Backtest returns look great—so why am I still losing money in live trading?
A trading strategy performs well on historical data, with high annualized returns, an impressive win rate, and relatively small drawdowns. But once you invest real money, you find that the profits are nowhere near as good as expected—or even that losses keep piling up.
The problem may not be that backtesting is useless, but that there are several easy-to-overlook gaps between historical simulations and real trading.
First, fees steadily eat into returns.
If a strategy trades frequently, even seemingly small fees on each trade can add up and significantly change the final result. This is especially true for strategies with limited profit margins per trade, where trading costs can wipe out the edge entirely.
Second, slippage can cause execution prices to differ from what you expect.
A backtest may assume that trades can be executed at the signal price, but liquidity, order size, and market movements can all cause the actual execution price to differ. The gap can be even more pronounced during sharp market swings.
Third, parameter optimization can create the illusion of strong performance.
If you repeatedly adjust parameters just to maximize a strategy’s returns over a particular stretch of historical data, you risk overfitting. The strategy may seem to have learned market patterns, when in reality it has merely adapted to noise in the past.
Fourth, market conditions are constantly changing.
A trading logic that worked in the past may not continue to work in the future. As market participants, liquidity, and volatility patterns change, a strategy’s performance may change as well.
A more robust way to validate a strategy is to divide the data into training, validation, and out-of-sample test periods; account for reasonable fees and slippage; and use paper trading to observe how it performs in execution. Even if a strategy passes these tests, future profitability is not guaranteed.
The value of backtesting isn’t to prove that a strategy will definitely make money, but to help us identify where it might fail.
In the next article, we’ll explore how to tell whether a string of trading losses is a normal drawdown or a sign that a strategy has lost its edge.
A trading strategy performs well on historical data, with high annualized returns, an impressive win rate, and relatively small drawdowns. But once you invest real money, you find that the profits are nowhere near as good as expected—or even that losses keep piling up.
The problem may not be that backtesting is useless, but that there are several easy-to-overlook gaps between historical simulations and real trading.
First, fees steadily eat into returns.
If a strategy trades frequently, even seemingly small fees on each trade can add up and significantly change the final result. This is especially true for strategies with limited profit margins per trade, where trading costs can wipe out the edge entirely.
Second, slippage can cause execution prices to differ from what you expect.
A backtest may assume that trades can be executed at the signal price, but liquidity, order size, and market movements can all cause the actual execution price to differ. The gap can be even more pronounced during sharp market swings.
Third, parameter optimization can create the illusion of strong performance.
If you repeatedly adjust parameters just to maximize a strategy’s returns over a particular stretch of historical data, you risk overfitting. The strategy may seem to have learned market patterns, when in reality it has merely adapted to noise in the past.
Fourth, market conditions are constantly changing.
A trading logic that worked in the past may not continue to work in the future. As market participants, liquidity, and volatility patterns change, a strategy’s performance may change as well.
A more robust way to validate a strategy is to divide the data into training, validation, and out-of-sample test periods; account for reasonable fees and slippage; and use paper trading to observe how it performs in execution. Even if a strategy passes these tests, future profitability is not guaranteed.
The value of backtesting isn’t to prove that a strategy will definitely make money, but to help us identify where it might fail.
In the next article, we’ll explore how to tell whether a string of trading losses is a normal drawdown or a sign that a strategy has lost its edge.