Financial Data Science¶
Applying data science methods to financial markets. Covers portfolio theory, risk metrics, derivatives basics, and quantitative analysis patterns.
Portfolio Theory¶
Key Metrics¶
- Expected return: weighted average of asset returns
- Portfolio volatility: sqrt(w^T * Cov * w) where w = weight vector
- Sharpe ratio: (return - risk_free_rate) / volatility
- Beta: sensitivity to market movements
Efficient Frontier¶
Set of portfolios maximizing return for given risk level. Found via Monte Carlo simulation or quadratic programming.
# Monte Carlo portfolio optimization
port_return = weights @ daily_returns.mean() * 252
port_vol = np.sqrt(weights @ cov_matrix @ weights) * np.sqrt(252)
sharpe = port_return / port_vol
Financial Derivatives¶
Options¶
- Call: right to buy at strike price. Payoff = max(0, spot - strike)
- Put: right to sell at strike price. Payoff = max(0, strike - spot)
- Premium: price paid for the option
Forwards/Futures¶
Lock price for future delivery. Mark-to-market: daily P&L based on spot vs contract price.
Risk Metrics¶
Value at Risk (VaR)¶
Maximum expected loss at confidence level over time period.
# Historical VaR (95%)
var_95 = np.percentile(returns, 5)
# Parametric VaR
var_95 = returns.mean() - 1.645 * returns.std()
Volatility¶
Standard deviation of returns. THE fundamental risk measure in finance.
68-95-99.7 rule applied to returns: 95% of daily returns within +/- 2*sigma.
Financial Constants¶
- Trading days per year: 252
- Trading days per month: ~20
- Trading days per quarter: ~60
- Annualization factor: sqrt(252) for volatility
Asset-Backed Securities (ABS)¶
Securitization: convert illiquid assets (mortgages, loans) into tradable bonds. - SPV (Special Purpose Vehicle): legally separate entity holding asset pool - Tranching: redistribute cash flows into bonds with different risk/return profiles - Senior tranche: first to be paid, lowest risk (AAA rating) - Equity tranche: residual after all other tranches paid, highest risk/return
Key Financial Ratios¶
| Category | Metrics |
|---|---|
| Profitability | EBIT, EBITDA, net margin |
| Liquidity | Current ratio, quick ratio |
| Solvency | Debt/equity, interest coverage |
| Valuation | P/E, EV/EBITDA, P/B |
| Cash flow | FCF, operating cash flow |
Free Cash Flow (FCF): most important metric for cash generation capacity. Accounts for working capital changes and CapEx.
Gotchas¶
- Financial returns are NOT normally distributed - heavy tails (kurtosis > 0)
- Past performance does not predict future returns
- Correlation between assets changes during crises (correlation goes to 1)
- VaR underestimates tail risk - use CVaR (expected shortfall) for extreme scenarios
- Always use log returns for multi-period analysis, simple returns for single period
See Also¶
- monte carlo simulation - simulation techniques for finance
- probability distributions - distribution theory
- time series analysis - financial time series
- descriptive statistics - analyzing return distributions