PAK
github.com/nicola-king/PAK
Local-first, modular Polymarket trading system exploiting Asia-US timezone lag with 9 core strategies and EIP-712 signed CLOB execution.
Timezone-arbitrage and multi-strategy: Python framework with 9 pluggable strategies including momentum, mean reversion, and timezone lag exploitation (trading US events during Asian hours when liquidity is thin).
Polymarket Alpha Kernel (PAK) is a modular trading framework with a focus on exploiting information asymmetry across timezones. The core thesis: during Asian trading hours, US-centric event markets on Polymarket see reduced liquidity and slower price discovery, creating opportunities for informed traders. PAK implements 9 strategy modules, each in its own Python class inheriting from a base `Strategy` interface. Strategies include: TimezoneTide (buys dips during off-hours), EMACross (trend following), MeanReversion (pairs within markets), and DeltaHedge (gamma scalp on nearing-resolution events). Execution is fully on-chain via Polymarket's CLOB using EIP-712 typed data signing—no centralized API. The code is clean with type hints, async/await patterns, and a configuration YAML file. A notable feature is the 'sandbox' mode that replays historical CLOB snapshots to simulate execution without real funds. Malware risk is low: no obfuscated binaries, standard Python dependencies, and the repo is tiny (94KB) with visible source. The main weakness is the lack of documentation for the strategy modules—the Chinese-language README is sparse on parameter explanations. Additionally, no backtesting results are published, making it a research project rather than a production bot. For quantitative developers interested in non-traditional Polymarket strategies, PAK offers fresh ideas.