A strategy can look profitable because you measured the winners correctly—and the losers too generously.

Consider exit assumptions.

In research, a stop at -3% often becomes exactly -3%.

In live markets, the price can move through that level before sufficient liquidity is available. The realized loss becomes -3.4%, -4%, or worse.

That difference looks small until it repeats across hundreds of trades.

This is why serious backtesting should model execution asymmetry: entries may occur under normal conditions, while exits often happen when volatility and urgency are already elevated.

For eligible new users, CODE2026 can reduce qualifying Binance Spot trading fees by 20%, helping lower one known component of execution cost.

But conservative fee assumptions are not enough.

Your model should also assume that the market will occasionally make leaving more expensive than entering.

A backtest that gives every stop the exact price requested is testing the strategy you wish you could trade.

The real test is whether the edge survives the execution you are actually likely to receive.