Information Ratio — The Metric That Tells Luck from Skill in Alpha
Two funds each delivered +6% alpha over their benchmark. One produced +5%, +6%, +7% in steady years; the other swung +25%, −15%, +8% and landed at the same average. Which is real skill? Information Ratio (IR) is the metric designed to answer exactly this question. This article covers the definition, intuition, what 0.5 / 1.0 / 2.0 actually mean, and how IR differs from the Sharpe Ratio.
1. Definition — Alpha divided by volatility
One line:
Two terms to pin down precisely:
Alpha (α)
Beta-adjusted excess return over the benchmark. Not a simple return difference, but the "excess that your beta cannot explain." Full definition in The True Meaning of Alpha.
Tracking Error (TE)
Standard deviation of (portfolio return − benchmark return). "How varied was your divergence from the benchmark." Typically computed from one year of daily returns; unit is pp.
So IR asks: "Is your alpha large enough to justify its volatility?" Same alpha — lower volatility ↑ IR, higher volatility ↓ IR.
2. Intuition — by example
Two hypothetical funds, built to illustrate the point. Assume their annual alpha over three years looks like this:
| Fund | 2024 α | 2025 α | 2026 α | Mean | TE (std. dev) | IR |
|---|---|---|---|---|---|---|
| A — Steady | +5% | +6% | +7% | +6% | 0.82pp | 7.32 |
| B — Volatile | +25% | −15% | +8% | +6% | 16.65pp | 0.36 |
The figures in this table are invented to show how IR is computed. They are not the record of any real fund, and the 2024–2026 labels are only for convenience.
Both funds: same +6% mean alpha. But IR: 7.32 vs 0.36 — roughly a 20× gap.
- A delivers similar alpha each year (5–7%) — consistent capability. Next year ~+6% is a reasonable expectation.
- B has a blowout year then a blowup — same average, but no predictive value for next year. Closer to luck.
3. Interpreting IR levels
Industry rule of thumb:
| IR | Interpretation | Example |
|---|---|---|
| < 0 | Consistently trails the benchmark | Index tracking would be better |
| 0.0–0.5 | Alpha drowned in noise | Majority of active funds (90%+) |
| 0.5–1.0 | Meaningfully active | Solid portfolio manager |
| 1.0–2.0 | Excellent | Top-5% manager |
| > 2.0 | Legendary / suspicious | Madoff-style fraud check warranted |
Multiple studies estimate only 10–20% of active funds achieve IR > 0.5 over the long run — the empirical basis for the claim that most active management is "expensive randomness."
4. IR vs Sharpe Ratio
Sharpe and IR look similar in form but measure different things:
| Metric | Numerator | Denominator | What it measures |
|---|---|---|---|
| Sharpe | Return − risk-free rate | Std. dev of return | Risk-efficiency of absolute return |
| IR | Alpha (return − benchmark) | Std. dev of (return − benchmark) | Risk-efficiency of excess-over-benchmark |
Sharpe answers "Is this fund's absolute performance good?" IR answers "Does it add value over the benchmark?"
- Index ETFs (S&P 500, etc.) can have meaningful Sharpe but IR is by definition near 0 (benchmark = itself).
- Active funds need both for honest judgment. Sharpe-only-good funds may just be running high beta in a bull market; IR-only-good with mediocre Sharpe may be a market-down period.
On Sharpe's limits and Sortino / Calmar alternatives, see The Sharpe Ratio Trap (Korean only for now).
5. Four IR-calculation pitfalls
① Measurement window — too short is noise
IR from 1–3 years of data is statistically weak. Academic convention is 5+ years. A fund with IR 3.0 one year often shows IR −0.5 the next.
② Benchmark choice — wrong pick warps everything
Comparing a large-cap fund to S&P 500 is fine; comparing it to NASDAQ small-caps makes both alpha and TE meaningless. See The True Meaning of Alpha section 5.
③ Return frequency — daily / monthly / annual
TE computed from daily returns differs in scale from monthly TE, so IR comes out different. Convention is to compute daily and annualize by ×√252. Always normalize frequency / annualization when comparing funds.
④ Fees — gross vs net
Marketing IR is often gross-of-fees. For an investor, net IR is what matters. A 1%/year management fee shaves IR by roughly 0.2–0.5.
6. Practical IR for individual investors
IR calculation is heavy for most retail investors, but it's useful in:
- Picking active funds / robo-advisors — don't stop at marketing returns; check IR (or the TE in the appendix).
- Self-assessment — once you have 2–3 years of data, compute your own IR honestly. Is your alpha consistent vs. an index?
- Strategy comparison — when backtesting momentum vs. value vs. dividend strategies, look at IR for consistency, not just alpha — otherwise you reward strategies that won in one specific regime.
Multifolios currently provides visual benchmark comparison (Return mode + SPY / KOSPI line). Automatic IR computation is a planned addition (personal alpha-analytics card).
Summary
- IR = Alpha / Tracking Error. Measures "can I expect a similar alpha next year?"
- Same alpha can give a 20× IR gap depending on volatility.
- IR 0.5 = meaningful, 1.0 = excellent, 2.0 = verify.
- Sharpe = absolute-return efficiency; IR = excess-over-benchmark efficiency. Read both.
- Four pitfalls: window, benchmark, frequency, fees.