What This Bot Does
Fully Automated — 24/7 Operation
This is not a directional trading bot. It does not try to predict whether Bitcoin goes up or down. Instead, it exploits the structural inefficiency of Polymarket's short-duration crypto markets — specifically the 5-minute and 15-minute price windows — by acting as a quasi market maker, capturing odds drift, and delta-hedging in real time.
The bot was observed active at 2–3AM ET with no signs of human involvement: the speed, consistency, and pattern of trades at sub-second intervals make manual operation impossible. This is a fully autonomous HFT system adapted for prediction markets.
$1.4M
All-Time P&L
37K+
Trades Executed
4
Assets Traded
5min
Core Timeframe
STRATEGY COMPONENTS
The 10 Strategy Pillars
01 · Market Focus
Exclusively targets short-duration crypto "Up or Down" markets: 5-minute windows (3:00–3:05, 3:05–3:10, 3:10–3:15 AM ET) and 15-minute windows (3:00–3:15 AM ET). Hourly candle markets are also included as secondary targets.
02 · Delta-Neutral Straddle
Simultaneously buys both Up AND Down on the same market window near 50/50 odds. This is not directional betting — it's capturing the spread and profiting when the odds drift away from equilibrium after entry.
03 · Ladder / DCA Entry
Instead of one large order, the bot places multiple incremental buys at successive price ticks: 50¢ → 51¢ → 52¢. This mimics traditional limit-order accumulation, achieving a better average entry price under slippage.
04 · Dynamic Delta Hedging
As the market moves, the bot sells the deteriorating side and doubles down on the winning side. Observed: sells Up at 32¢ after buying at 50¢, then loads heavily into Down. Classic real-time delta management.
05 · Odds-Gated Entries
Entry aggressiveness scales with how close to 50/50 the market is. Near-even odds see maximum capital deployment. Extreme odds (e.g. Up 6¢ / Down 94¢) see minimal or zero position. The bot never chases deep favorites.
06 · Dynamic Position Sizing
Position size is inversely proportional to odds extremity. At 50/50: 40–120 shares. At 70/30: 10–40 shares. At 90/10: 1–5 shares only. This Kelly-inspired sizing protects capital on asymmetric bets.
07 · Correlated Multi-Asset
BTC, ETH, SOL, and XRP are traded simultaneously in matching time windows. The bot exploits crypto correlation — if BTC sentiment turns bearish, it positions Down across ETH, SOL, XRP in the same window.
08 · Micro-Window Scalping
10–20+ trades within a single 5-minute window, scalping fractions of a cent in odds movement. This requires real-time orderbook monitoring and sub-second execution — impossible without full automation.
09 · Instant Redemption
Resolved positions are redeemed immediately. Zero idle cash. Capital recycled into the next window within seconds of settlement. This compounds the effective capital utilization rate dramatically.
10 · Low-Liquidity Hours
Operates heavily during 2–4AM ET, when human trading volume drops. Thinner orderbooks mean the bot's orders have more price impact, and inefficiencies persist longer before being arbitraged away by other participants.
EXECUTION LOGIC
Pseudocode Logic
Here is the reconstructed execution logic based on observed trade patterns. This is the closest approximation to the bot's core decision engine:
# ── POLYMARKET SCALPING BOT — RECONSTRUCTED LOGIC ──
ASSETS = ["BTC", "ETH", "SOL", "XRP"]
WINDOWS = ["5min", "15min", "1hr"]
loop every_second():
for asset in ASSETS:
for window in upcoming_windows(asset):
up_odds, down_odds = get_current_odds(asset, window)
imbalance = abs(up_odds - 0.50)
# ── ENTRY GATE: Only trade near 50/50 ──
if imbalance > 0.20: continue
size = calc_size(imbalance)
# size ∝ 1 / imbalance (max ~120 shares @ 50/50)
# ── STRADDLE: Buy both sides ──
if not has_position(asset, window):
ladder_buy("UP", asset, window, size, ticks=5)
ladder_buy("DOWN", asset, window, size, ticks=5)
# ── DELTA HEDGE: Manage live positions ──
else:
pos_up, pos_down = get_positions(asset, window)
delta = pos_up.value - pos_down.value
if delta > THRESHOLD: # Up side overweight
sell("UP", asset, window, qty=pos_up * 0.6)
buy("DOWN", asset, window, qty=size * 1.5)
elif delta < -THRESHOLD: # Down side overweight
sell("DOWN", asset, window, qty=pos_down * 0.6)
buy("UP", asset, window, qty=size * 1.5)
# ── CORRELATION: Check cross-asset signal ──
dominant_sentiment = get_btc_sentiment()
if dominant_sentiment == "BEARISH":
bias_down(assets=["ETH", "SOL", "XRP"], window=window)
# ── SETTLEMENT: Redeem resolved markets instantly ──
for market in get_resolved_markets():
redeem(market)
recycle_capital(market.payout)
Position Sizing Table
| Odds Range | Imbalance From 50/50 | Typical Share Size | Risk Level |
|---|---|---|---|
| ~50/50 (48¢–52¢) | < 2% | 80–130 shares | LOW ■ |
| 55/45 (45¢–55¢) | 2%–5% | 40–80 shares | MEDIUM ■ |
| 65/35 (35¢–65¢) | 5%–15% | 10–40 shares | MEDIUM ■ |
| 75/25 (25¢–75¢) | 15%–25% | 3–10 shares | HIGH ■ |
| 90/10 (10¢–90¢) | > 25% | 0–3 shares | EXTREME ■ |
BUILD GUIDE
How To Build This
Phase 01
Polymarket API Integration
Connect to Polymarket's CLOB (Central Limit Order Book) API. Authenticate via wallet (Polygon / MATIC). Subscribe to real-time orderbook WebSocket feeds for BTC, ETH, SOL, XRP markets. Parse market IDs for upcoming 5-minute and 15-minute windows. Library recommendation: py-clob-client (official Python SDK).
