TurboTrendReactor
Fast trend detection and quick entries and exits on the shortest timeframe.
- model
- XGBoost regressor (FreqAI)
- alt
- ValueReclaimFlux
- slots
- 5 open trades
Project · Machine learning · Automation
HITT-A-TRADE
A multi-timeframe, machine-learning crypto trading system: three FreqAI bots, an automated market-selection pipeline, layered risk controls and a custom command center. Designed and built by Jason Hitt on top of the open-source Freqtrade engine.
Running in dry-run (paper trading) mode
How it works
3bots in parallel: 15m · 30m · 1h
6FreqAI strategies in the library
32major coins in the tradable universe
60+automation & ops scripts
// The bots
Three timeframes, one orchestrated system
Each bot runs its own FreqAI model and strategy, with up to five open positions in USD pairs on Kraken. A shorter timeframe reacts quickly; a longer one rides bigger moves.
SwingPulseTrader
Momentum and swing structure in the middle of the range.
- model
- XGBoost regressor (FreqAI)
- alt
- RegimeValueReclaimer
- slots
- 5 open trades
MacroMomentumHunter
Longer holds that follow broader momentum and market regime.
- model
- XGBoost regressor (FreqAI)
- alt
- TrendValueGuardian
- slots
- 5 open trades
// Architecture
From raw market data to managed positions
- DataKraken OHLCV feeds plus imported historical trade archives, gap-repaired and integrity-audited.
- UniverseA fixed set of 32 major coins, ranked by live USD volume and filtered by health gates into each bot's whitelist.
- SignalsFreqAI trains XGBoost models per timeframe; a signal generator and fusion step combine model output with market-regime context.
- ExecutionThree Freqtrade bots act on signals with per-bot position limits, stops and exposure caps.
- OperationsWatchdogs, cron schedules, health checks, alerts and the Night Ops command center keep it all running.
// Under the hood
Built like infrastructure, not a script
Machine-learning signals
FreqAI pipelines with XGBoost models per timeframe, retrained on fresh data, plus signal fusion with market-regime and external market context.
Automated market selection
A whitelist pipeline ranks coins by volume, scores their health and syncs each bot's tradable list, with a global blacklist.
Risk management
Position limits per bot, capital-utilization tracking and stop-losses, with portfolio exposure and drawdown shown live.
Night Ops command center
A custom terminal dashboard in tmux, plus an HTML dashboard and API: bots, positions, P&L, risk and alerts on one screen.
Self-healing operations
Watchdogs restart missing bots, guard memory, free ports and pause or resume schedules, so the system recovers on its own.
Backtesting & hyperopt
Backtest suites, timeframe comparisons and hyperparameter optimization to test strategies before they ever touch a live bot.
// Stack
Tools & technologies
Credit where it's due: the trading engine is Freqtrade, the open-source algorithmic trading framework, with its FreqAI machine-learning module. Everything around it, from the strategies and pipelines to operations and the UI, is Jason's work.
Note: Hitt-A-Trade is a personal engineering project and currently runs in dry-run (simulated) mode. Nothing on this page is financial advice or an offer of any service.
// Contact
Questions about the system?
Happy to talk architecture, tooling or machine-learning pipelines.