Author
A backtest is one draw from a distribution of possible outcomes. With typical daily data and a few years of history, the 95% confidence interval on annualized return spans tens of percent. The headline number is honest. By itself, it is almost uninformative.
A backtest is an experiment producing data, not a screenshot proving a strategy. Every backtest return splits into predictive power plus drift-times-long-bias. Without detrending, a null model, and a p-value, an equity curve is decoration, not evidence.
A trading rule is a hypothesis if it is specific, falsifiable, and quantitative. The single hypothesis-test framework (compute a test statistic, simulate the null, count what fraction of simulations beats it) turns a TA claim into a falsifiable result or exposes it as vibes.
Excitement in trading is a warning sign. If your day produces emotional spikes, your positions are too big, your frequency is too high, or you are overriding the system. Good trading is engineered boredom: pre-computed signals, batched orders, vol-targeted sizing, and a fixed review cadence.
Trading is not about being right on every trade. It is about managing probabilities over hundreds of trades. A single win or loss means almost nothing. The edge appears only through repetition, discipline, risk control, and positive expectancy over time.
A trader running one system is one regime change away from irrelevance. Real longevity comes from portfolios of uncorrelated systems with different decay cycles. The goal is not finding the perfect strategy. The goal is surviving long enough to replace dying ones before they take you down with them.
The best-looking backtests are often the most fragile. Rules optimized to fit one market, one period, and one parameter set rarely survive live trading. Robust systems behave like loose pants: imperfect, flexible, and stable across many instruments, regimes, and parameter choices.
Markets do not repeat in exact shapes, regimes, or participants. What survives are deeper statistical properties: momentum, mean reversion, volatility clustering, fat tails, and lead-lag behavior. The edge is in measuring the rhyme, not memorizing the pattern.
Most traders focus on predicting markets instead of controlling risk. But survival, not prediction, is what compounds capital. Position sizing, drawdown limits, stop placement, and kill criteria matter more than being right about direction.
Using math, code, or machine learning does not automatically make trading scientific. A strategy becomes scientific only when it is falsifiable, benchmarked, tested against a null hypothesis, and replicable. Most retail “quant” content fails all four.
Most traders confuse a rule, a strategy, and a portfolio as the same thing. They are not. A rule generates forecasts, a strategy manages risk and sizing, and a portfolio allocates capital across systems. Mixing the layers is why many traders diagnose the wrong problem.
Most traders do not lose because their ideas are bad. They lose in the gap between signal and execution. Costs, sizing, psychology, bias, concentration, and strategy decay quietly destroy profitable systems long before the idea itself fails.