Sankar

Markets as a
learning environment

A paper-trading system for NSE stocks combining entropy analysis, representation drift detection, and adaptive decision control.

agent.py
# Initialise RL agent
agent = PPOAgent(
gamma=0.99
adaptive=True
)
status: "LIVE_PAPER"

Problem

Markets are chaotic and non-stationary, yet most trading systems assume stability. This leads to overfitting and failure under changing conditions.

Solution

Arbitrix evaluates not just signals, but whether the system should trust those signals. It integrates entropy (PIEC), drift detection (RLFS), and adaptive control (S-ADR).

System Architecture

DATA

Market Feed

Real-time + historical data

+ expand

ANALYSIS

Signal Engine

TA + entropy modeling

+ expand

CONTROL

Decision Layer

Adaptive execution

+ expand

EXECUTION

Trading System

Paper trading + AI reasoning

+ expand

Why this matters

In uncertain environments, knowing when not to act is more important than acting correctly.

Key Features

Entropy-aware trading signals
Drift-based reliability scoring
Adaptive position sizing
AI-assisted trade explanations

Build Journey

Concept

Explored RL + markets

System

Built simulation + TA engine

PIEC Integration

Added entropy modeling

Refinement

Improved robustness + modularity

Tech Stack

ReactNode.jsMongoDB

Related Systems