Where markets meet machine learning
Research notes, model explainers and market analysis from the team building TradePredictor. No hype, no signals-for-sale — just how the maths actually works.
Trading Psychology: The Biases That Cost Money
Most trading mistakes are not analytical failures. They are predictable cognitive patterns that appear reliably under financial pressure.
Market Analysis
Seasonality in Markets: Signal or Noise?
Calendar patterns are the easiest thing in finance to discover by accident. Some survive scrutiny; most are the product of testin…
Market Analysis
Reading Volume: What It Tells You and What It Doesn't
Volume is the most available secondary data most traders have and among the most misinterpreted. It confirms; it does not predict.
Market Analysis
How Central Bank Decisions Move Forex Pairs
Currency markets do not trade the decision. They trade the difference between the decision and what was already expected.
Market Analysis
What Actually Moves Bitcoin
Beyond the halving narrative, bitcoin's price has behaved increasingly like a high-beta risk asset. The data supports that more t…
AI & Machine Learning
Overfitting in Trading Models: How to Spot It
Overfitting rarely announces itself. It looks exactly like success, right up to the moment you deploy.
Market Analysis
Gold and Real Yields: The Relationship That Actually Matters
Gold pays no income. That single fact explains most of its price behaviour, and it runs through real interest rates rather than i…
AI & Machine Learning
Walk-Forward Validation vs Train-Test Split
The standard machine learning split assumes your rows are interchangeable. Time series rows are not, and that assumption quietly …
AI & Machine Learning
Confidence Scores in AI Predictions Explained
A prediction without a confidence estimate is nearly useless. But most confidence numbers are not probabilities, and treating the…
AI & Machine Learning
Why Machine Learning Models Fail in Live Markets
The gap between backtest and live performance has predictable causes. Most of them are known before deployment and ignored anyway.