A first example: returns and a moving average

Two of the most common transformations in all of quant work are returns and a moving average. Here is a compact, correct example on a clean OHLCV DataFrame.

Computing returns

A simple (percentage) return is today's close divided by yesterday's, minus one. Log returns are often preferred because they add cleanly across time.

import numpy as np

df["ret"] = df["close"].pct_change()
df["log_ret"] = np.log(df["close"]).diff()

The first row of each is NaN — there is no prior close to compare against. Always drop or handle that first row before summing or modelling.

A simple moving average

A moving average smooths price to reveal trend. The rolling method makes this trivial:

df["sma_20"] = df["close"].rolling(window=20).mean()
df["sma_50"] = df["close"].rolling(window=50).mean()

The first 19 values of sma_20 are NaN because the window is not yet full — that is correct behaviour, not a bug.

Why this matters for the rest of the track

These two lines already contain a trading idea: when sma_20 crosses above sma_50, momentum is turning up. We will turn exactly that into a backtest in the next chapter.

One warning baked in early

Notice that rolling and pct_change only ever look backwards. The moment a calculation uses a future value — even accidentally — your backtest becomes fiction. We make that pitfall explicit next.