How to know if your trading strategy has an edge

Most traders judge a strategy by how the last few weeks felt. That is the least reliable evidence there is. Whether a strategy has an edge is a statistical question, and it has a fairly simple answer once your trades are recorded the right way. This guide walks through the test step by step, shows it on a worked example, and covers the traps that make a strategy look better or worse than it is.

Step by step: testing your strategy for an edge

1. Fix your rules and write them down. An edge belongs to a specific set of rules: entry, stop, target, sizing, the sessions you trade. If the rules shift from week to week, there is nothing stable to measure. Note the date the current version started; only trades from that date count as evidence for it.
2. Record every closed trade in R. R is the result as a multiple of what you risked. A full stop-out is −1R, a winner that pays twice your risk is +2R, a scratch is about 0R. Use the result you actually got after spread, commission and slippage, not the one the chart promised. R takes position size out of the picture, so you are measuring the strategy rather than your account balance.
3. Calculate your expectancy. Expectancy is your average R per trade: add up every trade’s R and divide by the number of trades. Above zero means your wins outweigh your losses on average. It combines how often you win with how much you win and lose, which is why it says more than win rate alone.
4. Measure the spread. Two strategies can share an average and still differ wildly in how scattered their results are. A strategy whose trades land between −1R and +2R is easier to read than one that mixes −1R losers with occasional +10R runners. The more scattered the results, the more trades you need before the average means anything.
5. Put a 90% range around the average. The range combines the average, the spread and the number of trades into one honest statement: “the true expectancy most likely sits between here and here.” If the whole range is above zero, luck becomes an unlikely explanation. If it reaches below zero, a positive average could still be chance.
6. Check the sample size before reading any verdict. Below 20 trades there is no verdict worth reading; the range is simply too wide. From 30 trades, a range that stays above zero is promising. An edge only counts as validated at 100 or more trades with the whole range above zero.

That is the whole test. The rest of this guide is about doing it on real numbers and avoiding the ways it goes wrong. The exact formulas, including how the range is calculated, are in the methodology.

A worked example: same strategy, two sample sizes

Take a strategy that wins 45% of its trades at +2R and loses the rest at −1R. Its expectancy is 0.45 × 2 − 0.55 × 1 = +0.35R per trade, and that average is identical whether you have 40 trades or 100. What changes is how sure you can be.

Illustrative example, computed with the same engine as the free calculator.

At 40 trades: developing

The average is a healthy +0.35R, but the range runs from −0.06R to +0.76R. Because it still reaches below zero, the good average could be a lucky run. The right move here is not to change the rules and not to size up. It is to keep trading the rules exactly as written and keep recording. If the average and spread stay the same, the range clears zero at about 54 trades, which would read as promising; it would count as validated at 100.

At 100 trades: validated

Same average, but now the range runs from +0.10R to +0.60R. The whole range sits above zero and the sample has reached 100 trades, so the edge counts as validated. That still isn’t a promise about the next trade, or even the next month. It means the evidence so far is strong enough that luck is an unlikely explanation for the result.

What to record for each trade

The test is only as good as the records behind it. You don’t need much, but you need the same few fields on every trade, written down when the trade closes rather than reconstructed later:

• The date and the rule version the trade was taken under, so versions never mix.
• The setup, if your strategy has more than one, so each can be measured on its own.
• The realized result in R, after costs, including the small scratches and the trades you’d rather forget.
• Whether you followed your rules: a plain yes or no. This one field is what later lets you separate the strategy’s results from your execution.

Skipped trades matter too. If you passed on a valid setup, note it, even though it has no result. A pattern of skipping the setups that would have won is an execution problem the trade list alone will never show.

Common traps that fake an edge

Judging by win rate alone

A 70% win rate loses over time if the average loss is three times the average win: 0.7 × 1 − 0.3 × 3 is −0.2R per trade. A 35% win rate can carry a real edge if the winners are large enough. Win rate is only half of the picture; expectancy is the whole of it.

Cherry-picking a good stretch

Every strategy has runs, including strategies with no edge at all. Count every trade from the day the current rules started, not just the month that went well, and never restart the count after a bad week.

Testing on the same data you built the rules from

Rules tuned on past charts will always look good on those same charts. The evidence that counts is trades taken after the rules were fixed, whether live or forward-tested, because the rules had no chance to be fitted to them.

Mixing rule versions

If you changed your stop method halfway through, you now have two strategies. Blending them into one average tells you about neither. Measure each version separately, even though it means each one has fewer trades.

Counting trades where you broke your rules

A trade where you moved the stop or entered early tests your discipline, not your strategy. Keep those trades apart. Otherwise a strategy problem and an execution problem blur into one number, and you can’t tell which one to fix.

Leaving out costs

A small edge on paper can disappear once spread, commission and slippage are counted. Recording each trade at the R you actually realized keeps the costs inside the numbers instead of hiding them.

When the answer is “not yet”

Most traders checking for the first time won’t have 100 clean trades under one set of rules. That is normal. “Not enough data yet” is a real result: it tells you to keep the rules fixed, keep recording, and check again later. It is far cheaper to learn that slowly than to scale up on an edge that was never there.

If the range sits clearly below zero, though, more trades won’t rescue it. Look at the rules first, then check whether you followed them on the losing trades, and only then look at how consistently you sized positions. Those are the three places results go wrong, and the edge validation guide covers each one.

An edge can fade

A validated edge describes the past. Markets change, and an edge that held for a year can weaken. Keep recording after validation and watch whether the range keeps holding above zero; if recent trades start pulling it down, that is your early warning. Measuring an edge is not a one-off exam but a habit.