Phase 02
Odds Monitor + Signal Engine
Poll odds every 500ms for all target markets. Calculate the imbalance = |up_price − 0.50|. Trigger entry logic when imbalance drops below 0.20 (20% from center). Track odds history to detect drift direction and speed, which informs delta hedging decisions.
Phase 03
Order Execution Engine
Implement ladder buying: split target size into N incremental orders spaced 1–2 ticks apart. Use limit orders, not market orders — this captures spread rather than paying it. Implement retry logic and slippage guards. Target execution latency under 200ms per order.
Phase 04
Delta Hedging Loop
Every tick, calculate net delta of each position. If delta exceeds ±THRESHOLD (tune empirically, start at 15%), execute hedge: sell overweight side at market, buy underweight side with limit order. This is the P&L engine — capture the mean reversion to 50/50.
Phase 05
Cross-Asset Correlation Filter
Monitor BTC as the leading sentiment indicator. If BTC "Down" odds spike above 65¢, apply a directional bias to ETH/SOL/XRP positions — reduce Up exposure, increase Down. This adds alpha beyond pure market-making by following macro crypto momentum.
Phase 06
Settlement + Capital Recycling
Subscribe to resolution events via API. On resolution, call redeem() immediately. Track net payout and reinject into the capital pool. Log every trade to a database (PostgreSQL recommended) for performance analysis and strategy iteration.
# ── TECH STACK RECOMMENDATION ──
Language: Python 3.11+ (asyncio for concurrency)
API Client: py-clob-client (official Polymarket SDK)
Websockets: websockets / aiohttp
Scheduling: APScheduler (per-second polling)
Database: PostgreSQL (trade log + analytics)
Wallet: eth-account (Polygon MATIC signing)
Monitoring: Grafana + InfluxDB (live P&L dashboard)
Hosting: AWS EC2 us-east-1 (low latency to Poly servers)
Where The Edge Comes From
"The bot doesn't predict the future. It exploits the present — specifically, the temporary mispricing that occurs as human traders overreact to short-term price ticks on 5-minute markets."
— Strategy Analysis, March 2026
Edge Source 1: Odds Mean Reversion
On Polymarket's micro-window markets, odds frequently overshoot 50/50 as emotional traders pile in directionally. The bot fades these overreactions — buying the oversold side, selling the overbought side — and collects when odds revert. This is statistically robust on short windows where true probability really is ~50/50.
Edge Source 2: Speed Advantage
The bot executes at sub-second speed during 2–4AM ET when human liquidity is thin. It is effectively the fastest participant in these markets at those hours, allowing it to set the price rather than take the price.
Edge Source 3: Laddered Execution
By splitting orders into ticks, the bot achieves a better average price than any single market order would get. On a thin orderbook, this alone captures 1–3 cents of edge per position — which compounds enormously across 37,000+ trades.
Edge Source 4: Instant Capital Recycling
Settling and redeploying capital within seconds means effective capital utilization is dramatically higher than it appears. A $5,000 bankroll recycled 6x per hour is equivalent to $30,000 in deployed capital.
Key Risks To Manage
Directional Black Swan Risk: If BTC flash-crashes 10% within a 5-minute window, both Up and Down positions may lose (the Down bet wins but the odds may have been bought at the wrong price level). Implement a max-loss-per-window circuit breaker.
Liquidity Drying Up Risk: The strategy relies on being able to ladder in and out of positions. If a market suddenly has no counterparty, orders hang unfilled. Implement order timeout and cancellation logic after 10 seconds.
Copycat / Competition Risk: As more bots discover these micro-window markets, the edge compresses. Monitor win rate weekly — if it drops below 52%, the market is efficient and the strategy needs adjustment.
Risk Mitigation: Never deploy more than 15% of bankroll in a single 15-minute window across all assets. Set a daily drawdown limit at 8% of total capital — if hit, the bot pauses for 4 hours. Log every trade and review metrics weekly.
Final Architecture
╔══════════════════════════════════════════════════════╗
║ POLYMARKET SCALPING BOT — COMPLETE ARCHITECTURE ║
╚══════════════════════════════════════════════════════╝
INPUTS:
├─ Polymarket CLOB WebSocket (real-time odds)
├─ Polygon wallet (order signing)
└─ BTC sentiment feed (cross-asset signal)
CORE LOGIC:
├─ [1] Odds Monitor → detect near-50/50 markets
├─ [2] Entry Engine → straddle + ladder buy both sides
├─ [3] Size Calculator → Kelly-scaled by imbalance
├─ [4] Delta Hedger → sell drift, buy lag every tick
├─ [5] Correlation Filter → BTC bias → ETH/SOL/XRP
└─ [6] Settlement Engine → instant redeem + recycle
GUARD RAILS:
├─ Max 15% bankroll per window
├─ Daily drawdown circuit breaker (8%)
├─ Order timeout: cancel after 10s unfilled
└─ Win rate monitor: pause if below 52%
OUTPUTS:
├─ Trade log (PostgreSQL)
├─ Live P&L dashboard (Grafana)
└─ Profit (compounding)
This bot's $1.4M all-time profit over 37,781 trades is not luck. It's the result of a disciplined, systematic approach to exploiting structural inefficiency in prediction market microstructure — the same techniques used in traditional HFT, adapted for an emerging and less saturated arena.
The window to exploit these strategies is open now, while Polymarket's micro-window markets are still inefficient enough to generate edge. Build fast, monitor obsessively, and iterate weekly.