Markets may look chaotic, but under the intrinsic noise and unpredictability, some patterns quietly repeat. These seasonal effects — recurring behaviors tied to the calendar — can create subtle yet measurable edges for quantitative traders.

When you’re running a backtest, every line of code depends on a silent assumption: "The data actually reflects what happened in the market". But that assumption can break down — not because of your model, but because of something much smaller and easier to overlook: price adjustments.

In quantitative trading, like in physics, every model starts with data — and often, that means historical prices. But not all price data is created equal. One of the most overlooked — and misunderstood — aspects of market data is the distinction between adjusted and unadjusted stock prices. It might seem like a small technical detail, but it’s often the difference between a strategy that looks profitable —and one that truly is making money. In this post, I’ll unpack why that distinction matters.

When I first started backtesting strategies, I thought one long backtest was enough. It wasn’t. Markets change, conditions shift, and what worked in one year might fail in the next. That realization led me to walk-forward analysis — a systematic way to simulate how a strategy learns and adapts over time. In this post, I’ll show you how to generate the date sequences that make this method possible using Python.