Funding professionals are conversant in this downside. A backtest might present a sample. An element might seem vital. A mannequin might determine construction in returns. But as soon as turnover, buying and selling prices, slippage, market impression, regime shifts and operational constraints are included, the obvious alternative might disappear.
AI doesn’t eradicate this downside. In some instances, it may possibly intensify it.
A reinforcement studying agent is skilled to behave. If it detects weak indicators, it could attempt to exploit them via frequent place adjustments. However when these indicators are near noise, buying and selling exercise can change into pricey. The agent might overfit short-lived fluctuations, interact in extreme buying and selling and erode returns via prices.
Alternatively, the agent might study the alternative lesson. If the atmosphere is noisy and transaction prices are significant, probably the most rational coverage could also be to do nothing. On this case, the AI system doesn’t change into a superior dealer. It turns into inactive.
Each outcomes are vital. They present that failure shouldn’t be random. It’s informative.
If an AI agent can’t rework a detectable sample into optimistic web efficiency, which will inform us one thing in regards to the market atmosphere. The sign could also be too weak. The price construction could also be too excessive. The commentary house could also be inadequate. The mannequin could also be too constrained. Or the sample could also be statistically actual however economically untradable.
That is the essence of the Learnability Threshold.
