Backtesting and Simulation

Free CFA Level II lesson in Portfolio Management. 14 min read, ~2,057 words.

Three core biases: survivorship (failed entities excluded), look-ahead (future data used), data-snooping (best of many tests reported). Out-of-sample testing is the strongest defense, in-sample performance alone proves nothing. Walk-forward analysis beats a single split because it mimics real-time model updating. Bonferroni correction = significance level / number of tests, raises...