Information Coefficient (IC)
Computes the information coefficient (IC): the correlation between the returns (or rankings) predicted by a model and those actually realized. It's the measure of predictive skill at the heart of the fundamental law of active management — an IC of zero means worthless forecasts, and the closer to one, the better the signal anticipates the future. In practice, real ICs tend to be low, around 0.05 to 0.15. Enter the lists of predicted and realized values, in the same order.
Resultado
—
Information Coefficient (IC)
Computes the information coefficient (IC): the correlation between the returns (or rankings) predicted by a model and those actually realized. It's the measure of predictive skill at the heart of the fundamental law of active management — an IC of zero means worthless forecasts, and the closer to one, the better the signal anticipates the future. In practice, real ICs tend to be low, around 0.05 to 0.15. Enter the lists of predicted and realized values, in the same order.
The cold measure of forecasting skill
Every active manager believes they can predict which assets will rise. The information coefficient puts that belief to a direct test: it correlates the forecasts with what actually happened. An IC of zero means the forecasts have no relation to reality; the closer to one, the better the model sees the future.
The surprise for many is how low real ICs are. Even excellent managers rarely exceed 0.1, which seems pitiful until you remember that getting just a little more than half right, consistently and across many bets, is enough to make money. It's the logic of the fundamental law of active management: small skill, multiplied by breadth, becomes return.
Enter the list of predicted values and that of realized values, in the same order. The tool returns the Pearson correlation between the two, which is the IC. Beware of short series: few points can produce a high IC by chance, so confidence in the number grows with sample size.
Related Tools
Modified Dietz Return
Computes a portfolio's return with the Modified Dietz method, which adjusts the result for deposits and withdrawals made mid-period. Instead of ignoring the cash that came in and out, it weights each flow by the time it stayed invested: R = (ending value − beginning value − flow)/(beginning value + flow·weight). It's an approximation of the time-weighted return widely used by managers before the era of daily calculation. Enter the beginning value, the ending value, the net flow and the flow's weight in the period.
Tracking Error
Computes a portfolio's tracking error: the standard deviation of the differences between the portfolio's returns and the benchmark's, period by period. It measures how much the portfolio diverges from its reference index — an index fund aims for tracking error near zero, while an active fund has a higher one, reflecting its bets. It's the denominator of the information ratio. Enter the lists of portfolio and benchmark returns, in the same order, separated by commas.
Parametric Expected Shortfall (CVaR)
Computes the Expected Shortfall (ES), also known as CVaR, by the parametric Gaussian method. While VaR answers what the minimum loss is in the worst cases, ES goes further and answers what the average loss is when the worst happens, summing what lies in the tail beyond the VaR. The formula is −μ + σ·φ(Φ⁻¹(c))/(1−c), with Φ⁻¹ the inverse normal. It's a coherent risk measure, required under Basel, precisely because it captures the severity of extreme losses. Enter the mean return, the standard deviation of returns and the confidence level.
The results provided by this tool are for general informational and educational purposes only and do not constitute professional, financial, medical, legal, tax or accounting advice. Always confirm important decisions with a qualified professional and official sources.