This is a practitioner’s engineering journal on building Machine B—a machine-learned trailing stop engine designed to make exits systematic, consistent, and adaptive.
Thesis: If trailing stops are systematic, consistent, and adaptive, the equity curve will take care of itself.
The Core Idea
Most traders obsess over entries.
Most trading systems succeed or fail because of exits.
This series documents how I approached trailing stops as a risk engineering problem, including the mistakes and data traps that don’t show up in most ML trading tutorials.
Series posts
Part 0 — I Had Rules for Entries and Feelings for Exits
Part 1 — Why Trailing Stops Are Harder Than Entries
Part 2 — Random Forests for Trailing Stops: Labels Without Lookahead Bias
Part 3 — The Futures Rollover Trap: Why Your ML Model Breaks Every Quarter
Part 4 — RTH vs ETH: The Data Distribution Mismatch That Broke My Stop Model
Part 5 — Building Machine B: From Research Model to Production System
Machine A and Machine B
Machine A decides what to trade (signals, regimes, entries).
Machine B decides how to manage risk after entry (trailing stops, exits).
Entries predict direction.
Exits shape the distribution.
If you’re building systematic trading systems, this is the blueprint I wish I had when I started.