Python and data for quants

6 lessons · 4 equations. Each lesson below gives its formulas and key rules; open the lesson for the full explanation.

Idiomatic Python: generators, itertools and the gotchas

Key rules

Cross-sectional standardisation

z_{it}=\frac{x_{it}-\bar x_t}{s_t}

Compute the mean and spread across assets at each timestamp; broadcasting must align on that axis.

Remember

Join row count by key

N_{\mathrm{joined}}=\sum_k n_{k,\mathrm{left}}n_{k,\mathrm{right}}

Duplicate keys multiply rows. Check join cardinality before trusting a backtest or aggregate.

Remember

Key rules

Shift the signal, not the return, and lag by the publication delay rather than by one bar.
merge_asof with direction="backward" is the "what did I know then" join.
Survivorship, restatement and selection biases sit alongside look-ahead — name all four.

Annualised Sharpe after costs

\mathrm{SR}_{\mathrm{ann}}=\frac{\bar r_{\mathrm{net}}}{s_{\mathrm{net}}}\sqrt{252}

Use net daily returns after spread, fees and turnover costs; the square-root rule assumes weak serial dependence.

Remember

Amdahl’s law

\mathrm{speedup}=\frac{1}{(1-p)+p/s}

If only fraction $p$ of runtime is improved by a factor $s$ , the untouched work caps the total speedup.

Remember