What is a good information coefficient?

There is no universal threshold, and in stock selection ICs are small. In Opulence Alpha's research papers on popular signal families since 1995, the families judged proven have headline rank ICs between about 0.01 and 0.03 in absolute value. A small IC that keeps its sign across years and sectors is worth more than a large one seen in a short sample.

What does a negative information coefficient mean?

It means higher readings of the signal were followed by relatively worse returns. That is information pointing the other way, not an absence of information. In Opulence Alpha's short-term reversal study, the more a stock had beaten its industry over three sessions, the more it tended to trail other stocks the following week. The rank IC was −0.027, with a t-statistic of −12.4 across 1,633 Wednesdays.

Does a signal's IC stay the same over time?

Often not. In Opulence Alpha's short-term reversal study, the next-week rank IC was −0.0445 in 1995–2004, −0.0221 in 2005–2014 and −0.0170 since 2015. The sign held in every decade, but the effect got weaker each time. That is why the studies report IC by decade, by sector and among S&P 500 members, not only as one average.

FAQ · Information coefficient

What is the information coefficient (IC)?

Reviewed against the platform's code on Sep 26, 2026

The information coefficient (IC) is the correlation between a signal's readings and the returns that followed. It is measured across many stocks on the same date and averaged over dates. The rank IC uses the Spearman correlation of the two rankings. Zero means no relation. A positive IC means higher readings were followed by relatively better returns; a negative one, by relatively worse returns.

Why it matters

IC scores a signal's skill before any portfolio, cost or sizing choice, so very different signals can be compared on one scale. In stock selection it is small, often only a few hundredths, such as 0.02. What matters is whether it keeps its sign across dates, years and sectors, not how large one reading is. Grinold's fundamental law of active management explains why a small IC can still matter: the information ratio is roughly the IC times the square root of the number of independent bets.

How it works

On each date, rank the stocks by the signal and by their return over a fixed horizon, correlate the two rankings, then average across dates (Grinold and Kahn, 2000). Ranks stop one extreme return from dominating. They also do not change when the market's or the median stock's return is subtracted. Long-horizon returns overlap from one date to the next, so the t-statistic should use non-overlapping dates or a correction for autocorrelation.

How Opulence Alpha applies it

Opulence Alpha's research papers on signal families report a rank IC. On each Wednesday since January 1995, each paper takes the Spearman correlation between a signal and the return over the next 1, 5, 21 or 63 trading sessions. It uses every stock in the 1,767-company universe that has both, gives tied values their average rank, and needs at least 50 stocks. The t-statistic uses non-overlapping dates only. A relation counts as proven only when |t| ≥ 3, the sign is the same in at least 60% of years, and the IC is at least 0.01 in absolute value.

How the studies measure rank IC →

Related concepts

Multiple testingMultiple testing (data snooping) is what happens when many strategies are tested on the same data: some will look significant by chance. How reality checks and step-down tests correct for it.Survivorship biasSurvivorship bias: testing only on companies that still exist. Why it flatters backtests, how to correct it, and how the Opulence Alpha studies handle it.Backtest overfittingBacktest overfitting is when a strategy is tuned to the noise of past data, so its backtest looks strong but its live performance decays. How to detect and avoid it.

All concepts →Sector behaviour in every market regime →

Questions

What is a good information coefficient?: There is no universal threshold, and in stock selection ICs are small. In Opulence Alpha's research papers on popular signal families since 1995, the families judged proven have headline rank ICs between about 0.01 and 0.03 in absolute value. A small IC that keeps its sign across years and sectors is worth more than a large one seen in a short sample.
What does a negative information coefficient mean?: It means higher readings of the signal were followed by relatively worse returns. That is information pointing the other way, not an absence of information. In Opulence Alpha's short-term reversal study, the more a stock had beaten its industry over three sessions, the more it tended to trail other stocks the following week. The rank IC was −0.027, with a t-statistic of −12.4 across 1,633 Wednesdays.
Does a signal's IC stay the same over time?: Often not. In Opulence Alpha's short-term reversal study, the next-week rank IC was −0.0445 in 1995–2004, −0.0221 in 2005–2014 and −0.0170 since 2015. The sign held in every decade, but the effect got weaker each time. That is why the studies report IC by decade, by sector and among S&P 500 members, not only as one average.

References

Grinold, R. C. (1989). The Fundamental Law of Active Management. Journal of Portfolio Management, 15(3), 30–37.
Grinold, R. C. & Kahn, R. N. (2000). Active Portfolio Management, 2nd ed. McGraw-Hill.
Harvey, C. R., Liu, Y. & Zhu, H. (2016). …and the Cross-Section of Expected Returns. Review of Financial Studies, 29(1), 5–68.

Educational content about research methods. Not investment advice.