pablodiegoo/time-series-analysis

Time-Series Analysis Skill

This skill provides a robust framework for validating any logic applied to time-series data (financial, operational, or behavioral). It focuses on avoiding look-ahead bias and ensuring statistical robustness through walk-forward validation.

Capabilities

1. Strategy Validation Engine (time_series_backtest_engine)

Vectorized engine to simulate decisions over time.

• Supports custom entry/exit logic.
• Implements time-based and threshold-based stops.
• Handles massive datasets via streaming output.

2. Resource Allocation Simulator (resource_allocation_simulator)

Event-driven simulator to test how resources (capital, inventory, etc.) are distributed across multiple signals.

• Fixed-percentage or dynamic allocation.
• Concentration limits per signal.
• Realistic constraint modeling (leverage caps, minimum allocations).

3. Performance Metrics

Calculates standard risk-adjusted return metrics:

• Sharpe Ratio / Sortino Ratio.
• Maximum Drawdown (MDD).
• Win Rate and Expectancy.

Usage

from scripts.time_series_backtest_engine import run_backtest
from scripts.resource_allocation_simulator import simulate_allocation

# 1. Validate a strategy over historical data
results = run_backtest(
    data=df, 
    entry_threshold=2.0, 
    exit_threshold=0.0,
    stop_loss=3.0
)

# 2. Simulate resource distribution based on signals
sim_report = simulate_allocation(
    trades=results,
    initial_resource=100000,
    max_per_signal=0.1
)

Best Practices

• Walk-Forward: Always separate the formation (training) window from the trading (test) window.
• Look-ahead Bias: Ensure signals at time t only use data available at t-1.
• DPI: Use performance_curve_builder for high-resolution result visualization.

Detailed References

• Methodology: See time_series_validation_methodology.md for backtest logic and spread calculation.
• Metrics: See financial_metrics_pairs_trading.md for financial-specific indicators (Half-Life, Sharpe).
• Validation: See walk_forward_validation.md for gold-standard validation patterns.

Dependencies

pandas, numpy, scipy.