Back-Testing Without Fooling Yourself

“The brochure shows a new smart-beta index that beat the market by 5% a year for fifteen years. But the index was launched last month. How can it have a fifteen-year record?”

• What a back-test is, and the questions it can and cannot answer
• Look-ahead bias and point-in-time data
• Survivorship bias
• Transaction costs, capacity and unrealistic prices
• Data snooping, multiple testing and overfitting
• The deflated Sharpe ratio, explained without heavy mathematics
• In-sample, out-of-sample and walk-forward testing
• Why live performance usually falls short of the back-test, and a checklist for reading other people's results