1+ week, 4+ day ago   (1023+ words) Every backtest report has a line that says maximum drawdown. Mine said 15.3R. I used that number to size positions for years, and it took me an embarrassingly long time to notice that it is not a risk estimate. It is…...

2+ week, 5+ day ago   (285+ words) For a long time, the natural question in financial machine learning was: which model is better — gradient boosting, LSTM, Transformer, or something else? In 2026, more and more research is pushing toward a different question: can we actually trust the experiment…...

3+ week, 3+ day ago   (966+ words) Every automated system that still has a human with override permissions is not an automated system. It is a hybrid, and the human half never gets profiled. None of that shows up in a trade log. A trade log records…...

3+ week, 4+ day ago   (381+ words) In Part 1 (Linked: Part 1), I walked through the mechanics and math of a custom Retrieval-Augmented Generation (RAG) pipeline, including why high-dimensional vector spaces rely on cosine similarity to measure semantic connections. The takeaway was optimistic: replace rigid keyword matching with…...