Use the 1/N rule for diversification under deep uncertainty

When you cannot estimate the value of each option reliably, spread resources equally.

Why it works

Under deep uncertainty — where the probabilities and payoffs of outcomes cannot be estimated with confidence — optimizing an allocation requires assumptions you don’t have data to support. The 1/N heuristic (split equally among N options) outperforms optimized portfolio allocations in out-of-sample real-world tests because it avoids the overfitting that makes optimization strategies fragile when conditions change.

How to do it

1. When you face a resource allocation decision with genuine deep uncertainty (time between projects, budget across experiments), default to equal allocation.
2. Only deviate from equal allocation when you have reliable data showing differential returns.
3. Revisit allocation after you have real outcomes — shift toward what’s working with evidence, not optimism.

Evidence

DeMiguel, Garlappi & Uppal (2009) compared 1/N portfolio allocation to 14 optimized strategies across multiple stock market datasets and found 1/N competitive with or better than the optimized approaches out-of-sample. (observational)

This result is specific to financial portfolio contexts with limited data and uncertain means. For decisions where returns are meaningfully different and estimable, optimized allocation can outperform 1/N.

Sources

• DeMiguel, Garlappi & Uppal (2009), Optimal versus naive diversification, Review of Financial Studies

Common mistake

Applying 1/N even when you have reliable data showing that options differ substantially — the heuristic is for deep uncertainty, not a substitute for evidence when evidence exists.