LuckOrEdge: the weekly lab for traders who want proof, not hype.

Issue 4 · Tuesday, November 10, 2026 · 4-minute read

Good morning. Last week, a strategy with no edge made nearly 30R in five months, then gave it all back. This week: why trading it live can't tell you which kind of strategy you have, and what can.

WEEKLY QUIZ Beginner

Last week's answer: C, about 1 time in 3

In our simulation of 22 stocks with no edge at all, the best one averaged +0.64R a trade or more 35% of the time. TSLA's three good years were about what luck produces when you look at 22 stocks at once.

The answer is hidden in today's issue. Find it, then vote. The answer arrives next Tuesday.

Your strategy truly wins 60% of its trades at 1:1. About how many trades does a test need to reliably detect that edge?

Login or Subscribe to participate

LUCK OR EDGE? · THE WEEKLY LESSON

Live trading is too slow to tell luck from edge

Plenty of traders skip backtesting: "I'll just trade it live and see how it does." It sounds practical. The problem is that the market hands out evidence slowly, and it charges you for every lesson.

One assumption. As in Issue 1, every example here uses a 1:1 risk-to-reward and ignores trading costs. Costs only make the numbers worse.

Two traders, 100 trades

Picture two traders who each take two trades a day. Trader A has a real edge: she wins 55% of her trades. Trader B has none. He flips a coin. After 100 trades, about ten weeks of trading, who's ahead?

Usually Trader A. But the coin flipper is ahead about 1 time in 5, and they're tied another 4% of the time. After ten weeks of real money and real stress, a trader with a genuine edge can easily look worse than someone guessing.

Bar chart: how often the coin flipper is ahead of the trader with a real 55% edge, by number of trades each. 32% at 20, 27% at 50, 22% at 100, 12% at 250, 5% at 500, 1% at 1,000.

How often the coin flipper is ahead of the trader with a real 55% edge, by number of trades each. At 100 trades, it's still about 1 in 5. At 1,000 trades, about 1 in 100.

How long it takes to prove an edge

To have an 80% chance of detecting a real edge at the 0.05 standard from Issue 1, you need far more trades than most people expect. At two trades a day, a 55% edge needs about 654 trades: roughly 16 months of live trading. A 52% edge needs nearly 4,000 trades, close to eight years.

Chart of trades needed to detect an edge by win rate: 65% needs 76 trades, 60% 169, 58% 258, 55% 654, 52% 3,945.

Trades needed for an 80% chance of detecting a real edge at 1:1 and the 0.05 standard, and how long that takes at two trades a day.

Backtesting: years of evidence in an afternoon

A backtest runs your exact rules over years of past data. In an afternoon, you can see hundreds or thousands of trades before risking a dollar. Our 3 Bar Play test covered 764 trades on 22 stocks. At two trades a day, that's about a year and a half of live trading.

It also shows what a few months live can't: your worst losing streak, your deepest drawdown, and how the strategy behaved in very different markets.

One warning: a backtest is only as honest as its design. Tweaking the rules until the results look good, using information that wasn't available at the time, or ignoring costs can turn a backtest into a luck machine. That's why we write our rules down before a test and seal the last year, as we did with the 3 Bar Play.

Verdict: more data. A hundred live trades can't reliably tell luck from edge. Even with a real 55% edge, a 100-trade test detects it only about 1 time in 4. A backtest gives you the sample size live trading would take years to build.

APPLY IT THIS WEEK

  1. Write your rules down. Make them precise enough that someone else could follow them exactly. If you can't, you can't backtest the strategy, and you probably aren't trading it consistently either.

  2. Backtest before you risk money. Aim for a few hundred trades across different markets, with commissions and slippage included. Can't code? Scroll back through your charts bar by bar and log every setup in a spreadsheet. It's slow, but it works.

  3. Use live trading to confirm, not to discover. Compare your live results with the range your backtest predicts: average result per trade, win rate and losing streaks. If they fall well outside it, stop and investigate before trading bigger.

TERM OF THE WEEK

Statistical power

If your strategy really has an edge, statistical power is the chance your test will detect it. Small samples have low power: 100 trades give a genuine 55% edge only about a 24% chance of showing up as significant. That's why "it didn't work after 50 trades" proves very little in either direction.

NEXT WEEK

The gold strategy that won 7 trades out of 10

I rebuilt a popular Micro Gold futures strategy trade for trade, then ran it once on three years of data it had never seen. Edge or luck?

Your turn

Hit reply: do you backtest before trading a new strategy? If not, tell me what stops you: time, data or coding. I'll tackle the most common answer in a future issue.

P.S. Want a strategy tested? Reply with its name and where you found it. Your requests decide what we test next.

LuckOrEdge is educational content only, not investment advice. The examples assume a 1:1 risk-to-reward and no trading costs; real results depend on your strategy, costs and trading. Past performance does not guarantee future results.

Keep Reading