Skip to content

CLAUDE.md

Latest commit

 

History

History
163 lines (125 loc) · 4.34 KB

CLAUDE.md

File metadata and controls

163 lines (125 loc) · 4.34 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

NostalgiaForInfinity (NFI) is an advanced cryptocurrency trading strategy framework for Freqtrade. It implements a multi-mode trading system supporting both long and short positions across various market conditions.

Key Characteristics:

  • Main strategy file: NostalgiaForInfinityX7.py (~73,000 lines, latest version)
  • Required timeframe: 5 minutes (timeframe = "5m")
  • Multi-timeframe analysis: 5m, 15m, 1h, 4h, 1d + BTC reference
  • Startup candles: 800

Common Commands

Testing

# Run all tests (parallel execution)
pytest

# Run specific test file
pytest tests/unit/test_NFIX5.py -v

# Run backtests only
pytest tests/backtests/ -v

# Run with coverage
pytest --cov=. --cov-report=html

Code Quality

# Format code
ruff format .

# Lint check
ruff check .

# Fix linting issues
ruff check --fix .

# Run pre-commit hooks
pre-commit run --all-files

Docker Development

# Start trading bot
docker compose up -d

# Run tests in container
docker compose -f docker-compose.tests.yml up

# View logs
docker compose logs -f

Backtesting (via Freqtrade)

# Basic backtest
freqtrade backtesting --strategy NostalgiaForInfinityX7 --timerange 20240101-20250101

# With specific config
freqtrade backtesting --strategy NostalgiaForInfinityX7 -c configs/exampleconfig.json

Architecture

Core Lifecycle Methods (NostalgiaForInfinityX7)

Method Purpose
populate_indicators() Calculates all technical indicators across timeframes
populate_entry_trend() Generates long/short entry signals with mode tags
populate_exit_trend() Generates exit signals
custom_exit() Routes exits by trading mode via enter_tag
adjust_trade_position() Handles DCA (averaging down) and profit-taking

Trading Mode System

Trades are tagged with enter_tag to route to appropriate exit logic:

Mode Long Tags Short Tags Purpose
Normal 1-13 501-502 Standard trading
Pump 21-26 521-526 Quick upward momentum
Quick 41-53 541-550 Short-term profits
Rebuy 61-63 561 Position averaging
Rapid 101-110 601-610 High-frequency entry
Grind 120 620 DCA-focused
Scalp 161-163 661 Micro-profit trades
Top Coins 141-145 641-642 Large-cap focused
BTC 121 N/A Bitcoin-specific

Data Flow

populate_indicators() → Calculate 5m + merge 15m/1h/4h/1d + BTC indicators
         ↓
populate_entry_trend() → Evaluate conditions → Assign enter_tag
         ↓
Trade Execution → adjust_trade_position() (DCA/profit-taking)
         ↓
custom_exit() → Route by enter_tag → mode-specific exit method

Key Design Patterns

  1. Strategy Pattern: Trading modes via enter_tag routing
  2. Configuration Pattern: NFI_SAFE_PARAMETERS whitelist for JSON overrides
  3. Cache Pattern: process_only_new_candles = True + persistent profit targets

Configuration

Required Settings (in Freqtrade config)

{
  "use_exit_signal": true,
  "exit_profit_only": false,
  "ignore_roi_if_entry_signal": true
}

Strategy Parameter Override (via nfi_parameters)

{
  "nfi_parameters": {
    "futures_mode_leverage": 3.0,
    "stop_threshold_spot": 0.10
  }
}

Config Files Structure

  • configs/exampleconfig.json - Base template
  • configs/trading_mode-*.json - Spot/futures specific
  • configs/pairlist-*.json - Exchange pair lists
  • configs/blacklist-*.json - Excluded pairs

Code Style

  • Line length: 119 (Black)
  • Linter: Ruff (rules: E4, E9, F, B, Q)
  • Max complexity: 12 (McCabe)
  • Pre-commit: Ruff check + format

Testing Strategy

CI/CD Backtests (.github/workflows/backtests.yml)

  • Triggered on X7 strategy changes
  • Tests across 24 months (2024-01 to 2026-01)
  • Three exchange configurations: Kucoin Spot, Binance Spot, Binance Futures
  • Generates JUnit XML reports

Unit Tests

  • Located in tests/unit/
  • Focus on cache functionality and strategy components

Version History

Multiple strategy versions exist (X through X7). X7 is the active development version.

  • Signal tuning commits follow pattern: "signal X: fine tune"
  • All changes validated via automated backtesting