The Fundamental Law of Active Management

Two managers both claim genuine forecasting skill. Manager A makes one big macro call a year and is right 55% of the time. Manager B makes a thousand small stock-specific calls a year and is also right 55% of the time. Intuitively, which one should end up with the better, more reliable track record? The Fundamental Law of Active Management, developed by Richard Grinold, turns that intuition into an equation: skill alone doesn't determine performance, skill combined with how many independent times you get to apply it does, and the law shows the second factor can matter just as much as the first.

Why a casino trusts a thousand small bets over one big one

A casino doesn't win because any single hand of blackjack is a sure thing, the house edge on one hand is tiny and the outcome is essentially a coin flip weighted slightly in its favor. The casino wins reliably because it repeats that tiny edge across an enormous number of independent hands, and the law of large numbers turns a small, noisy edge into a highly predictable long-run profit. A single huge bet with the same edge would be far riskier, one unlucky outcome and the edge tells you nothing. Active management works the same way: a small forecasting edge, applied across many genuinely independent decisions, compounds into a much more reliable result than the same edge applied to one or two big calls.

The equation

$\text{IR} = \text{IC} \times \sqrt{\text{BR}}$

In words: your information ratio (IR, risk-adjusted active return, see Information Ratio) equals your information coefficient (IC, the correlation between your forecasts and what actually happens, a direct measure of raw skill) times the square root of your breadth (BR, the number of independent bets you make per year). A manager doubles their expected information ratio either by doubling their skill, or, just as effectively, by quadrupling the number of genuinely independent bets they make, because breadth enters under a square root.

This is why systematic, high-turnover strategies with modest per-trade skill (a small IC applied across thousands of independent positions) can post information ratios that rival or beat concentrated, high-conviction managers with much higher per-trade skill but far lower breadth. It also explains why a manager can't just trade the same idea a thousand times and call it high breadth: the formula requires the bets be independent. Ten correlated bets on the same underlying view count, for this formula's purposes, as closer to one bet than ten.

Worked example

Manager A: IC = 0.10 (a genuinely strong forecasting edge), makes 4 independent macro calls per year. $\text{IR} = 0.10 \times \sqrt{4} = 0.10 \times 2 = 0.20$.

Manager B: IC = 0.03 (a much weaker edge, barely better than noise on any single call), but makes 400 independent stock-specific calls per year (a systematic strategy trading a broad universe). $\text{IR} = 0.03 \times \sqrt{400} = 0.03 \times 20 = 0.60$.

Despite having roughly a third of Manager A's per-decision skill, Manager B's information ratio is three times higher, purely because of breadth. This is the core, counterintuitive result of the law: raw forecasting skill is not the whole story, and can be the smaller half of it.

Worked example: the cost of correlated bets

Manager C claims 200 "independent" bets per year, but on inspection, they are all variations on the same three or four macro views (rates direction, dollar strength, and two related sector calls), expressed across many individual positions that are highly correlated with each other. Grinold & Kahn's framework treats effective breadth, not the raw position count, so if those 200 positions really reduce to roughly 4 truly independent views, Manager C's actual IR is $\text{IC} \times \sqrt{4}$, not $\text{IC} \times \sqrt{200}$, a factor of 7 difference. A due-diligence process that counts position tickets instead of independent views will badly overestimate this manager's expected consistency.

What this means in practice

• Explains why quant strategies scale differently from concentrated stock-picking. A systematic strategy with a small, noisy edge per trade can still be viable, even attractive, if it can be applied across a genuinely large, diversified opportunity set.
• "Breadth" is a claim to interrogate, not accept at face value. Counting the number of positions is not the same as counting independent bets; correlated positions inflate the apparent breadth without delivering the corresponding IR benefit.
• Guides capacity and universe-expansion decisions. A manager whose edge is thinning per trade can often defend performance by expanding into more, genuinely uncorrelated opportunities, rather than by trying to sharpen an already-thin per-trade edge further.

Practice

1. Manager D has IC = 0.05 and makes 100 bets/year, but those 100 bets have an average pairwise correlation of 0.5, reducing effective breadth to roughly 20. Compute the naive IR (using 100) and the corrected IR (using 20). How large is the overstatement?
2. Explain, using the square-root relationship, why it's generally easier for a systematic strategy to improve its IR by expanding its investable universe than by improving its forecasting model.
3. If IC and breadth trade off through a square root, why can't a manager simply claim infinite breadth (trade everything, all the time) to drive IR arbitrarily high? What real-world constraint does the formula ignore?