Trading Strategy Backtesting
Test a written trading hypothesis against historical examples and retain the assumptions needed to interpret the result.
Connect practice with review
The preview shows how a trading record can be presented alongside performance metrics. All displayed numbers are illustrative. Open the application to check the tools and data currently available.
Open the TGSFX application →Define the rules before the test
Write the entry condition, invalidation, exit rule, sizing method and eligible market hours before collecting outcomes. Specify the instruments and historical interval. If a rule needs a discretionary judgement, describe how that judgement is recorded so that another review can identify the same event.
A backtest result depends on the data and implementation. Record commissions, spread assumptions, slippage and treatment of gaps or missing bars. A chart that touches both a stop and target inside one bar may not reveal which order would have filled first.
Separate development from evaluation
Use one sample to develop an idea and retain another for evaluation. Repeatedly adjusting parameters to improve the same historical result can produce a model that fits that sample rather than a repeatable process. Keep a version log when you change a rule and do not silently combine results from different versions.
Read backtesting versus market replay to separate rule evaluation from execution rehearsal.
Review more than the final return
- Number of trades and the markets or sessions represented.
- Net expectancy after the stated costs.
- Average win and loss, profit factor and losing sequences.
- Maximum drawdown and the time spent below the previous peak.
- Sensitivity to nearby assumptions rather than one ideal parameter.
Use the statistics reference and calculators to interpret the sample. Historical or simulated performance does not guarantee future results.
Questions about this topic
What is look-ahead bias?
It occurs when a test uses information that would not have been available at the decision time.
Does a profitable backtest prove a strategy works?
No. Data quality, costs, rule changes and sample selection can materially change the result. A historical result remains conditional on its assumptions.