docs(book): Complete comprehensive Chapter 8 - Time Series Analysis

Complete pedagogical rewrite of Chapter 8 with production-ready OVSM code
for algorithmic trading time series analysis.

## Contents (~23,000 words substantive content)

**Opening:** LTCM case study ($4.6B collapse from broken stationarity)

**Section 1:** Why Prices Aren't Random Walks
- Empirical evidence (SPY autocorrelation, volatility clustering, fat tails)
- Three exploitable patterns: momentum, mean reversion, volatility clustering

**Section 2:** Stationarity - The Foundation
- Intuitive explanations (dog on leash analogy)
- Complete ADF/KPSS test coverage
- 225-line OVSM implementation with detailed comments
- Worked example: SPY 2020 prices → returns
- Transformation strategies

**Section 3:** ARIMA Models - Capturing Temporal Patterns
- AR models (Bitcoin hourly returns example, φ = 0.142)
- MA models (shock persistence, bid-ask bounce)
- Box-Jenkins methodology (3-step process)
- Ethereum daily returns worked example (ARIMA selection via BIC)
- Production-ready AR(p) implementation (175 lines OVSM)

**Section 4:** Cointegration - Foundation of Pairs Trading
- Economic intuition (KO vs PEP, ETH vs BTC)
- Engle-Granger two-step method
- Complete ETH/BTC worked example (β = 0.062, half-life = 3.4 days)
- Full backtest with transaction cost reality check
- 180-line OVSM implementation with trading signals
- Error Correction Models (ECM)
- Rolling window stability analysis

**Section 5:** Kalman Filters - Dynamic Parameters (Brief)
- Time-varying hedge ratio problem
- State-space intuition
- 18-25% Sharpe improvement demonstrated

**Summary:**
- Three Questions Framework (stationarity, strength, stability)
- Common pitfalls checklist (6 major mistakes)
- When time series works/fails
- Production deployment checklist (4 categories, 20 items)
- Complete OVSM function reference

## Key Features

1. **Pedagogical Approach:**
   - 60% explanation, 40% code (as requested)
   - Real-world disasters/motivations before theory
   - Step-by-step worked examples with real data
   - Every code line explained (WHAT/WHY/HOW pattern)
   - Trading implications throughout

2. **Production-Ready Code:**
   - adf-test (125 lines with full documentation)
   - ar-model (175 lines with AIC/BIC selection)
   - engle-granger-test (180 lines with trading signals)
   - All functions include interpretation and error handling

3. **Practical Focus:**
   - Transaction cost reality checks on all strategies
   - BTC hourly momentum: statistically significant but unprofitable after costs
   - ETH/BTC pairs trading: $823 gross → $55 net after fees
   - Kalman filter benefits quantified: +18% Sharpe

4. **Risk Management:**
   - LTCM opening sets cautionary tone
   - Rolling window analysis for all strategies
   - Parameter stability requirements
   - Production deployment checklist

## References

10 academic citations including Engle-Granger (1987), Box-Jenkins,
MacKinnon cointegration critical values, LTCM case study.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Author: 0xrinegade