Bitcoin Trade Bot

Automatically trades Bitcoin via the Coinbase Advanced API.

What it does

• Uses Coinbase Advanced API market data + account data.
• Generates buy/sell signals from EMA trend + RSI momentum.
• Accounts for net fees by:
  • Reading your Coinbase Advanced Trade SPOT maker/taker fee tier.
  • Previewing order commissions before execution.
  • Requiring expected edge to clear fee + profit thresholds.
• Supports guarded commands for convert and short/close-position actions.
• Includes walk-forward backtest command for config evaluation.
• Supports optional auto-actions in loop mode (auto_actions):
  • auto short open
  • auto short close-position
  • auto convert
• Writes graph-ready analytics to tradebot_metrics.json:
  • equity_curve: time series for realized/unrealized/net PnL.
  • trades: normalized trade events with side, size, fees, and realized PnL.

Important constraints

• Profit is never guaranteed.
• Default mode is paper (simulated trading only).
• Live execution is blocked unless all are true:
  • config.mode = "live"
  • CLI uses --execute-live
  • Environment variable TRADEBOT_ENABLE_LIVE=true

Guardrails implemented

• Max order USD cap.
• Max total position USD cap.
• Max daily buy USD cap.
• Max daily short-open USD cap.
• Minimum USD reserve.
• Daily realized loss stop.
• Optional daily realized profit lock (daily_profit_target_usd).
• Max trades per day.
• Cooldown between trades.
• Product allowlist.
• Convert pair allowlist.
• Shorting disabled by default and limited by notional + leverage caps.

Files

• tradebot/ bot implementation.
• config.example.json safe default settings.
• tradebot_state.json runtime state (created automatically).
• tradebot_metrics.json analytics output for charting (created automatically).

Setup

Requirements:

• Python 3.11+.
• Node.js 18+ (used only for ES256 JWT signing helper).

python -m venv .venv
.venv\Scripts\Activate.ps1
pip install -r requirements.txt
Copy-Item config.example.json config.json

Save your Coinbase API key JSON in a file named cdp_api_key.json in the project root folder.

Optional GUI dependencies:

pip install -r requirements-gui.txt

GUI loop updates use Streamlit fragments, so only live runtime sections rerun.

Usage

Run one cycle (paper mode):

python -m tradebot --config config.json run

Run continuously:

python -m tradebot --config config.json run --loop

Launch optional GUI dashboard:

python -m tradebot --config config.json gui

GUI options:

• --host (default 127.0.0.1)
• --port (default 8501)
• --no-browser to run headless

GUI notes:

• Run One Cycle Now in the GUI calls the same run_cycle logic as CLI run.
• GUI includes Start/Stop loop controls (websocket event-driven; no manual loop interval).
• When GUI loop is ON, a cycle is run when new websocket market data arrives.
• If websocket data is unavailable, GUI loop falls back to conservative polling.
• Backtest panel can include a synthetic scenario suite to test bull/bear/chop/volatility regimes.
• Live execution from GUI still requires:
  • mode = "live" in config
  • explicit GUI live checkbox + confirmation
  • TRADEBOT_ENABLE_LIVE=true
• GUI reads/writes the same runtime files:
  • tradebot_state.json
  • tradebot_metrics.json

Loop output behavior:

• In --loop mode, TradeBot suppresses duplicate console lines.
• A new line is printed only when cycle output changes (new signal/action-relevant information).

Auto actions in loop:

• Set auto_actions.enable_auto_short, auto_actions.enable_auto_close_position, and/or auto_actions.enable_auto_convert to true in config.
• Auto actions still obey live-execution safety gate and guardrails.
• Auto short-open and auto-convert also obey guardrails.cooldown_seconds.
• Auto checks are throttled by:
  • auto_actions.short_check_interval_seconds
  • auto_actions.close_check_interval_seconds
  • auto_actions.convert_check_interval_seconds

Real-time checks:

• Set loop_seconds to 1.0 for one decision cycle per second.
• Websocket market-data feed is enabled by default:
  • websocket.enable: true
  • websocket.market_data_url: wss://advanced-trade-ws.coinbase.com
  • websocket.channel: ticker (or ticker_batch)
  • websocket.stale_seconds: max age before REST fallback
  • websocket.ping_interval_seconds / websocket.ping_timeout_seconds
  • websocket.max_reconnect_seconds
• price_refresh_seconds controls cache/REST fallback behavior when websocket data is stale.
• Keep granularity at ONE_MINUTE (strategy candles are still 1-minute bars).
• Use:
  • candles_refresh_seconds to control how often candle history refreshes (example: 10).
  • fees_refresh_seconds to control how often fee tier refreshes (example: 300).
• Spot execution mode:
  • execution.prefer_maker_orders: true enables post-only maker orders for spot BUY/SELL.
  • execution.maker_price_offset_bps controls post-only limit offset.
  • execution.maker_max_order_wait_seconds controls stale-pending warning threshold.
  • execution.max_spread_bps blocks spot entries/exits when bid/ask spread is too wide.
  • In maker mode, bot output may remain HOLD while pending fills are being reconciled.
• Enable sub-minute signal overlay:
  • strategy.enable_subminute_signals: true
  • strategy.subminute_window_seconds (recent tick window, e.g. 90)
  • strategy.subminute_min_samples (minimum ticks needed)
  • strategy.subminute_min_signal_strength_bps (micro-signal threshold)
• Regime adaptation (optional but enabled by default):
  • strategy.enable_regime_adaptation
  • strategy.regime_volatility_lookback
  • strategy.low_volatility_bps / strategy.high_volatility_bps
  • strategy.trend_regime_threshold_bps
  • strategy.low_vol_threshold_multiplier
  • strategy.trend_threshold_multiplier
  • strategy.choppy_threshold_multiplier
  • strategy.high_vol_threshold_multiplier

