Polymarket Btc Up Down Trader

Polymarket BTC Up-Down Trader

Trade Polymarket's BTC daily and weekly UP/DOWN markets with built-in exit discipline. Enters on Binance momentum signals, exits automatically before resolution tail risk bites.

🚨 Framework, not a production trading system. Read DISCLAIMER.md before connecting to a wallet with real funds.

Polymarket only. All trades execute on Polymarket with real USDC. Use --live for real trades; default is dry run.

Not for fast (5m/15m) markets. Use polymarket-fast-loop for those. This skill targets daily and weekly BTC direction markets with hours-to-days of duration.

What's New in v1.2.3

• Removed the "Oracle-Engineering Halt Gate" section. This skill never called that detector, and it does not trade the ≤5-minute markets the gate targets. No code change.

What's New in v1.2.2

• Live NO entry is YES-scale. avg_cost is the held-side price. A NO bought at 0.30 now stores entry 0.70 so target_hit can fire. Replay rows have no avg_cost and were already correct.

What's New in v1.2.1

• Exit monitor reads real Position objects. client.get_positions() returns dataclasses, not dicts. Replay rows have shares + cost_basis only. Side, size, and YES-scale entry are derived. A missing sources list under replay is this skill.

What's New in v1.2.0

• Replay plane (SIM-5442). Under simmer backtest (SIMMER_REPLAY=1) discovery, prices, and the clock come from the existing replay harness — same SIMMER_REPLAY / SIMMER_REPLAY_NOW pattern as weather. Live Gamma and CLOB midpoint stay dark. Candles were already replay-safe. The exit monitor reads SDK Position objects and replay share-rows (no .get crash after the first fill).
• Keep / kill / fix for this path is in Troubleshooting. Do not add a second capital lane until a full-tape run can produce an honest KEEP or KILL.

⚠️ BTC UP/DOWN markets carry Polymarket's crypto taker fee. Effective rate is 3.5% at 50¢, up to ~6.6% on cheap shares. Makers pay 0% and earn a 20% rebate from collected taker fees. Factor this into your minimum edge threshold.

Setup Flow

1. Install the Simmer SDK

   bashCopy code
   pip install simmer-sdk
   

2. Set your API key

   • Get from simmer.markets/dashboard → SDK tab
   • export SIMMER_API_KEY="your-key-here"

3. Set wallet private key (required for live trading)

   • export WALLET_PRIVATE_KEY="0x..."
   • Not needed for dry-run or managed wallets.

4. Configure exit discipline (optional — defaults are well-tuned)

   bashCopy code
   python strategy.py --config
   

Quick Start

bashCopy code
# Dry run — see what would happen
python strategy.py

# Go live
python strategy.py --live

# Monitor-only (check exits without new entries)
python strategy.py --live --monitor

# Quiet mode for cron
python strategy.py --live --quiet

How to Run on a Loop

Linux crontab:

Copy code
# Every 30 minutes
*/30 * * * * cd /path/to/skill && python strategy.py --live --quiet

OpenClaw native cron:

bashCopy code
openclaw cron add \
  --name "BTC Up-Down Trader" \
  --cron "*/30 * * * *" \
  --session isolated \
  --message "Run BTC up-down trader: cd /path/to/skill && python strategy.py --live --quiet"

Configuration

All settings configurable via config.json, environment variables, or --set:

bashCopy code
python strategy.py --set exit_before_resolution_hours=2.0
python strategy.py --set volume_spike_exit_multiplier=4.0
python strategy.py --set target_hit_capture_pct=0.75

Exit Discipline Knobs

Why these defaults? Exit behavior on short-dated prediction markets skews heavily toward early exits — the final hours before resolution compress remaining upside while tail risk (last-minute reversals, oracle variance) stays constant. The defaults are calibrated to exit early while capturing the bulk of the move. Empirical work on short-dated crypto UP/DOWN exit patterns informed the calibration; adjust based on your own backtest results.

Entry Knobs

Example config.json

jsonCopy code
{
  "exit_before_resolution_hours": 1.0,
  "volume_spike_exit_multiplier": 3.0,
  "target_hit_capture_pct": 0.85,
  "entry_threshold": 0.05,
  "min_momentum_pct": 0.3,
  "max_position": 10.0,
  "daily_budget": 50.0
}

Exit Logic

Each run cycle evaluates open positions against three triggers, in priority order:

1. time_cap — Time-based exit

Closes position when the market is within exit_before_resolution_hours of resolution.

Copy code
hours_to_resolution ≤ exit_before_resolution_hours → EXIT (time_cap)

This is the hard floor. Holding to settlement adds tail risk (oracle variance, thin liquidity, last-minute manipulation) with little remaining upside on a near-resolved market.

2. target_hit — Profit capture exit

Closes position when it has captured target_hit_capture_pct of the theoretical maximum gain.

Copy code
YES positions: (current_price - entry_price) / (1.0 - entry_price) ≥ 0.85 → EXIT (target_hit)
NO positions:  (entry_price - current_price) / entry_price ≥ 0.85 → EXIT (target_hit)

Example: Bought YES at $0.40. Maximum gain = 1.0 - 0.40 = $0.60/share. At $0.91, captured = (0.91 - 0.40) / 0.60 = 85% → exit.

3. volume_spike — Smart-money volume exit

Closes position when 10-minute CLOB volume exceeds volume_spike_exit_multiplier× the rolling baseline.

Copy code
current_10m_volume / avg_baseline_volume ≥ 3.0 → EXIT (volume_spike)

A sudden surge in volume near resolution typically signals that informed traders have positioned ahead of outcome certainty. Exiting into the spike captures the liquidity premium.

Exit Reasons

Every exit logs one of four reasons:

CLI Reference

bashCopy code
python strategy.py                      # Dry run (show positions + opportunities)
python strategy.py --live               # Execute real trades
python strategy.py --monitor            # Only run exit checks (no new entries)
python strategy.py --positions          # Show open positions
python strategy.py --config             # Show current config
python strategy.py --set KEY=VALUE      # Update config
python strategy.py --quiet              # Only output on trades/errors

Source Tagging

All trades are tagged source: "sdk:btcupdown". This keeps BTC UP/DOWN P&L separate from other strategies in your Simmer portfolio.

Troubleshooting

Replay KEEP / KILL (SIM-5442)

Reuse simmer backtest / SIMMER_REPLAY=1. Do not invent a second harness. Pinned path check:

bashCopy code
python -m pytest skills/polymarket-btc-up-down-trader/tests/test_replay_plane.py -q

"No active BTC UP/DOWN markets found"

• Live: Gamma API may be slow or markets may not be trading. Try again in a few minutes.
• Replay: the skill lists from the tape with q=bitcoin up or down. A high-volume slice with 0 Up/Down dailies is FIX, not "no edge".
• Check polymarket.com directly for active BTC UP/DOWN markets.

"Could not fetch BTC momentum"

• Binance API rate-limited. Increase lookback_minutes or wait.

"Momentum X% < minimum Y%"

• Signal not strong enough. Entry threshold working correctly. Adjust min_momentum_pct if too conservative.

Volume spike trigger not firing

• Default baseline needs 6 prior 10-minute windows of data. In low-volume markets, set volume_baseline_windows lower or increase volume_spike_exit_multiplier.

"External wallet requires a pre-signed order"

• WALLET_PRIVATE_KEY is not set. Export the private key for your Polymarket wallet.
• The SDK handles all signing automatically once the env var is set.

"Balance shows $0 but I have funds on Polygon"

• Polymarket V2 uses pUSD (PolyUSD). Migrate at simmer.markets/dashboard (one click, ~30s).