AI-Powered Crypto Trading Bot | Data To Money

| Metric | Value | |-------------------|-----------| | initial_balance | 10000.0000| | final_balance | 13509.9919| | total_return_pct | 35.0999 | | total_trades | 519.0000 | | win_rate | 0.6859 | | avg_profit | 6.7630 | | max_drawdown | -28.7804 | | Period | 1 month |

Video

Data To Money Documentation

Overview

This bot is an automated cryptocurrency trading system designed to operate on the Bitget exchange. It fetches real-time market data, applies predictive analytics using a pretrained neural forecast model, and executes buy/sell or exit strategies based on intelligent signal generation.

It uses MongoDB to store historical data, allowing for real-time inference, tracking, and decision-making. It also includes performance tracking and data export functionalities.

---

Features

• ✅ Real-time candle fetching from Bitget.
• ✅ Uses a neural forecasting model (TimeMixer) for price prediction.
• ✅ Intelligent signal generation based on historical + forecast analysis.
• ✅ Automatic TP/SL management for risk control.
• ✅ Supports both scalping and swing trading.
• ✅ Trade logging and performance metric reporting.
• ✅ Sends webhook alerts (optional) for executed trades.
• ✅ CLI command ("hey bot") to export current performance snapshot.

---

Architecture

1. Market Data Handling

• Source: Bitget's spot market endpoint (/api/v2/spot/market/history-candles).
• Granularity: ?min intervals.
• Stored in: MongoDB collection (default: bitget10K.ETHUSDT_??MA_timeseries).
• History Window: ??? data points used for modeling, updated every minute.

2. Prediction Model

• Model Type: NeuralForecast Model (TimeMixer).
• Input: Time series of ??-period average close prices (ten_avg).
• Output: 30-point future forecast used to derive predictive trends.

---

Core Components

Class: GoldCollector

Initialization

runbot = GoldCollector(token="ETHUSDT")

Main Parameters

| Name | Description | |------|-------------| | lookback | Historical points for context (???) | | forward | Prediction length (??) | | hold | Window for recent comparison (??) | | tp_percent / sl_percent | Take-profit / stop-loss thresholds | | volume_multiplier | Trade volume filter to confirm breakout | | position_size | Capital allocation per trade (as % of balance) | | initial_balance | Starting virtual capital (default $10,000) |

---

Trading Logic

Signal Generation (generate_signal)

• Historical Status:

  • "up" if current avg < all of last ??
  • "down" if current avg > all of last ??
  • "hold" otherwise

• Prediction Status:

  • Analyzed over 6 intervals (?, ??, ??, ??, ??, ?? points)
  • Encoded as ++++++ (bullish), ------ (bearish), or mix.

• Volume Check:

  • Must exceed volume_avg * volume_multiplier to trigger trade.

• Trade Decision:

  • BUY if hist = up & pred = ++++++ or +++++-
  • SELL if hist = down & pred = ------ or -----+

---

Trade Execution

Long/Short Buckets

• Trades are grouped and tracked in long_bucket or short_bucket lists.
• When enough time has passed since their entry, a collective exit (exit_buy or exit_sell) is triggered with computed TP/SL levels.

Exits (check_exits)

• Exit trades when:
  • Price hits either SL or TP.
  • OR trade duration exceeds 30 minutes.

Capital Management

• Every trade reduces available capital.
• Profits from closed trades are added back to the balance.

---

Logging & Notifications

Webhook Payloads

Payloads can be optionally sent to a webhook for real-time updates:

• Entry Notification (send_payload1)
• Exit Notification (send_payload2)

Webhook structure follows the ChartPrime format.

---

Backtest / Performance Metrics

Trigger "hey bot" in console to:

• Export all trades to CSV.
• Export performance summary.

Metrics Calculated

| Metric | Description | |--------|-------------| | total_return_pct | % change in balance | | win_rate | % of profitable trades | | avg_profit | Average PnL per trade | | max_drawdown | Maximum single-trade loss |

---

File Export

• Trades: {token}_trades_{timestamp}.csv
• Metrics: {token}_metrics_{timestamp}.csv

---

Execution Flow

1. GoldCollector.run_gold_collector() runs every 60 seconds:

   • Fetch historical + latest candle
   • Run signal & prediction
   • Check exits and update trades

2. schedule_gold_collector() runs in a daemon thread.

3. listen_for_input() listens for hey bot to save data snapshot.

---

Sample Trade Output

{
  "trade_id": 42,
  "ticker": "ETHUSDT",
  "entry_time": "2025-04-19T12:01:00Z",
  "action": "buy",
  "entry_price": 3185.42,
  "size": 1000.0,
  "sl": null,
  "tp": null,
  "status": "open",
  "hist_status": "up",
  "pred_status": "++++++"
}

---

Future Improvements

• Add live order placement with exchange API (Bitget)
• UI dashboard for real-time visualization
• Use dynamic risk allocation (Kelly criterion / volatility targeting)
• Integrate multi-token support (e.g., BTC, SOL, etc.)
• Add unit tests & logging module