Blending Fast and Slow Alphas

A short-term reversal signal might predict returns well for the next day or two and then fade to nothing, while a value or quality signal might predict returns weakly but steadily over many months. Trading only the fast signal means constant turnover and high costs; trading only the slow one leaves a source of edge on the table. Alpha horizon blending combines signals with different decay speeds into a single forecast so the portfolio captures both without either one drowning out the other.

The usual approach weights each signal's contribution to the forecast by its own risk-adjusted strength (often its information coefficient) rather than by intuition, and separately manages trading so the fast component turns over quickly while the slow component barely trades at all.

A worked example

Say a reversal signal has an information coefficient of 0.06 but decays to near-zero predictive power within 3 days, while a value signal has an information coefficient of 0.03 but stays useful for 6 months. A blended forecast might weight the reversal signal roughly 2x the value signal in the near term (matching its higher short-run IC), while the value signal dominates the average holding period because it persists far longer — the combined signal captures the reversal's punch and the value tilt's staying power, rather than forcing one uniform trading speed on both.