StructureX

Home › Guides › Practice

How do you know if your trading strategy works?

A strategy is not proven by a few wins or a beautiful chart. Start with written rules, test them without peeking, include costs and see whether the same idea holds on new data.

Beginner guide · 9-minute read · By Paspir · Updated 2026-11-14

The short answer

You need an honest record of how a fixed strategy behaved across different conditions, not a win streak. A useful test has five gates: clear rules, complete historical records, results after costs, a fresh sample and evidence you can follow the process. Failing one gate tells you what to fix next.

This is the decision page. How to backtest shows the mechanics of collecting trades. Here, you decide whether the evidence from that work is worth anything.

The five-gate decision flow

GateAsk this questionIf the answer is no
1 — RulesCould another person follow my entry, stop, target and skip rules?Rewrite the method; do not count trades yet
2 — RecordDid I log every qualifying historical setup without looking ahead?Redo the biased part of the test
3 — MathDid the full sample include losing trades, spreads, fees and slippage?Recalculate the result
4 — Fresh testDid I try the same rules on later, untouched charts or paper trades?Test a new sample before judging
5 — ExecutionCan I follow the plan and log rule breaks in practice?Drill the process; do not confuse execution errors with strategy evidence

There is no fixed pass percentage in this table. The gates decide whether you have a trustworthy test, not whether the strategy is certain to make money. A good historical record can still fail later.

Gate 1: a rule must be falsifiable

"Buy a strong chart" is an opinion, not a testable strategy. "Enter after the candle closes above a pre-marked level; stop below the last confirmed swing low; skip if the distance exceeds the risk budget" is something another person can check.

Write the conditions before you choose charts. Include what makes you not trade. Without skip rules, you may silently ignore all the bad-looking candidates.

If two people reading your plan would trade different candles, you are still at the rule-writing stage. The trading-plan guide helps make it specific.

Gate 2: count the opportunities you wish you hadn't seen

Suppose your made-up test contains 50 trades. You remember ten excellent ones and ten bad ones. If you throw out the bad ones because the chart looked "messy," the remaining sample no longer tests the original rule.

Keep the original chronological record, including days when the rule correctly said Wait. Tag any rule change as a new test version. Do not hide edits inside the same spreadsheet.

For why a sample of 20 or 50 can be noisy, see how many trades to test a strategy.

Gate 3: do the math in R, then include costs

R means one unit of the risk you chose before each trade. Here is a fictional fifty-trade test with every trade following the plan, before costs:

OutcomeTrade countAverage resultTotal
Winners22+2R+44R
Losers28-1R-28R
Combined50—+16R

The average is 16R / 50 = +0.32R per trade before costs. The win rate is 22 / 50 = 44%. The positive average comes from the average wins being larger than the average losses.

Now suppose the recorded total trading costs, expressed in the same risk units, add up to 20R across the fifty trades. The adjusted total is +16R - 20R = -4R, or -0.08R per trade. Costs have changed the verdict. You cannot declare the strategy sound based on the first total.

These totals are illustrations, not past performance. Converting fees into R requires each trade's actual planned risk. You can check a real sample with the expectancy calculator and expectancy guide.

Gate 4: test what you did not use to pick the rules

Suppose you tried five stop sizes on one year of charts and chose the version with the largest reported return. That year is the training sample. It is not an independent exam for your chosen settings.

Freeze the version and run it on later dates that were not part of the selection. Then forward-test with paper orders or a simulator, noting that simulated fills differ from real fills. Keep the forward-test rows separate from the backtest rows.

The backtest versus forward test comparison explains the two jobs.

Gate 5: can you repeat the plan under pressure?

A strategy can be measurable while the person running it ignores its rules. Track a separate process score: decisions that followed the plan divided by valid decisions. If a hypothetical 50-trade log followed the rules on 47 trades, that is 47 / 50 = 94%, but the three breaks still need labels and review.

An execution problem calls for practice. A weak result with consistent rules calls for a different kind of investigation. Mixing them makes both harder to fix.

Use the journal guide for the columns and how to measure trading progress for the process score.

When to reject, pause or revisit

Reject the version if its rules cannot be repeated, the test relies on seeing future candles, or it only looks good after hiding losing cases. Those are flaws in the evidence.

Pause and investigate if costs wipe out the result, the method only works in one market mood, or fresh samples disagree sharply with the first one. Do not keep increasing size hoping that a small sample will come right.

Keep studying if the records are honest, different conditions have been tested and the new sample behaves consistently. That still is not proof of future profit. It is a reason to learn the next thing carefully, not to skip the rest of the process.

StructureX Foundations offers simulated practice and journal tools; those can train the execution gate. A spreadsheet and free replay can cover much of the same decision process without a membership.

Questions

How many profitable trades prove a strategy works?

No fixed number proves future results. The rules, completeness of the test, costs and independent new samples matter more than a win count.

Can a low win rate still work?

In an honest arithmetic sample, larger average wins can offset more losses. Include trading costs before drawing conclusions.

Is paper trading proof of a strategy?

It can show whether you follow rules and how the method behaves without money at risk, but paper fills and emotions differ from live trading.

What if the backtest is good but the forward test is bad?

Keep them separate and inspect the mismatch. Possible causes include curve-fitting, changed market conditions, different fills and rule breaks.

Should I change the rules after every loss?

No. Changing rules after each outcome prevents you from collecting one consistent sample. Keep a versioned record and test changes on fresh data.

Follow along

XTradingViewFacebook