Out‑of‑sample R‑squared in finance is typically 0.03‑0.04, i...

Out‑of‑sample R‑squared in finance is typically 0.03‑0.04, indicating very low predictive power

Nang cites his ex‑wife’s research showing that a good out‑of‑sample R² for market predictions is around 0.03‑0.04.
This means that even the best models explain only a few percent of variance.
Consequently, relying on a single model’s predictions without diversification is risky.
Continuous model validation and ensemble approaches are needed to mitigate this limitation.

Rishi NangTitans Of Tomorrow01:18:05

Quote

“She saw something I was writing that talked about a good out of sample R squared in our world is like 0.03‑0.04.” — Rishi Nang

Quote

“Zero, remember, is the min and one is the max. And like 0.3‑0.4 is like successful.” — Rishi Nang

Nang explains why intuition is a subconscious synthesis of data, the limits of out-of-sample R-squared, and how over-confidence can be dangerous.

“Intuition is a subconscious aggregation of relevant data, not a mystical ability”

“Confirmation bias, recency bias, and gambler’s fallacy are amplified when traders over‑interpret small sample results”

“Financial markets provide a small, noisy data set that makes overfitting especially dangerous”