Fact sheet, five-year track record
Sharpe ratio
a year
worst fall
Made-up fund. Past performance of a coin is not a guide to anything.
The Noise Fund
Launch a hedge fund made of coin flips.
One button. The machine flips a coin for five years of make-believe market days, tries 20,000 trading rules on it, and keeps the best one. Then it writes you a fund.
The market above is a coin. Heads, the price goes up 1 percent. Tails, down 1 percent. No pattern in it can last, because there is none.
Trying rules…
What just happened
You did what a lot of real fund launches do. Try many rules on old prices, keep the winner, print the fact sheet. If you try enough rules, one always looks brilliant, even on a coin. The trick has a name: overfitting. The fact sheet never says how many rules were tried.
The Sharpe ratio is return divided by bumpiness. Funds love to quote it. Above 1 sounds like skill. But the best of many tries has a floor that comes from luck alone. We call it the luck bar. For 20,000 tries on five years it sits near .
This is not a cartoon
In 2017, Hou, Xue and Zhang re-tested 447 published stock-market patterns. With a fair bar for how many people had been searching, 85 percent of them failed. Campbell Harvey and colleagues argued that a new pattern needs a test score above 3, not the usual 2, for the same reason.
Before you trust any backtest, ask one question: how many did you try?
Sources
- Hou, Xue and Zhang, "Replicating Anomalies", NBER working paper 23394 (2017). 447 anomalies; 286 insignificant at 5 percent; 380 (85 percent) with a t-cutoff of 3.
- Harvey, Liu and Zhu, "…and the Cross-Section of Expected Returns", Review of Financial Studies (2016). 316 factors; new factors need t above 3.0.
- Bailey and López de Prado, "The Deflated Sharpe Ratio", Journal of Portfolio Management (2014). The expected best Sharpe ratio from many tries, used here for the luck bar.
- Bailey, Borwein, López de Prado and Zhu, "The Probability of Backtest Overfitting", Journal of Computational Finance (2017).