Test it before you trade it
A strategy you have not tested is a hypothesis you are funding with real money. Testing one is the cheapest thing you will ever do — and the results are frequently a relief, because most ideas die on history rather than in your account.
Your exits and your costs, not defaults
A backtest that assumes zero spread is a backtest of a market that does not exist. Costs are part of the run here, set to match your own broker, because the gap between "profitable before costs" and "profitable" is where most strategies actually live.
The same goes for exits. A backtest that only measures entries is measuring half a strategy — that is a different tool, and it is the entry check. A backtest here runs the whole thing: entry, exit, costs, and every trade that results.
How a run works
- Pick a strategy, an instrument and a timeframe. Any strategy you have built, on any market your plan covers.
- Set the costs. Spread and commission as your broker charges them.
- The run walks the history forward. Conditions are evaluated candle by candle, in order, with no access to prices that had not happened yet.
- You get the trades, not just a number. Win rate, drawdown, average return and the signal count — plus the window the run actually covered.
A finished backtest
- Strategy
- Mean Reversion
- Window
- 2019–2026
- Signals
- 412
- Win rate
- 63%
- Average return
- 1.9R
The signal count is deliberately near the top. A 78% win rate over 9 trades is noise; the same figure over 400 is evidence. Read the count before you read anything else.
The window is reported, not implied
Every result names the period it actually covered. A strategy tested over eight months and one tested over eight years are not the same evidence, and a result that quietly rounds the two together is the kind of thing that reads as fine right up until it costs you.
What this does not do
Backtested results do not indicate future results. A backtest is evidence about the past, bounded by the window tested and the costs you set. It cannot tell you that a strategy will keep working, and a strategy tuned until it looks good on history is usually tuned to that history — which is a real risk worth understanding before you use Refine. Treat a good backtest as permission to keep investigating, not as a forecast.
Questions
How far back does a backtest run?
Across the history held for that instrument and timeframe — typically several years on the majors and metals. The result reports the window it actually covered rather than implying more, because a backtest over eight months and one over eight years are not the same evidence.
Does it account for spread and commission?
Yes. Costs are part of the run, not a footnote applied afterwards, and you set them to match your own broker. A strategy that is profitable before costs and unprofitable after is the single most common way a backtest misleads.
What is the difference between a backtest and an entry check?
A backtest runs your full strategy including exits and costs, and reports the trades it would have opened and closed. An entry check ignores exits entirely and measures only what the market did in the bars after your conditions fired. Use the entry check to find out whether your entry has an edge at all, then the backtest to find out whether a complete strategy built on it makes money.
Can I trust the results?
As evidence about the past, within the window tested and the costs you set. Backtested results do not indicate future results, and a strategy tuned until it looks good on history is usually tuned to that history. Treat a good backtest as a reason to keep going, not as a forecast.