How a strategy is tested

Every strategy is tested thoroughly before it earns any trust:

• 1-minute bars first, a fast, coarse pass to see whether an idea has any edge at all.
• Tick-by-tick NBBO next, the strategy is replayed against the actual national best bid/offer, quote by quote, instead of tidy round-number prices.
• A purposeful execution delay, orders fill with a deliberate latency, so the test can’t cheat by trading on prices it couldn’t have reached in time.
• A ~40% fill assumption, not every resting order gets hit, so only a realistic fraction is counted as filled.
• Spread filters, names with wide or unstable spreads are screened out, since the quoted edge there evaporates on the round trip.
• Volatility-based sell stops, every position carries an adaptive stop and a timed exit, so losers are cut in the test exactly as they would be live.

Only after a strategy clears that gauntlet does it move to paper trading on live market data. If it holds up there, it earns real capital; if it fails at any stage, it goes back to research for a new idea. The distance between a naive backtest and this one is the whole point, it’s what keeps a strategy that looks good on paper from quietly losing money in the market.