The IC-Horizon Profile

A single information coefficient — the correlation between a signal's rank and the subsequent return — tells you the signal works at one horizon. It says nothing about whether the edge builds slowly and lingers, or fires immediately and evaporates, and those two shapes call for completely different trading strategies even when the peak IC is identical.

Reading the profile instead of the point

An IC-horizon profile is built by computing the same information coefficient repeatedly, once for each horizon in a range — say, 1 day, 5 days, 10 days, 20 days, 60 days out — and plotting IC against horizon. Three shapes show up often enough to be worth naming.

A signal that peaks early and decays fast (strong IC at 1–2 days, close to zero by 20) is capturing something the market absorbs quickly — likely a short-lived reaction, useful only if the strategy can trade fast and cheap enough to capture it before it's gone. A signal that builds slowly and peaks late (weak at 1 day, strongest at 40–60) is capturing a slower repricing — useful to a strategy with a longer holding period and patience, useless to one that turns over daily. A signal that stays roughly flat across horizons is unusual and worth double-checking; genuine signals almost always decay eventually, and a flat profile out to very long horizons can be a sign the "signal" is actually proxying for a slow-moving risk exposure rather than a decaying information edge.

Profile shape	Likely mechanism	Matching strategy
Peaks early (1–5d), decays fast	Fast-absorbed information, news-like	High turnover, needs cheap execution
Builds slowly, peaks late (40–60d)	Slow structural repricing	Low turnover, patient holding period
Roughly flat across all horizons tested	Possibly a risk exposure, not decaying information	Worth a factor-exposure check before trusting it

Why the shape matters more than the peak

Two signals can both report a peak IC of 0.05 — but one hits that peak on day 2 and is gone by day 10, the other hits it on day 45 and stays elevated through day 90. Sized identically without looking at the shape, the fast signal would be traded far too infrequently to capture its edge, and the slow signal would be traded far too often, paying costs on rebalances that add nothing because the signal hasn't moved. The horizon at which a strategy actually trades has to be matched to where the profile peaks, not chosen independently of it.