Selected with hindsight: the trap in most strategy track records
Selected with hindsight means picking strategies by last quarter's result and then showing last quarter's result. On our own catalog the top 10% looked brilliant that way; chosen on June 30 and measured in Q3 instead, it returned $0.9–3.9k while holding all 16 returned $26.4k (source: PV_EDGE Master Data, DATA 2026-10-10).
Why does it always look brilliant?
Because the selection and the measurement use the same data. Whatever ranked first over the window will, by construction, show the best result over that window. The chart is true and the conclusion is false.
How do you make the honest version?
Fix a date. Choose on it, using only data before it. Measure after it. Walk-forward. On a Forex Swing 100K account the shortlist chosen on June 30 gave a different Q3 answer at every lookback; the whole catalog gave the same one, +$26,356, at all five.
What does that mean for a basket on a prop account?
That the basket shouldn't be picked by last quarter. The PVE Index holds every strategy and weights by volatility, which is the point of the whole test, written up in We picked the top 10% strategies.
The method, applied to the whole catalogue.
Every strategy and portfolio with modeled and live shown separately and the forecast behind each one. One PDF, then our letters with what the systems did.
By email, from Puerto Viejo, Costa Rica. Unsubscribe anytime.
Related
FAQ
Isn't every ranking selected with hindsight? A ranking over a past window is a description of that window. It becomes hindsight selection when the same window is used to prove the pick.
Was the ranking metric the problem? No. Net over max drawdown with a one-unit floor is a reasonable metric. The problem was picking by any metric and expecting it to hold.
Modeled and live layers are labeled separately. Monte Carlo figures (1,500 paths × 3 years) are a forecast. Hypothetical performance; past results do not guarantee future returns. Not financial advice.