That strategy — did you backtest it?

Somewhere on social media tonight, someone is posting a chart with three arrows and the words "this setup prints money." And somewhere else, someone is about to fund a live account because of it. This column is for the second person — before the deposit clears.

In OKUGAI there is a question the residents ask before any other, and the workshop master Alma asks it with a smile that isn't entirely kind: "That strategy — did you backtest it?" Not "is it good," not "who taught you," but: have you counted what it actually does?

Why "trying it" is not "testing it"

Most beginners test strategies with real money, three trades at a time. Win twice — "it works." Lose twice — "it's broken," and the search for the next holy grail begins. The problem is sample size: two or three trades tell you nothing but luck. A coin that lands heads twice is not a heads-only coin.

Verification means separating the question "did I win?" from the question "does this rule have an edge?" The first is answered by your last trade. The second is only answered by a sample — and the practical minimum is 30 trades of one unchanged rule.

The minimum backtest, in five steps

Step 1 — write the rule in one line. "Buy when the 20-period average crosses above the 50, stop below the last swing low, take profit at twice the stop distance." If you cannot write it in a line, you cannot test it — vague rules produce vague data.

Step 2 — open past charts and scroll back. A free charting tool is enough. Hide the future (bar replay mode, or a sheet of paper on the screen if you must), and step forward candle by candle.

Step 3 — record every signal, not just the pretty ones. The rule fires, you log it: date, entry, stop, exit, result in R (multiples of your risk). Skipping ugly signals is how people falsify their own hopes.

Step 4 — collect 30 trades. Not 10. Ten trades still swing wildly on luck; thirty starts to show the rule's face.

Step 5 — tally five numbers: win rate, average win : average loss, expected value per trade, maximum drawdown, and the count itself. These five are the strategy's ID card. From here on you can speak about it in numbers instead of adjectives.

The three beginner traps

Trap one: peeking at the future. On a historical chart you can always "see" that the entry worked — because the right side of the screen is visible. Test bar by bar or the whole exercise is theater.

Trap two: changing the rule mid-test. Ten trades in, losing, you widen the stop "just a little." The data set is now garbage: it describes two different strategies. Note the improvement idea, finish the 30, then test the new version separately.

Trap three: forgetting costs. Spread and slippage eat a fixed bite out of every trade. A strategy that trades often can be profitable on paper and dead after costs — subtract them before you celebrate.

Verification is a substitute for talent

Here is the honest sales pitch: backtesting is boring. It is also the single most democratic tool in trading. You cannot buy reflexes or intuition, but anyone with a free evening and a spreadsheet can count 30 trades — and that count will quietly outperform most instincts on the market.

The residents of this city did not become residents by finding secrets. They became residents by refusing to trade anything they hadn't counted. The tower you're reading this from was built on tally marks.

Frequently asked questions