Adding More Tests Doesn't Make Your Backtest Safer

Half your validation layers buy you nothing. The ones that work double the edge you need.

Aug 21 • Lucas

4

June 2026

Max Drawdown Is a Distribution, Not a Number

Why the worst loss in your backtest tells you almost nothing about the worst loss you should plan for, and how a block bootstrap makes the gap visible.

Jun 12 • Lucas

25

2

6

May 2026

ML trading or data mining?

The line is thinner than you think

May 29 • Lucas

7

Live Crypto Data in 15 Lines of Python

No account, no API key, no cost. Just an open socket and Python.

May 22 • Lucas

10

1

90% of retail quants don't have a database. Just files.

The fix costs €0. The bug costs you months of research.

May 15 • Lucas

10

The Clip Effect: Is Your Backtest Real, or Just Lucky?

A 60-second robustness check that kills most backtests.

May 8 • Lucas

18

1

Building a Clean Research Environment with Oryon

How to combine features and targets in one dataframe before doing exploratory data analysis.

May 1 • Lucas

5

April 2026

Standardizing Features Without Breaking Your Pipeline

A More Coherent Way to Standardize Features in Quantitative Pipelines

Apr 24 • Lucas

7

Build once. Run live.

A unified approach to feature pipelines across research and production

Apr 17 • Lucas

5

From Quantreo to Oryon

Latency does not start at execution. It starts upstream, in the feature pipeline.

Apr 10 • Lucas

4

Why intra-bar features matter

How to enrich higher-timeframe bars with lower-timeframe information without changing your full research workflow.

Apr 3 • Lucas

9

March 2026

Why I always try to simplify my alpha formulas

Simpler formulas will miss some cases. But in many situations, that is exactly what makes them more robust.

Mar 27 • Lucas

9