Arb signals on PolymarketBTC, ETH, SOL —audited in public.
Every arb our engine surfaces is logged on detection and reconciled at resolution. No survivorship bias, no hand-picked windows. Read the receipts before you read the pitch.
The receipts, refreshed continuously.
Two signal tiers. Both reconciled to the resolution outcome. Same data the dashboard shows paying users — public, no signup required.
From a Polymarket order book to a logged result.
Five steps. Each one is its own subsystem, each one writes its output to a queryable table. Nothing is opaque, nothing is hand-curated.
- 01Collect
Sub-second L2 snapshots of every BTC/ETH/SOL Up/Down market, joined with Binance spot, Bybit perp, and the Chainlink reference price.
- 02Model
Log-normal probability for each market — fitted on trailing realised vol, drift-adjusted.
- 03Detect
Logical arb when YES+NO walk-book < $1. Endgame when model disagrees with the market by > 15 points with τ < 5 min.
- 04Track
Every detection writes a row to arb_audit_log BEFORE any user sees it. Reconciled at resolution — wins and losses both.
- 05Publish
Public audit at /audit. Plus tier gets WebSocket push; Premium gets HMAC-signed webhook delivery to your URL.
The walk-the-book engine isn't a demo.
It's the same code that produces the audit above. Backtest any strategy against the same orderbook archive our detectors use. Paper-trade with the same fill model. Build your own conditions, run them on live snapshots — same engine, all the way down.
Walk-the-book backtest
Real fills, real fees, real spread. Mid-price simulations lie; this one tells you what an executing trader would actually realize.
Paper trader
Live snapshots, same walk_book() + platform_fee() the audit uses. Paper-vs-baseline tracking number flags when your simulation stops matching reality.
Strategy Builder
AND/OR condition trees, typed primitives, parameter sweeps. Save to the cloud, generate Python, lift it into your own bot.
The audit tells you what we won. Here's what it leaves out.
- ·Polygon gas (~$0.10–0.30/trade). Realized PnL on the audit is before on-chain execution cost.
- ·Execution lag. The detector logs at scan time (every 30 s). By the time a user clicks the row on the live page, the book has often moved.
- ·Adverse selection. When a logical arb fires it's usually because HFT bots haven't marked the cheap side yet; by the time a 30 s detector emits it, the window may already be closing.
Treat the "Realised / share" numbers on the audit as a generous upper bound on what an executing user actually captures. We log what we surface; we don't pretend the path from detection to filled order is frictionless.
Pay by how the audit reaches you.
Same engine, same data, same audit on every tier. You pay for delivery speed and integration depth — not for access.
See the audit. Browse all resolved markets. Build & backtest in the dashboard. 5 markets/backtest. 1 paper slot.
Free + REST poll /v1/arb/live for actionable signals. 20 markets/backtest, 2 concurrent.
Pro + WebSocket push /v1/arb/ws (sub-50 ms). 100 markets/backtest, 5 concurrent. 2 API keys.
Plus + HMAC-signed webhook delivery to your URL. 250 markets/backtest, 10 concurrent. 5 API keys.
Common questions before you sign up.
Is this a signal service?▾
Why audit publicly?▾
Is this affiliated with Polymarket?▾
Do you store my API key or strategy code?▾
What's the free tier good for?▾
What if my favourite feature isn't in any tier?▾
Read the audit. Decide for yourself.
No signup required to see every signal we've ever flagged, with the win or loss attached. We'd rather you check our work first.