In May 2026, I started my journey into trading, having arrived there by way of the related — but very different — world of investing.
I hit the books, including some of the often-cited classics such as Jack Schwager’s Market Wizards series and Mark Douglas’s Trading in the Zone.
While on a trip to Cape Town, I found myself killing some time in a shopping mall and picked up a copy of Ross Larter’s The Beginner’s Guide to the Stock Exchange. It explained, from the ground up, many of the basic principles of trading. I read it cover to cover within a few days and made copious notes along the way.
Larter also runs a trading community through SureTrading, which I subsequently joined. Through his book and the community, I encountered several trading strategies for the first time, including the Opening Range Breakout (ORB), Trend Following Momentum (TFM) and a strategy based around fractals.
Over the following weeks, I consumed trading content voraciously. I listened to countless hours of the Chat With Traders podcast on Spotify. Traders who particularly impressed me were followed on X or YouTube. Books led to podcasts, podcasts led to videos, and videos led to yet more strategies and variations.
I became increasingly interested in one thing: finding an edge.
The questions begin
Despite being eager to start trading — and, naturally, to start making money — I deliberately held off from committing real capital.
Instead, I watched stocks. I applied the principles of different strategies and observed what would have happened.
These were pretend trades, but they quickly started raising real questions.
Would a 15-minute opening range produce better results than a 30-minute one?
Should I enter immediately when price breaks the opening range, or wait for additional confirmation?
Would placing an entry some percentage beyond the opening-range high reduce false breakouts?
What looked at first like a relatively straightforward set of trading rules quickly became a collection of variables.
And every change raised the obvious question:
Is this actually better, or does it just sound better?
From trading ideas to software
My background is in software engineering, and it wasn’t long before I realised that trading strategies needed to be treated much like any other hypothesis.
They needed to be tested.
In trading, of course, that means backtesting — something repeatedly emphasised by many of the traders interviewed in the Market Wizards books.
And if a strategy is going to be tested properly, it needs data.
Lots of data.
That was when Gecko was born.
I started by using the Interactive Brokers API to download historical candle data for a basket of stocks listed on the London Stock Exchange, deliberately choosing companies from a variety of industries.
Initially, Gecko’s job was relatively modest: reproduce the rules of an Opening Range Breakout strategy and determine when a trade would have been entered.
That didn’t last long.
Once Gecko could enter a trade, it also needed to know how to exit one. So I added stop-loss and take-profit logic.
Then I realised that merely knowing whether a trade won or lost wasn’t enough. Gecko needed to simulate an account, define a risk budget for each trade and correctly size each position.
So a pricing and position-sizing service followed.
At that point I thought I was nearly finished.
I wasn’t.
The project keeps growing
Individual results were useful, but I needed to compare thousands of trades and entire backtest runs.
So I built aggregation functionality to calculate and persist performance metrics, allowing different configurations and strategy variants to be compared.
Then I wanted to understand why one run performed differently from another.
Gecko therefore began exporting the individual trades, the strategy configuration and the aggregated statistics. These were then submitted to OpenAI using a structured prompt for additional analysis and insights, with that analysis stored alongside the backtest results.
What had started as:
Did this stock break its opening range?
had gradually become an entire research pipeline.
And then I moved on to the Trend Following Momentum strategy.
That required Gecko to understand considerably more than raw candle data. I added calculations for indicators including EMA3, EMA15, MACD and the Stochastic Oscillator, enriching the historical candles so that strategies could reason about both price action and indicator state.
That was a substantial step forward.
And it still isn’t finished
I could continue listing the functionality that has been added since, but in truth the work never really ends.
Every answer seems to create another question.
What I have become increasingly interested in is not finding a single magical trading strategy, but asking very specific questions and then using historical data to find out whether the assumptions behind them actually hold up.
Sometimes the answer is encouraging.
Sometimes an apparently sensible change makes a strategy worse.
Sometimes changing a parameter barely affects the result at all.
And sometimes an interesting observation turns out not to be a tradeable edge.
All of those outcomes are useful.
In the posts that follow, I’ll share some of these experiments in much more detail: narrowly focused questions about trading strategies, tested across Gecko’s historical dataset, together with what the results actually showed.
Because one of the most important things Gecko has taught me so far is this:
An insight does not have to be a trading edge to be worth recording.