Backtesting a Trading Strategy: What It Shows and Hides

What does backtesting a trading strategy tell you?

Backtesting applies a defined set of trading rules to historical market data. It estimates which trades the strategy would have taken, how often it traded and what its gains, losses and account fluctuations might have looked like under the test assumptions.

However, a backtest does not prove that a strategy will be profitable in the future. It is a historical simulation, not a live trading record or forecast. Its value depends on the quality of the data, the accuracy of the code and the realism of assumptions about spreads, commissions, slippage and order execution.

A useful backtest answers: “How did these exact rules behave under these historical conditions?” It does not answer: “How much money will I make next month?”

What a backtest can show

Whether the rules are precise enough to test

A testable strategy needs objective instructions. “Buy when the market looks strong” cannot be reproduced reliably. “Buy when the closing price crosses above a specified moving average and exit at a defined stop or target” can be translated into code and checked across historical data.

This process often exposes missing decisions. For example, the trader may not have specified what happens when a signal appears during an existing position, whether overnight positions are allowed or which price is used for entry.

Trade frequency and market exposure

A backtest can show how many signals occurred, how long positions remained open and how much time the strategy spent in the market. This helps distinguish an active intraday system from one that may wait days or weeks for a valid setup.

Trade count also provides context for the results. A large gain produced by a few unusual trades should be interpreted differently from a result distributed across many market conditions. More trades do not automatically make a strategy better, but a very small sample gives less evidence about how consistently the rules behave.

Historical gains, losses and losing periods

A backtest can calculate gross profit, gross loss, net result, average trade, consecutive losses and historical equity declines. These figures help traders evaluate whether the strategy’s behaviour would have been tolerable and whether one exceptional period generated most of the result.

Win rate alone is not enough. A strategy can win frequently but lose more on each losing trade than it makes on each winner. Another strategy may lose more often while keeping losses smaller than profitable trades. The complete distribution matters.

Performance in different market conditions

Historical testing can reveal whether a strategy behaved differently during trends, ranges, volatile periods or quieter markets. A trend-following system, for example, may experience repeated small losses in a sideways market and stronger results when sustained price movement develops.

This does not guarantee that the same pattern will continue. It does help identify the conditions on which the strategy appears to depend.

A simplified backtesting example

Consider a hypothetical forex strategy that enters at 1.1000, places a stop loss at 1.0950 and therefore has a 50-pip planned stop distance. Assume the account is $10,000 and the rule risks 1%, or $100, on each trade.

In a simplified USD-denominated example where one standard lot is treated as $10 per pip, risking $100 across 50 pips would require 0.20 lots: $100 divided by 50 pips equals $2 per pip. The test must recalculate position size if the stop distance or account equity changes.

If the software instead uses a fixed one-lot position, the simulated risk would be about $500 for the same stop distance before costs. That is 5% of the hypothetical account, not 1%. A backtest may therefore appear attractive or dangerously volatile simply because its position-sizing logic differs from the intended rules.

The simulation should also account for the possibility that the stop is filled below 1.0950 because of spread changes, a market gap or slippage. Assuming every order receives the requested price can understate losses.

What backtesting can hide

Look-ahead bias

Look-ahead bias occurs when the test uses information that would not have been available at the time of the trade. For example, code may use the final high or low of a candle to make a decision at that candle’s opening price.

The resulting trade could not have happened as simulated. Indicators should normally use completed data unless the strategy explicitly models how values changed during the open candle.

Overfitting and excessive optimisation

Overfitting happens when rules are adjusted repeatedly until they match historical noise. A developer may test many indicator periods, entry thresholds, stop distances and trading hours, then select the combination with the best past result.

The final settings may look precise but fail on unseen data because they describe accidental features of the test period. A robust strategy should not collapse when a parameter changes slightly. For example, performance that appears only with a 27-period setting but disappears at 26 and 28 deserves extra scrutiny.

Trading costs and execution limits

A backtest may omit or simplify:

• Bid-ask spreads and spread expansion
• Broker commissions and exchange fees
• Slippage between the requested and executed price
• Delays in sending or confirming orders
• Partial fills or rejected orders
• Financing, swap or contract rollover effects

These costs are particularly important for strategies targeting small price movements. A modest difference in entry or exit price can turn a small simulated gain into a live loss.