Show status:

python -m tradebot --config config.json status

Convert (preview only unless live + --execute-live):

python -m tradebot --config config.json convert --from-currency USD --to-currency USDC --amount 25

Short open (guarded, derivative products only):

python -m tradebot --config config.json short-open --product-id BTC-PERP --base-size 0.001 --leverage 1

Close derivative position:

python -m tradebot --config config.json close-position --product-id BTC-PERP

Run walk-forward backtest:

python -m tradebot --config config.json backtest --lookback-candles 3000 --train-candles 600 --test-candles 180 --step-candles 90

Run backtest with synthetic scenario suite:

python -m tradebot --config config.json backtest --lookback-candles 3000 --train-candles 600 --test-candles 180 --step-candles 90 --include-scenarios --scenario-length 1800

Run synthetic scenarios only (skip historical download):

python -m tradebot --config config.json backtest --scenarios-only --scenario-length 1800 --train-candles 600 --test-candles 180 --step-candles 90

Scenario suite includes deterministic market regimes:

• bull_trend_pullbacks
• bear_trend_bounces
• sideways_chop
• breakout_then_reversal
• volatility_spike_regime

Metrics JSON

tradebot_metrics.json has two chart-friendly arrays:

• equity_curve: one point per cycle (ts, price_usd, realized_pnl_usd, unrealized_pnl_usd, net_profit_usd).
• trades: one point per executed trade (side, price_usd, base_size, quote_notional_usd, fee_usd, realized_pnl_usd).

Guardrail note:

• Set guardrails.max_trades_per_day to 0 to disable trade-count limiting.
• Set guardrails.daily_profit_target_usd to a positive number to stop opening new buy/short/auto-convert risk once realized daily PnL reaches that target.

Runtime output note:

• Each run cycle now includes:
  • price_source (websocket, cache, rest, cache_fallback)
  • signal_source (candle or subminute)
  • spread_bps, max_spread_bps
  • signal_threshold_bps, signal_volatility_bps, signal_regime
  • candle_signal, candle_signal_reason
  • candle_threshold_bps, candle_volatility_bps, candle_regime
  • subminute_signal, subminute_signal_reason, subminute_ticks (when sub-minute is enabled)
  • subminute_threshold_bps, subminute_volatility_bps, subminute_regime (when sub-minute is enabled)
  • pending_order_id, pending_order_status, pending_order_side (when maker order is awaiting fill)
  • daily_buy_used_usd
  • daily_buy_limit_usd
  • daily_buy_remaining_usd
  • daily_short_open_used_usd
  • daily_short_open_limit_usd
  • daily_short_open_remaining_usd so you can monitor progress toward the daily spend cap in real time.
• Spot execution quality checks may return HOLD reasons:
  • spread too wide for entry (...)
  • spread too wide for exit (...)
• status now includes websocket health/connection details under websocket.
• Terminal output is colorized by default (use_color_output: true in config). Disable with use_color_output: false or environment variable NO_COLOR=1. Additional color cues:
  • signal/action fields and signal source (candle vs subminute)
  • trend/RSI/edge and fee fields
  • auto-action reason fields (auto_close_reason, auto_short_reason, auto_convert_reason) Daily spend colors:
  • daily_buy_remaining_usd: green when available, yellow when low, red at 0.
  • daily_buy_used_usd: yellow as it accumulates, red when cap is reached.
  • daily_short_open_remaining_usd: green/yellow/red by remaining short-open budget.
  • daily_short_open_used_usd: yellow while opening shorts, red at cap.

Going live safely

1. Keep mode as paper until logs look stable for multiple days.
2. Use very low limits first (max_order_usd, max_position_usd, max_daily_loss_usd).
3. Only then set:

$env:TRADEBOT_ENABLE_LIVE="true"

and use run --execute-live.

Disclaimer

This software is intended for educational use only and does not constitute financial advice. The author assumes no responsibility for any financial gains or losses resulting from its use.