Just Built a Dutch Book Arbitrage Bot for Polymarket's 15-Minute Crypto Markets

#polymarket #arbitrage #crypto #bot

Just shipped a Dutch book arbitrage bot for Polymarket's 15-minute crypto markets.

This bot delivers guaranteed profits (in theory) by exploiting brief pricing inefficiencies in the binary up/down contracts — no market direction prediction needed.

How the Arbitrage Works

Polymarket's 15-minute markets (e.g., "Will BTC be up or down in the next 15 minutes?") have two complementary outcomes:

UP (yes shares)
DOWN (no shares)

In a perfectly efficient market, the ask prices should satisfy:

UP_ASK + DOWN_ASK = 1.00

But due to liquidity imbalances, lag, or temporary mispricings, sometimes:

UP_ASK + DOWN_ASK < 1.00

When that happens → buy both sides → pay less than $1 total → get exactly $1 back at resolution (one outcome always wins).

That's classic Dutch book arbitrage — risk-free profit (minus fees/gas, assuming fast execution).

Bot Architecture

Real-time monitoring via Polymarket's WebSocket API for order book updates on BTC, ETH, SOL, XRP 15-min markets (and potentially more).
Detect when UP_ASK + DOWN_ASK < threshold (e.g., < 0.995 to account for fees).
Automatic trade execution: Place simultaneous buy orders on both sides.
Handle resolution automatically (profits realized when market resolves).

Early runs show opportunities appearing multiple times per hour during volatile periods, especially around news or high-volume trading.

Tech Stack (feel free to customize this section)

Python (or Rust/Go for lower latency?)
WebSocket client for Polymarket's streaming API
Order placement via Polymarket's trading API (authenticated)
Basic risk checks: position sizing, fee estimation, slippage guards

Caveats & Reality Check

Fees eat into tiny edges — need low-latency setup and optimized sizing.
Liquidity — thin markets can move against you mid-execution.
Resolution risk — Polymarket oracles are solid, but edge cases exist.
Rate limits & bans — don't hammer the API.
Not financial advice — just a fun engineering experiment!

This is a great example of how prediction markets create programmable financial primitives. Prediction markets + bots = DeFi's next frontier?

Repo coming soon (or already open-source? link if yes). Would love feedback or ideas for improvements — e.g., multi-market scanning, cross-chain arb, or integrating with other pred markets like Kalshi.

What do you think — viable strategy long-term, or will the market makers close these gaps forever?

Contact Info
Questions, Support, or strategy ideas? Reach out: