Alpha Decay: Why Your Best Signal Gets Weaker Over Time

Here's an uncomfortable truth: every trading signal you have is decaying right now. Not dramatically — not overnight — but continuously, as market participants discover the same pattern, arbitrage the same inefficiency, and push the edge toward zero.

This process is called alpha decay, and it's the fundamental reason why "set it and forget it" trading systems fail. Your edge has a half-life, and you need to measure it, track it, and replace it before it hits zero.

The Mechanics of Alpha Decay

An alpha signal exists because of an information asymmetry or a behavioral pattern that hasn't been fully priced in. Decay happens through three channels:

1. Arbitrage: Other traders discover the pattern and trade it, pushing the price to fair value. The mispricing closes. The signal dies.
2. Adaptation: Market participants change their behavior in response to the pattern. If everyone starts buying dips at RSI 30, the dip never reaches RSI 30 anymore.
3. Structural change: Market microstructure, regulation, or technology changes the conditions that made the signal work. HFT changed momentum. Options market-making changed mean reversion.

Measuring the Decay Rate

The standard approach: track signal performance in rolling windows.

Typical Half-Lives

• High-frequency statistical arbitrage: Days to weeks. These are the first to decay as HFT firms replicate them.
• Short-term momentum (1–5 day): 3–6 months. Decays as quantitative funds add similar signals.
• Medium-term momentum (21–126 day): 1–3 years. Slower decay because it's harder to arbitrage at this timescale.
• Value/fundamental signals: 5–10+ years. The slowest to decay because they rely on behavioral patterns (overreaction to earnings) that are deeply ingrained.
• Macro/regime signals: Essentially permanent. Central bank behavior and business cycles don't get arbitraged away.

The Rotation System

Professional quant funds don't have one signal — they have a portfolio of signals at different points in their decay curves. At any given time:

• ~30% of active signals are "fresh" (discovered in the last 6 months, full edge)
• ~40% are "mature" (6–18 months old, 60–80% of original edge)
• ~20% are "decaying" (18+ months, <40% of original edge)
• ~10% are "zombies" (detected as dead, being retired)

The system continuously generates new signals, monitors existing ones, and retires the dead ones. The overall edge of the portfolio stays stable even as individual signals decay.

How to Detect That YOUR Signal Is Decaying

If you're running a personal trading system, here's the minimum monitoring:

1. Monthly hit-rate: What % of your signals are winning? If it's been dropping for 3 consecutive months, the edge is decaying.
2. Monthly average R: What's the average profit/loss in R-multiples? A steady decline is the clearest decay signal.
3. Drawdown frequency: Are you getting stopped out more often? That's the market adapting to your entry pattern.

GemStox's Approach to Decay

GemStox's 10-class cross-validation system is inherently a decay-detection mechanism. If a signal that passed all 10 classes last month now only passes 6, that's a decay alert — the strategies that used to agree are starting to disagree because the underlying edge is weakening.

The system automatically downgrades or retires signals whose cross-validation score drops below threshold. You never trade a zombie signal.