Missing intrabar price order

Basic candle data shows the open, high, low and close, but not necessarily the sequence of prices inside the candle. If both a stop and a profit target fall within the candle’s range, the test may not know which was reached first.

Some engines make an optimistic or pessimistic assumption. Higher-resolution data can reduce this ambiguity, but it still needs accurate timestamps and realistic order rules.

Poor or inappropriate data

Missing candles, incorrect prices, inconsistent time zones and duplicate records can distort results. US stock tests may need to handle corporate actions and changes to the list of tradable companies. Futures research may need contract rollover rules, while forex testing should distinguish bid and ask prices where possible.

Survivorship bias is another concern. Testing only instruments that remain actively traded today can exclude those that disappeared or became unsuitable, making the historical opportunity set look better than it was.

Future market changes

Markets change as volatility, liquidity, participation and economic conditions evolve. A rule that benefited from one historical regime may become less effective or behave differently in the next.

No amount of historical testing can include future events that have never occurred. This is why a backtest should be treated as evidence about past behaviour rather than proof of future performance.

Human and operational behaviour

A simulation follows every rule without hesitation. A person may skip trades after a losing streak, close a position early or increase size after a gain. Live software may also face internet interruptions, platform updates or incorrect account settings.

Backtesting usually hides these behavioural and operational risks unless they are modelled explicitly.

How to build a more reliable backtesting process

1. Write the rules before testing. Define the instruments, timeframe, entry, exit, stop, position sizing, maximum exposure and trading hours.
2. Check the data. Confirm the period, price type, time zone, missing records and treatment of spreads or contract changes.
3. Add realistic costs. Include commissions, spreads, financing and conservative slippage assumptions where relevant.
4. Separate development and validation data. Build the strategy on one period and evaluate it on historical data that was not used to select the rules.
5. Test different conditions. Examine trending, ranging, volatile and quieter periods rather than relying on one favourable interval.
6. Perform sensitivity checks. Change parameters, costs and execution assumptions slightly. Results that disappear after a minor change may be fragile.
7. Use walk-forward testing. Repeatedly develop on an earlier window and test on the next unseen window to approximate how the process would have advanced through time.
8. Forward-test before relying on it. Paper trading or controlled live observation can expose execution and operational issues that historical data cannot show.

Forward testing is not proof either. It simply provides another layer of evidence using current prices and real-time signal generation.

How to read a backtest report critically

Do not judge a report by its final net result or equity curve alone. Ask:

• What market, timeframe and historical dates were tested?
• Were the strategy rules fixed before the final test?
• How many trades contributed to the result?
• Did a small number of trades produce most of the gain?
• Were spreads, commissions and slippage included?
• Was position size fixed or adjusted according to risk?
• Was any data reserved for out-of-sample validation?
• How did the strategy behave when assumptions became less favourable?

Useful metrics may include average gain, average loss, expectancy, maximum historical equity decline, consecutive losses and time in the market. Expectancy can be expressed as the win rate multiplied by the average win, minus the loss rate multiplied by the average loss. It is an estimate based on the tested sample, not a promised return.

Evaluating backtests supplied with trading software

If a software provider presents backtesting material, confirm whether it represents a historical simulation, a live record or a hypothetical illustration. Check the instrument, dates, account assumptions, position-sizing method and included costs. Screenshots without methodology provide limited information.

Finoways builds algorithmic trading software and research tools for FOREX, COMEX and US markets. Its services information and frequently asked questions can help readers understand the scope of its tools. Finoways does not guarantee results, hold client funds or execute trades as a counterparty; clients use their own brokerage accounts.

Backtesting is a filter, not a forecast

A good backtest can reject unclear ideas, expose risk and show how consistently defined rules behaved in the past. It cannot remove uncertainty, reproduce every live execution condition or guarantee that a historical edge will continue.

The strongest approach combines realistic historical testing, unseen validation data, robustness checks and forward observation. Even then, position limits and predefined risk controls remain necessary.

Trading carries a risk of loss, and simulated performance can differ materially from live results. Review the risk disclosure before using trading software or placing trades.