Signal Decay, Capacity, and the Half-Life of Alpha

Every alpha signal has a characteristic time horizon over which it works: momentum signals are profitable at 1-month to 12-month horizons but lose edge (or even reverse) at 1-day or 3-year horizons. A signal calibrated to the wrong horizon either turns over too fast (trading costs consume the alpha) or holds too long (the signal edge expires and positions drift into loss). The IC decay curve — how fast the rank correlation between signal and realised return falls with horizon — is the empirical fingerprint of a signal's time horizon.

Beyond horizon lies capacity. As a strategy scales up, it moves prices against itself (market impact), reducing alpha at the margin. The maximum capital deployable before impact costs exceed alpha is the strategy's capacity. These two limits — time and size — define the operational boundaries of any stat-arb approach.

• plot and interpret an IC decay curve;
• estimate the signal half-life from the lag-1 IC autocorrelation;
• set position holding periods and turnover targets from the half-life;
• explain the square-root impact law and use it to estimate capacity;
• describe alpha decay from crowding and identify the quant-quake as a crowding event.

1. 1. The IC decay curve
2. 1. Turnover and the holding-period trade-off
3. 1. Capacity and market impact
4. 1. Alpha decay from crowding