS strategies in simulated trading until you/we find the holy grail of trade strategy.
Paul Giamatti is a Golden Globe award winning actor and producer. You probably know him from his role on Billions, but my favorite character is the role he played on Lodge 49 as author extraordinaire and fellow alchemist, Lamar Marvin Metz.
As much as I love Giamatti, I am an even bigger fan of Jim Gavin, the creator of Lodge 49. When I found out that Gavin was going to be a guest on Giamatti’s podcast, it became a must-listen event.
Here’s the podcast—like Lodge 49, this particular episode is about secret societies and alchemy, two of my favorite subjects:
I listened to this episode and was struck by the parallels between the alchemical hunt and our collective hunt for the holy grail of automated trading strategies.
Here’s a short list of alchemists you may (or may not have) heard of:
Wei Boyang (2nd century AD) - Chinese alchemist, wrote the "Cantong qi"
Maria the Jewess (circa 1st-3rd century AD) - Alexandrian alchemist, invented several distillation devices
Nagarjuna (c. 150-250 AD) - Indian alchemist and Buddhist philosopher
Zosimos of Panopolis (late 3rd/early 4th century AD) - Egyptian alchemist and Gnostic mystic
Cleopatra the Alchemist (not the Egyptian queen, 3rd-4th century AD) - wrote on practical alchemy
Abu Bakr Muhammad ibn Zakariyya al-Razi (Rhazes) (854-925) - Persian physician and alchemist
Roger Bacon (1219/20-1292) - English philosopher and Franciscan friar
Ramon Llull (1232-1315) - Majorcan philosopher and logician
Paracelsus (1493-1541) - Swiss physician, alchemist, and astrologer
Mary Sidney Herbert, Countess of Pembroke (1561-1621) - English alchemist and writer
Robert Boyle (1627-1691) - Anglo-Irish natural philosopher, chemist, and physicist
Tycho Brahe (1546-1601) - Danish nobleman, astronomer, and alchemist
Isabella Cortese (16th century) - Italian alchemist, author of alchemical secrets
Isaac Newton (1643-1727) - English physicist and mathematician
These scholars contributed to the development of alchemy and, in many cases, to the transition from alchemy to modern chemistry and other scientific disciplines. Like traders, most were polymaths. Their work often combined elements of philosophy, pattern recognition in the natural world (math, biology, astrology, physics), mysticism, and early scientific inquiry.
Wouldn’t this group be surprised to learn that in 2024 The Nobel Prize for Chemistry would go to the CEO of Google DeepMind, or that the Nobel Prize for Physics would go to Geoffrey Hinton, also known as the "Godfather of AI"? It would seem that the modern-day alchemist is also a computer scientist.
What is Alchemy
“You are an alchemist; make gold of that.”
-William Shakespeare
Often dismissed as mad scientists, and perhaps with good reason, alchemists are known for many things, including crossing the line.
For some, alchemy is about being able to make gold. For others it’s about immortality. For others like Paracelsus—the eccentric Swiss wanderer who swapped cow dung for cleanliness in medicine, and subsequently died in a tavern brawl—it’s also about healing people. Still other alchemists, like the one depicted in Mary Shelley's Frankenstein, are interested in the creation of life, or homunculus. From homunculi to golems, ancient and medieval text is full of mankind’s attempt at “artificial” or unnatural creation.
And today, we’re in the midst of a Cambrian explosion in the world of artificial intelligence. If generative AI is as helpful to other disciplines as it is to trading, the next five years are going to be what could only be described as magical.
In a reverse Tower of Babel event, large language models (LLMs) are already performing a kind of linguistic magic. GPT4 can generate code, process images, and interpret 26 languages. Claude is self-described as a “translation tool” and supports over 50 languages, not including programming languages. Most recently Claude introduced the ability to communicate directly with your computer via Computer Control.
From the quest for immortality to the creation of digital lifeforms, AI is reshaping the world around us. Even the US government is concerned about the threat our new “baby Frankenstein" might pose to national security. Here’s an excerpt from the most recent memo on Harnessing Artificial Intelligence to Fulfill National Security Objectives:
Included in this memorandum is an invitation to the world’s best AI talent to join team USA in exchange for expedited entry.
LLMs have certainly changed my life. I’ve gone from just imagining certain trading strategies to actually creating and running them in minutes. I also use LLMs for other interests like bird watching, ancient text translation, plant identification and a host of random questions that probably should never be asked out loud, like: “If there are multiple timelines, how can I test the theory?”
Of course the hard part is knowing where to draw the line and I’ve added a few AI policy experts like Miles Brundage to my daily reading list for good measure, but convenience and high utility as a tool-set are highly addictive. I certainly wouldn’t have been able to share the following strategy with you without the help of LLMs. With over 700 lines of code, Strategy 87 is the most comprehensive to date, but it is also designed to do more than automate a trade.
Like LLMs, it uses a weighting system to provide context, and then assigns those weights a score based on a set of predefined conditions. The hope is to use this discrete weighting system as a first step toward an adaptive one. If you’re new to trading, the matrix can also be used as a way to think about manual trades.
Strategy 87: The Trader’s Homunculi (A Blueprint)
If we think of trading algorithms as modern-day, machine-based "homunculi," what I’m about to share with you could be seen as a blueprint for creating one; a set of rules and considerations that guide its decision-making process.