Projects

This repository contains small quantitative experiments focused on decision theory, probability, and incentives.

The main work is inside the Quant Projects folder.

---

Quant Projects

This folder contains self-contained simulations and analyses related to betting strategies, probabilistic forecasting, and information value.

Kelly Betting Simulation (Kelly Betting Simulation.py)

Simulates repeated betting using the Kelly criterion to study long-run capital growth.

Core ideas:

• Optimal bet sizing under known probabilities
• Tradeoff between growth rate and volatility
• Risk of ruin under different leverage levels

What it demonstrates:

• Monte Carlo simulation
• Log-utility maximization
• Sensitivity of outcomes to probability estimates

This script is useful for understanding why Kelly betting is theoretically optimal but practically dangerous when probabilities are misestimated.

---

Probability Calibration (Probability Calibration.py)

Evaluates how well predicted probabilities match observed outcomes.

Core ideas:

• Calibration vs accuracy
• Overconfidence and underconfidence in probabilistic models
• Reliability of forecasts

What it demonstrates:

• Binning predicted probabilities
• Comparing predicted vs empirical frequencies
• Basic diagnostics for probabilistic predictions

This is directly relevant to forecasting, risk modeling, and any system that outputs probabilities rather than point estimates.

---

Scoring Rules and Incentives (Scoring rules incentives.r)

Explores how different scoring rules incentivize honest probability reporting.

Core ideas:

• Proper vs improper scoring rules
• Incentive compatibility
• Why some metrics encourage truthfulness and others do not

What it demonstrates:

• Expected score analysis
• Comparison of scoring rules under different reported probabilities
• Use of R for statistical reasoning and visualization

This project connects decision theory with practical evaluation metrics used in forecasting and machine learning.

---

Value of Information (Value of Information.py)

Simulates how additional information changes expected utility and decision quality.

Core ideas:

• Expected value of perfect and imperfect information
• When more data is actually worth paying for
• Diminishing returns of information

What it demonstrates:

• Decision making under uncertainty
• Utility-based comparisons
• Simulation of information updates

This is relevant to domains like trading, experimentation, and data acquisition strategy.

---

Requirements

Python:

• Python 3.8+
• pandas
• matplotlib

R (for the scoring rules analysis):

• R base installation

---

Setup and Usage

Clone the repository:

git clone https://github.com/hemv-857/Projects.git
cd Projects

Create a virtual environment and install Python dependencies:

python -m venv venv
source venv/bin/activate        # Linux/macOS
.\venv\Scripts\Activate         # Windows PowerShell
pip install numpy matplotlib

Run any Python script:

cd "Quant Projects"
python "Kelly Betting Simulation.py"

Run the R script from R or RStudio:

source("Scoring rules incentives.r")

---

Scope and Intent

This repository is intentionally small and focused.

It exists to:

• Explore quantitative decision-making concepts
• Test ideas with simulation
• Demonstrate reasoning under uncertainty

It does not aim to:

• Be a reusable library
• Provide optimized or production-ready code
• Hide assumptions behind abstractions

---

License

No license is currently specified.
Without a license, reuse is not permitted by default. Add one if reuse is intended.

---

Clear reasoning beats large codebases. Every file here exists to answer a specific question, not to pad a portfolio.