How I turn a paper into a tradeable system

StatsEdgeTrading

Just like every normal person, I spend my spare time reading academic studies about markets and indicators.

This week’s rabbit hole: a mean reversion paper using the True Strength Index (TSI). I’d never really touched TSI before, which is exactly why it caught my eye. New indicator, clean premise, and (on their charts) a system that looked competitive in up years while doing a noticeably better job during drawdown regimes.

But here’s the part that matters: the paper is not the product. The process is the product.

My workflow looks like this:

1. Read the study myself
   Yes, AI can summarize anything. I still read the methodology first. You can make any backtest look great if you’re creative enough. I’m looking for “does this make sense” and “is there any funny business” before I ever touch code.

2. Use an LLM to accelerate understanding, not outsource it
   Once I understand the paper, I’ll drop it into a tool like NotebookLM and have it walk me through the rules, assumptions, and signal logic. I’ll even listen to it on a dog walk. If the summary doesn’t match what I just read, that’s a red flag.

3. Rebuild it in my environment and over-test it on purpose
   Next step is my real test: I recreate the logic on my own data (25 years) and I start with an intentionally huge capacity test. I’m talking thousands of potential positions, because I want to see if the “edge idea” survives reality at scale before I start babying it with optimization.

4. Translate the idea to the instrument I actually trade
   The paper focused on indices. I care about individual stocks. So I move it to weekly charts, add basic guardrails (stop loss, profit target, time-based logic), and see if it behaves like a mean reversion system should: frequent exits, controlled pain, and no catastrophic bleed in ugly markets.

5. Stress test the years that matter
   I don’t care if a strategy looks cute in a straight-up bull market. I care what it did in 2008, 2020, and the 2022 to 2023 bear. Mean reversion’s whole risk is obvious: you’re buying things that are selling off. So we check the periods where that hurts.

If it passes the smell test, it goes into my library of “columns” for future optimization and strategy building. If it doesn’t, it goes into the trash where most ideas belong.

That’s the job behind the scenes at StatsEdgeTrading. The trading is on rails. I wait for the Discord alert bot to yell at me. The work is building better systems.

If you want the outputs of this process (and the free newsletter and courses), head to www.statsedgetrading.com.