The seduction
Complexity sells itself. A system with twelve indicators, four filters, and a machine-learning overlay feels like it must know something a two-rule system doesn't. The equity curve in the backtest is smoother. The pitch sounds smarter. And the whole edifice quietly commits the deadliest error in quantitative trading: it mistakes fitting the past for predicting the future.
Overfitting, in plain English
Historical price data contains some signal and a great deal of noise. Every parameter you add (another indicator, another threshold, another exception) gives the system one more degree of freedom to wrap itself around the noise. Wrap tightly enough and the backtest becomes flawless, because the system has effectively memorized history. Then you trade it live, the noise arrives in a pattern history never showed, and the "edge" evaporates.
The tell is fragility. Change the 14-period lookback to 15 and the profits vanish? Then there was never an edge; there was a coincidence with a parameter attached. Real effects, like trend and momentum, are robust: they show up across a wide band of parameters, across markets, and across decades. The 200-day moving average works about as well at 150 or 250 days. Nothing about it is precise, which is exactly the point.
Complexity hides risk
A second, sneakier failure: every added component is a place for risk to hide. A system you can't fully explain is a system whose losses you can't diagnose. When a two-rule system draws down, you know exactly what happened: the trend reversed, the stop paid. When a twelve-input model draws down, you know nothing: is this normal variance or is the thing broken? You can't tell, so you can't act, so you either freeze or abandon it, usually at the bottom.
Which is the third failure: abandonment. Every system's real test is whether its owner keeps trading it through the inevitable losing streak. Confidence during a drawdown comes from understanding, and understanding shrinks as complexity grows. Complicated systems don't just fail statistically; they get fired by their own creators at the worst possible moment.
The professionals already ran this experiment
The most famous demonstration is the Turtle experiment: complete beginners trading a simple, fully disclosed breakout system to reported nine-figure profits (we like the story so much we built a whole site about it). Decades of managed-futures results tell the same story: the strategies that survived forty years (trend following, momentum, carry) are embarrassingly simple, executed with obsessive discipline. Andreas Clenow's Following the Trend shows a full professional implementation and it fits on a page. The sophistication of serious shops lives in risk management and execution, not in secret signals.
What simple looks like
- Few parameters, none of them precise, all of them survivable when wiggled.
- A one-sentence reason the edge exists, involving other people's predictable behavior.
- Rules a stranger could execute from the written page, without asking you anything.
- Risk decided in advance (1% and a stop), so no single trade matters.
Ready to subtract?
The playbook walks through simplifying your trading down to what earns its place.