TradeSim AI

TradeSim AI: Slow and Steady Wins the Trade

#devchallenge #hf26challenge #hacktoberfest #mlh

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

TradeSim AI is an AI-powered paper trading and strategy backtesting platform that helps retail traders and finance students experiment with trading strategies using historical market data — without risking real money.

Instead of chasing quick wins, the platform encourages a slower, more disciplined approach to strategy development. Users can evaluate market signals, observe simulated portfolio performance, and refine their trading logic over time.

It combines traditional technical indicators like SMA and RSI with an open-weight language model that provides BUY, SELL, or HOLD decisions based on the current market state.

The goal is simple: test patiently, measure consistently, and let the results speak for themselves.

Demo

The platform runs locally, allowing users to experiment with simulated trades and track virtual portfolio performance.

Code

GitHub: anupam99jain/paper-trader

TradeSim AI is a lightweight simulated trading engine that combines technical analysis with LLM-guided decisions.

Key features include:

• Simulated order execution with virtual capital.
• Built-in technical indicators, including SMA and RSI.
• AI-assisted BUY/SELL/HOLD decisions.
• Portfolio value tracking throughout simulated trading.
• A modular architecture that makes it easy to experiment with additional indicators and strategy logic.

How I Built It

The backend is built in Python, with Qwen 2.5 7B serving as the open-weight AI model through Ollama.

The trading engine follows a straightforward workflow:

1. Fetch market price data.
2. Calculate technical indicators, including the 20-period SMA, 50-period SMA, and 14-period RSI.
3. Feed the current market state into a structured prompt.
4. Ask the model to return a BUY, SELL, or HOLD decision.
5. Execute simulated trades, account for transaction fees, and update the virtual portfolio.

Keeping the system modular makes it easier to compare different strategies, adjust the decision-making logic, and experiment with alternative open-weight models.

Rather than treating an individual trade as proof that a strategy works, the platform makes it possible to observe how decisions accumulate over a longer testing period.

Why Does Open Innovation Matter?

Open innovation makes experimentation more accessible.

Running an open-weight model locally through Ollama reduces dependence on metered, proprietary AI APIs and gives users greater control over how their market data and strategy logic are processed.

It also makes the underlying AI component easier to inspect, replace, and experiment with as new models become available.

For a project centered on repeated testing, these advantages matter. Developers should be able to run experiments, compare outcomes, and improve their strategies without every iteration introducing another API bill.

The broader idea is that useful AI-powered financial tools should be approachable for students and independent developers, not just people with expensive infrastructure or proprietary subscriptions.

My Agent Session

The model's decision-making loop is integrated into the trading engine, where it receives technical indicators and returns a structured trading action.

The Lazy Payoff

Backtesting removes much of the repetitive work involved in manually checking trading ideas against market data.

Instead of watching charts continuously or executing every hypothetical trade by hand, users can let the engine process market updates, calculate indicators, simulate orders, and track portfolio performance.

That leaves more time for the part that actually matters: understanding why a strategy works, where it fails, and whether its results hold up across a longer test.

The platform is designed to support experimentation, not promise profits. Simulated returns do not guarantee future performance in live markets.