Simulation Models Backlog
Candidate simulation approaches for stock market prediction, ordered roughly by implementation complexity and relevance to the current stack.
Prioritized
| Model |
Status |
Notes |
| Monte Carlo |
✅ Done |
10k price path scenarios; output highest win-rate moves as probability distribution on top of RF |
| Bootstrap Resampling |
✅ Done |
Resample historical returns instead of assuming a distribution; more realistic fat tails |
| Regime-Switching |
✅ Done |
Bull/bear/sideways state transitions; integrates with existing GARCH + K-means vol regime |
| VaR (Value at Risk) |
Backlog |
Maximum expected loss at a given confidence level (e.g. 95%); pairs directly with existing GARCH and vol regime work |
| Parkinson Volatility |
Backlog |
Alternative vol estimator using high/low prices instead of close-to-close; easy addition alongside existing realized vol calculations |
On Radar
| Model |
Notes |
| Geometric Brownian Motion (GBM) |
Classic Black-Scholes path model; assumes log-normal returns and constant vol — fast but unrealistic; useful as a baseline comparison |
| Historical Scenario Replay |
Replay known crashes (2008, COVID) against current positions for stress testing |
| Agent-Based Model (ABM) |
Simulate thousands of traders with different strategies; captures herding, panic selling, microstructure effects — high complexity |
| Game Theory Simulation |
Model market as a strategic game between participants — experimental |
| Quantum Monte Carlo |
Used by some hedge funds; experimental and overkill for this stack |