IdiotProof: An Autonomous AI Trading System Based on Custom DSL and Dual ML Methods

IdiotProof is an autonomous AI trading system developed by the mindattic team, built using the C#/.NET 10 tech stack. Its core innovation lies in automatically converting traders' natural language strategy descriptions into executable automated trading strategies. Through the dual machine learning self-calibration mechanism (combining genetic algorithm optimizer and neural network weight learning) and 24/7 uninterrupted monitoring, it achieves a complete closed loop from idea to automated trading execution.

Project Background

IdiotProof aims to address traders' need to convert natural language ideas into automated strategies, enabling a "set-it-and-forget-it" passive trading experience without requiring users to stay up late at night monitoring the market.

Tech Stack

Built with C#/.NET 10, it includes two main modules: Blazor Server Web application and an independent monitoring console, ensuring system stability and scalability.

Blazor Server Web Application

Offers three ways to create strategies: visual flow chart editor, raw IdiotScript code input, and Claude Opus-generated natural language strategies. The interface includes features like strategy management, learning center, and API settings. Authentication and data persistence are implemented via ASP.NET Core Identity and Entity Framework Core.

Independent Monitoring Console

An independently running console application that loads active strategies 24/7, evaluates conditions, and triggers trading signals. The decoupled design ensures automated execution.

Natural Language to Code

IdiotScript is a fluent text DSL designed specifically for trading strategies. It supports near-natural language syntax (e.g., "Go long when price breaks previous high and RSI is less than 70") to directly convert into an executable condition tree.

Conditional Algebra System

Builds multi-level condition trees via .And/.Or/.Not operators. The StrategyBuilder and Conditions static directories provide pre-built condition templates for price behavior, technical indicators, etc.

Genetic Algorithm Optimizer

Automatically searches for optimal strategy parameter combinations: creates parameter population → backtests to evaluate fitness → selection/crossover/mutation → iteratively evolves to the best configuration, adapting to different market environments.

Neural Network Weight Learning

The dual ML method enables self-calibration of indicator weights for each stock. The neural network learns the predictive power of indicators from historical data, dynamically adjusts weights, and optimizes strategies in a targeted manner.

Indicator Calculation Engine

Provides classic technical indicators like ADX, ATR, Bollinger Bands, EMA, and candlestick pattern recognition. All calculations are stateless pure functions, facilitating testing and backtesting.

Data Feed System

The IMarketDataFeed abstraction layer supports Polygon real-time data and simulated data. SwitchableMarketDataFeed enables seamless switching between live and simulated trading, while MockDataFeed provides historical playback.

Broker Abstraction Layer

Implements the Alpaca interface (simulated/live). SandboxBrokerClient serves as a safe fallback, and the IBKR adapter can be enabled when dormant.

Three-tier storage strategy:

1. %APPDATA% shared keyring stores LLM and broker API keys;
2. %LOCALAPPDATA% stores application-specific state;
3. SQL Server is the primary storage, persisting core data such as strategies, user preferences, and audit logs, ensuring data consistency between Blazor and Monitor.

Application Scenarios

• Quantitative traders: complete toolchain from conception to execution;
• Part-time traders: "set-it-and-forget-it" passive experience;
• Strategy developers: Claude-driven to lower prototype development barriers.

Summary

IdiotProof represents cutting-edge exploration in AI quantitative trading. It builds a complete closed loop through custom DSL, dual ML optimization, and 24/7 monitoring. Its modular architecture and security design provide a reference implementation for financial AI applications.