“We examine not to claim that any model works, but to understand when they do — and when they don't.”
Every study here is pre-registered: universe, step size and method sealed before the walk begins, point-in-time index membership, and a deflated-Sharpe correction that counts every attempt — including the discarded ones. Subscribe via RSS · methodology essays live in the Knowledge Base.
A pre-registered, point-in-time walk-forward study of S&P 500 in 1-year steps. Stitched out-of-sample Sharpe 0.45; deflated to 36% for 11 trials — it does not survive the multiple-testing correction — most likely an artifact of the search.