Chapters in Sentiment, News & AI Research
Chapter 5 of 5 · 5 min read
AI Signals: a next-day model and its track record
What the model predicts, how it is scored and why the published record matters more than any single call.
AI Signals estimates the probable direction for the next trading day using price-based features. What makes it credible is not a clever model, but a published, time-stamped track record you can check.
How to judge a signal model
- Hit rate and payoff over many calls, not a few.
- Out-of-sample results, not just in-sample fits.
- Performance across market regimes.
- Transaction cost and slippage assumptions.
A 55% hit rate can be very valuable or worthless depending on payoff and costs.
Key takeaways
- Judge a model by its track record, not its story.
- Edge is small; costs can erase it.
- Signals are information, not instructions.
Put it into practice
Open the live page on Market Intelligence and see this concept on real data.
Keep learning
- Backtesting: testing ideas on history · Technical Analysis & Trading Tools
Education only, not investment advice. Updated 2026-10-09.