What Is an Edge, Really?

intermediate7 min read

A repeatable reason your trades win more than they lose. If you cannot name it, you do not have it.

An “edge” is a repeatable, structural reason your trades make money over time — a genuine statistical advantage, like the small built-in edge a casino has over players. Without one, trading is just gambling with extra steps.

The brutal test of a real edge: if you can’t clearly articulate WHY it works, you probably don’t have one. A true edge has a cause — a behavioural bias others repeat (momentum, overreaction), a structural quirk (forced selling, index rebalancing), or an informational/speed advantage. “It worked in my backtest” is not an edge; it’s a pattern that could easily be luck or overfitting (covered later). An edge must answer two questions: why does this inefficiency exist, and why hasn’t it been arbitraged away? If your only answer is “the numbers looked good,” you’re standing on noise. Naming the mechanism is what separates a durable edge from a coincidence that will vanish the moment you bet real money.

Behavioural edge — exploiting a persistent human bias (e.g. herding, overreaction) that keeps repeating.
Structural edge — exploiting forced or mechanical flows (index rebalances, fund redemptions, tax selling).
Informational/speed edge — knowing or reacting faster (largely the domain of institutions/HFT).
The test — you must name why it works and why it persists; “the backtest looked great” is not an edge.

ExampleA momentum edge has a named cause: investors under-react to good news, so winners keep drifting up — a behavioural bias that’s persisted for decades across markets. Contrast “buy on Tuesdays when RSI is 47 and the moon is waxing” — it may fit the backtest beautifully, but with no mechanism, it’s almost certainly noise that won’t survive live.

Key takeawayAn edge is a repeatable, structural reason you win — behavioural, structural, or informational — like a casino’s built-in advantage. The test: name why it works and why it persists. “The backtest looked good” is not an edge; an unnamed edge is usually just luck waiting to disappear.

FAQs

How do I know if my edge is real or just luck?

Demand a *mechanism* (a clear reason it works and persists), then test it rigorously out-of-sample and across regimes (later modules). Real edges have an economic or behavioural cause and survive on data the strategy never saw; lucky patterns have no explanation and fall apart out-of-sample. If you can’t explain it, be deeply skeptical.