Monte Carlo Stock Simulator

Probabilistic stock price forecasting using Geometric Brownian Motion. 90% validated accuracy on historical backtests.

Live Demo | Documentation

---

Results

Metric	Value
Validation Accuracy	90.0% (30+ backtests)
Performance	1000 simulations in <1 sec
Statistical Significance	p-value 0.67 (within expected range)

---

Features

• Monte Carlo Simulation — Normal and Student-t (fat-tailed) distributions
• Risk Metrics — VaR (95%, 99%), Sharpe Ratio, confidence intervals
• Validation Dashboard — Automated backtesting with regime analysis
• Model Comparison — Side-by-side Normal vs Student-t performance

---

Quick Start

git clone https://github.com/XCODESSS/Monte-Carlo-Simulation
cd Monte-Carlo-Simulation
pip install -r requirements.txt
streamlit run app.py

---

How It Works

Model: Geometric Brownian Motion S(t+1) = S(t) × exp((μ - 0.5σ²)Δt + σ√Δt × ε)

• μ, σ estimated from historical log returns
• ε ~ Normal(0,1) or Student-t(df) for fat tails
• 90% confidence intervals from simulation percentiles (5th, 95th)

Why Student-t? Real markets have fatter tails than Normal distribution predicts. Student-t captures extreme events (crashes, rallies) more accurately.

---

Assumptions

1. Constant volatility — Model uses historical σ, but real volatility changes over time
2. Log-normal returns — Assumes returns follow GBM; ignores jumps, mean reversion
3. No fundamental factors — Doesn't consider earnings, P/E, macroeconomic data
4. Historical parameters — Future may not resemble past

---

Where It Fails

Condition	Behavior
High volatility regimes	Hit rate drops ~5-10%
Regime changes (bull→bear)	Model lags behind transitions
Black swan events	Even Student-t underestimates extreme tails
Long forecast horizons (>3 months)	Intervals become very wide (50%+)

Directional accuracy: ~50% — The model quantifies uncertainty well but doesn't predict direction better than chance.

---

When To Use This

Good for:

• Understanding price uncertainty ranges
• Risk assessment (VaR, downside scenarios)
• Comparing volatility across assets
• Educational purposes

Not good for:

• Trading signals
• Predicting direction
• Long-term forecasts (>3 months)

---

Tech Stack

Python · NumPy · Pandas · SciPy · Streamlit · yfinance

---

Contact

Shreyarth — LinkedIn · GitHub

Seeking Data Science / Financial Analysis internships for Summer 2026.

⭐ If this helped you understand Monte Carlo simulations, please star the repo!

Built with: Precision · Validated with Data · Ready for Production