Methodology
About these notes
We publish when we have something specific to say about no-code backtesting, version traceability, or systematic research workflow — not on a fixed schedule.
How curve fitting makes a backtest look great in-sample yet fail live, the warning signs to watch for, and how version traceability makes over-optimization visible.
The silent biases that make a backtest look better than reality, with crypto examples and a short audit checklist.
How to split crypto history into in-sample and out-of-sample periods, why a strategy that only shines in-sample is probably overfit, and how to run the split in Traseq.
A high win rate feels reassuring, but a 90%-win strategy can still lose if the rare losses are large. Here is how to read win rate alongside expectancy and profit factor.