Time-of-Day Conditioning of Signal Decay
Why the same intraday trading signal can have a completely different useful lifespan depending on whether it fires at the open, midday, or the close, and why strategies condition signal decay on time of day rather than treating decay as constant.
An intraday signal's decay is how quickly its predictive edge fades after it fires — a signal that predicts the next five minutes' return is useless if you act on it ten minutes later. What's easy to miss is that decay isn't a fixed property of the signal alone; it depends heavily on when in the trading day the signal fires, because volume, volatility, and the number of other participants reacting to the same information all vary systematically across the session. A signal firing in the first thirty minutes after the open, when volume is high and many participants are actively repricing overnight news, typically decays much faster than the identical signal firing in the quiet early-afternoon lull, when fewer participants are trading and a mispricing can persist longer simply because fewer people are around to correct it.
Time-of-day conditioning means estimating decay curves separately for different intervals of the trading day — open, mid-morning, lunch, pre-close — rather than fitting one decay curve to the whole session and applying it everywhere. A strategy using a single average decay rate risks holding a position too long when it fires near the close (where a decayed signal offers little edge but full risk) or exiting too early when it fires during a quiet midday stretch where the edge would have persisted longer.
The close deserves special treatment: elevated volume from index rebalancing and closing auctions changes both the signal's decay and the cost of acting on it, which is why many intraday strategies explicitly disable or heavily discount signals in the last few minutes of trading regardless of how strong they look on paper.
Signal decay in intraday trading isn't a fixed number — it varies by time of day because volume, volatility, and participant density vary across the session, so estimating separate decay curves for the open, midday, and close produces materially better sizing and holding-period decisions than assuming one decay rate applies all day.
Further reading
- Kissell, The Science of Algorithmic Trading and Portfolio Management, ch. 3