Writing
On Structural Reasoning
What a Mechanism Owes You That an Artifact Never Will
Every instrument you trust to validate an edge reads the same thing: the returns. Deflated Sharpe, PBO, CPCV, walk-forward. Different tools, one input.
Deploying MCMC Outside of Bayesian Frameworks
The protection against overfitting is procedural. The sampler is allowed to propose relationships, but it is not allowed to justify them.
You Can't Backtest Your Way to a Reason.
Most models are discovered inside datasets. The structural ones are derived before the dataset is ever opened.
You Didn't Overfit. You Just Never Understood Why The Pattern Was There.
Most quantitative researchers are using machine learning the same way an eight year old uses laser eyes. Impressive until nobody tells him when to stop.