From Idea to Live: The Strategy Lifecycle

Most traders treat a new idea like a light switch — they hear about a setup, like the look of it, and start trading it with real money the same week. Professional strategy builders treat the same idea like a candidate going through a hiring process: several stages, each with a clear bar to clear, and most candidates rejected along the way. The stages have a name — the strategy lifecycle — and walking your ideas through it is the single biggest thing that separates a trader with a plan from a trader with a hunch.

The lifecycle was formalised for quant funds running automated systems, but the logic is the same whether you trade by hand or by code. The stages are: idea, then testing on history, then paper trading on live data, then small live, then full size. The point of the staged structure is brutal but kind: it is designed to kill weak ideas cheaply, before they cost you anything real.

The stages, in plain terms

• Idea. You can say why the edge should exist — what mistake other traders are making, or what structural quirk you are exploiting. "It goes up after it breaks out" is not a reason; "breakouts work because other traders chase momentum and I get in before them" is closer. Write the hypothesis down before you touch a chart.
• Testing on history. You check whether the idea would have worked on past data, with realistic costs and a fixed set of rules. This is where most ideas quietly die, and that is the system working.
• Paper trading. The rules are frozen and you trade them on live prices with fake money. This catches problems history can't show you — slippage, hesitation, rules that were ambiguous until a real situation forced a decision.
• Small live, then full size. Real money, deliberately undersized at first, scaled up only once live results track what testing promised.

Industry surveys of automated strategies find that only a small fraction of ideas that look good on history survive honest out-of-sample testing, and fewer still survive going live. The funnel is supposed to be narrow.

Why each gate exists

The reason the stages are separate is that each one catches a different lie. History tells you whether the idea ever had an edge — but a backtest is a perfect world where you always got filled at the price you wanted and never hesitated. Paper trading drags the idea into the real world of spreads and live data, but with no money on the line, so a mistake costs nothing. Small-size live is the first time your own psychology is fully in the loop, because now it is real money and the fear is real too. Skip a gate and you simply discover its lesson later, with more capital at risk.

A second reason the gates matter: the more versions of an idea you tried before settling on one, the more likely the winner is just luck. A trader who tweaks the moving-average length twenty times and trades the best one has effectively run twenty lottery tickets and kept the winner. The lifecycle's later stages — fresh data the idea has never seen — are how you find out whether you discovered an edge or just fit the noise.

Running the lifecycle inside FSP

You don't need a quant fund to run this loop — FSP is built around it. A Strategy in FSP is the written-down idea: its entry and exit rules, and crucially its validation and invalidation rules (what would prove the edge is real, and what would prove it has broken). Pairing that with a risk plan turns a vague setup into something testable.

From there the stages map directly onto features. The portfolio backtest stands in for "testing on history." Connecting a demo account lets you forward-test the frozen rules on live prices — paper trading, exactly as the lifecycle prescribes. And once you go live, tagging every trade to its strategy means Analytics can show you whether the live edge matches what the backtest promised, which is the real graduation test.

Put it to work in FSP: pick one idea, write it up as a Strategy with explicit validation and invalidation rules plus a risk plan, run it through the portfolio backtest, then forward-test it on a demo connection before a single real dollar is at stake.