Rolling ICIR: measuring the persistence of signal ranking

A signal can look brilliant on one date by chance. Capital should depend on whether that ranking skill repeated often enough—and steadily enough—to look different from a lucky observation.

Rolling ICIR summarizes that recent history by scaling average information coefficientInformation coefficient (IC)The cross-sectional correlation between a signal score and a subsequent return. Rank IC uses ranked values and measures whether the signal orders securities correctly.Open glossary entry → by its variation through time. It turns a sequence of dated results into a moving view of signal persistence.

Definition

Let \(\operatorname{IC}_{i,t}\) be the cross-sectional information coefficient for signal \(i\) at observation time \(t\). For a trailing window of \(T\) observations, define

\[\overline{\operatorname{IC}}_{i,t} = \frac{1}{T} \sum_{s=t-T+1}^{t} \operatorname{IC}_{i,s}\]

and the sample standard deviation

\[s_{i,t} = \sqrt{ \frac{1}{T-1} \sum_{s=t-T+1}^{t} \left( \operatorname{IC}_{i,s} - \overline{\operatorname{IC}}_{i,t} \right)^2 }.\]

The annualized rolling ICIR is

\[\operatorname{ICIR}_{i,t} = \sqrt{A}\, \frac{ \overline{\operatorname{IC}}_{i,t} }{ s_{i,t} },\]

where \(A\) is the annualization convention. strategynet.ai uses \(A=252\) for daily IC observations.

Adjustment for IC variability

Signals with the same mean IC can differ substantially in stability. The example below gives both signals a mean daily IC of \(0.02\):

Signal A receives the higher ICIR because its ranking performance varies less through time. ICIR describes the historical stability of mean IC relative to the variability of the IC series. Portfolio return is measured after position sizing and costs.

Trailing-window estimation

A full-history ICIR can combine periods in which a signal represented different economic conditions. A trailing window allows the estimate to respond when:

• a signal becomes crowded;
• its underlying economic relationship weakens;
• the investment universe changes;
• a market regime changes the relevant forecast horizon; or
• data quality or coverage changes.

Shorter windows introduce more sampling variation. A 20-day estimate can change substantially when one observation enters or leaves the window. A 252-day estimate changes more slowly and may retain evidence from a stale regime.

Window length is part of the model specification. Out-of-sample tests should set it before the reported performance period.

Application in allocation

For each FMP candidate, the current allocator maps rolling ICIR to an expected-return score:

\[\alpha_{i,t} = c\, \operatorname{ICIR}_{i,t},\]

with \(c=0.01\) in the current service. The scaling constant places FMPs on a common allocation scale before covariance and portfolio constraints are applied.

The score follows the stability of the underlying cross-sectional forecast. A strong recent portfolio return has little effect when its ranking evidence is inconsistent.

The score is then used in the nominal objective

\[\underset{w}{\operatorname{maximize}} \quad \alpha_t^{\mathsf T}w - \frac{\lambda}{2} w^{\mathsf T}\Sigma_t w,\]

or in the corresponding worst-case and CVaR extensions described in robust portfolio optimization.

Units and statistical limits

Rolling ICIR and the Sharpe ratio use different inputs. ICIR divides a mean cross-sectional correlation by the variability of that correlation through time. The Sharpe ratio divides mean portfolio excess return by portfolio-return volatility.

The conventional \(\sqrt{A}\) scaling annualizes ICIR. Formal statistical inference also requires treatment of serial dependence, overlapping return labels, changing universes, and common market shocks, all of which can reduce the effective sample size.

A complete report should state:

• IC convention: rank or Pearson;
• forecast horizon and signal lag;
• trailing window length;
• annualization convention;
• minimum cross-sectional sample size;
• handling of missing values and ties; and
• overlap of return labels.

Diagnostic checks

A rolling ICIR series should be reviewed alongside its components. At minimum, plot:

1. daily IC;
2. rolling mean IC;
3. rolling IC standard deviation;
4. rolling ICIR; and
5. the number of securities in each cross-section.

A rising ICIR can reflect a stable positive mean or a temporary decline in the denominator. The component series identify which effect dominates.

Further reading

• How Does Information Coefficient Decay Across Forecast Horizons?. How IC changes across cumulative and incremental return windows, and why the estimated peak does not mechanically set a portfolio holding period.
• When Is an Information Coefficient Reliable?. Why uncertainty in mean IC must account for serial dependence, overlapping return labels, sample size, and specification search.
• What Is a Meaningful Information Coefficient?. How magnitude, persistence, effective breadth, constraints, and costs affect the practical interpretation of an IC estimate.
• Information Coefficient vs Information Ratio. How ICIR over a forecast history differs from Information Ratio over a portfolio's benchmark-relative return history.
• Richard C. Grinold, “The Fundamental Law of Active Management”, The Journal of Portfolio Management, 1989. Relates forecast skill and independent breadth to expected portfolio Information Ratio.
• Edward E. Qian, Ronald H. Hua, and Eric H. Sorensen, Quantitative Equity Portfolio Management: Modern Techniques and Applications, Chapman and Hall/CRC, 2007 (ISBN 978-1-58488-558-0). Develops information-ratio estimation and stability in a practitioner portfolio-management context.