The Backtester

The backtest is beautiful. High Sharpe, low drawdown, eye-popping annualised return. A handful of live trades on it. Already considering modifying the parameters. You'll build a hundred backtests. You'll never run one long enough to find out if it works.

The backtest is the only part of trading where you're in control. You set the rules. You run the data. Output confirms or refutes without requiring you to sit in a losing position wondering if you're wrong. The backtest has no emotional cost. That's exactly why you keep running it.

Live trading asks a different question. Not "what would have happened?" but "can you execute this plan while it's happening?" After a handful of live trades you don't have an answer — you have noise. No strategy survives a tiny sample. You've pivoted before the distribution even started to reveal itself.

Technical term is overfitting, but not to market data. You're overfitting to the backtest environment — perfect information, no slippage, no emotional drag, infinite patience to find the parameter set that looks cleanest in hindsight. The live account is not that environment. You'll never know whether you have an edge because you've never run anything long enough to find out.

The next backtest isn't research. It's avoidance in a spreadsheet.

New backtest running. Mean-reversion signal with a volume filter on a broad-market proxy, multi-year sample. Annualised return looks excellent. Max drawdown looks tame. Sharpe looks elite. Found it.

Open live account. Capital allocated. Take the first couple of trades. One wins. One loses. Sample insufficient. Open a new backtest window. What if you added a trend filter? Different lookback window? Run a pile of parameter combinations. One improves the Sharpe further. Restart paper.

The original strategy, run live for a quarter, would have produced enough trades to start saying something. You have a handful. You pivot.

Somewhere in a brokerage account the original strategy is still trading — in the data, in the backtest, in the spreadsheet. Never in the live account. Never long enough to know.

Pick the current best strategy. Write the rules — all of them, on paper, today. Underneath: "I will not modify, pause, or pivot until I have 100 live trades recorded." Sign it. Date it. Photograph it.

Close the backtest software. No new backtest until 100 live trades on the current strategy. The urge to refine isn't intellectual curiosity — it's discomfort dressed as research. Research is what you do before you have data. After a handful of trades, you don't have data. After 100, you do.

One hundred trades. At your frequency, a few months. In that time you'll know more about whether the strategy works than you could from any amount of additional backtesting. The backtest can't tell you how you behave in a losing streak. The 100 trades will. Go get them.