Every Algo Has a Decay Rate

Here is the thing almost nobody selling an algorithm will tell you plainly: every edge decays. Not some of them. Not the bad ones. All of them, including the ones that are genuinely good right now. Decay is not a defect in a strategy — it is a property of markets. Once you accept that, a lot of the noise in this industry becomes very easy to see through.

Why an edge decays

An edge exists because some inefficiency exists — a behaviour that repeats, a structural quirk, a group of participants doing something predictable. The moment enough capital notices the same thing, the inefficiency gets competed away. That is not a conspiracy; it is just what markets do. They adapt.

Regimes also change. A strategy calibrated on a trending, low-volatility stretch is being fed different market physics when volatility doubles. It has not broken. It is simply no longer operating in the conditions that made it work.

And some "edges" never existed at all. They were overfitting — a set of parameters tuned so precisely to historical data that they described the past rather than explaining it. Those do not decay. They were never alive.

Decay is measurable, and that is the point

This is the part that turns a vague warning into a usable tool. You can measure decay. Hold back data the strategy was never tuned on and test on it. Compare performance in the tuned window against the untouched window. Run the thing forward on live data and watch whether the live distribution resembles the backtest distribution or quietly drifts from it.

A strategy that performs beautifully in-sample and mediocrely out-of-sample is telling you something specific: much of what you measured was fit, not edge. That is not a reason to despair. It is a reason to size accordingly and keep measuring.

Now apply that to a bot for sale

Sit with the economics for a moment. If someone possesses a genuinely durable, high-performing automated strategy, the returns from trading it compound privately. Selling copies of it does two things: it generates revenue that does not depend on the strategy working, and it accelerates the decay of the edge by putting more capital into the same inefficiency.

So when a strategy is offered for sale, ask the honest question — not cynically, just clearly. What is the seller's actual business? If the answer is "selling the strategy" rather than "trading the strategy," you have learned something important about which one they believe in.

There is a further problem. A closed system cannot be inspected. You cannot see the entry logic, so you cannot know which market conditions it depends on. You cannot see the exit logic, so you cannot know how it behaves in a regime it has never met. When it starts underperforming — and it will, because everything does — you have no way to tell whether it is normal variance, a regime it will recover from, or terminal decay. You are left guessing about a black box with your capital inside it.

The alternative is not "never automate"

Automation is genuinely useful. Rules executed by a machine do not get scared, revenge-trade, or move a stop because the candle looked frightening. That is a real advantage over a tired human at 3pm.

The alternative to a black box is a system you can see inside. When you understand why a strategy works, you can recognise the conditions it needs, notice when those conditions leave, and decide deliberately rather than hoping. You can measure its decay instead of being surprised by it. We work through what makes an edge real and explainable in how to find your trading edge and turning a trading edge into a repeatable system.

This applies to every tool, including ours and including yours. A strategy we build decays exactly like any other. The difference discipline makes is not exemption from decay — it is noticing it early, measuring it honestly, and retiring something before it costs you rather than after. That is the whole skill.

Trading carries risk of loss regardless of how a strategy is built or who built it.