Data-Driven Investment Strategies

import pandas as pd, numpy as np

returns = prices.pct_change().dropna()
vol_21d = returns.rolling(21).std() * np.sqrt(252)
# 21 trading days ~= 1 month; annualise by sqrt(252)

We use Python for day-to-day research (pandas, NumPy, scikit-learn, statsmodels) and R for deeper statistical work (time-series, regression diagnostics).

We run hypothesis tests on return predictability, build factor models, and test assumptions like normality, stationarity, and autocorrelation before acting on any signal.

Random forests, gradient boosting, and simple neural networks for factor prediction. Heavy emphasis on walk-forward validation to avoid look-ahead bias.

We run parametric VaR, historical simulation, and Monte Carlo. CVaR captures tail risk that VaR ignores — both reports feed into our quarterly risk review.

We implement mean-variance optimisation, Black-Litterman views, and Hierarchical Risk Parity — and we backtest each approach against simpler benchmarks like equal-weight.

Foundations for everything from volatility forecasting to option pricing. We rarely implement Black-Scholes directly, but understanding GBM is prerequisite to serious quant work.