About Dibs!

A forensic intelligence layer for Polymarket. We watch every trade, score the wallets behind them, and surface the ones with demonstrable information edge — before the crowd notices.

What we do

Dibs! is a continuously-running signal pipeline for prediction-market trading activity on Polymarket. We ingest the full firehose of public on-chain trades, score each one for "information edge" indicators (wallet track record, market context, coordination patterns, price-impact behaviour), and dispatch the highest-conviction signals to subscribers via Telegram.

The bet behind the project: a small subset of Polymarket wallets consistently know things the crowd doesn't, and their trading footprint shows it. Surfacing those wallets in real time is more valuable than any prediction model run from cold.

Where the data comes from

Everything we use is public:

• Polymarket CLOB WebSocket — the live order-book feed every Polymarket frontend already consumes. We just listen to a lot of it at once.
• Polymarket Gamma + REST APIs — market metadata, recent trades, market depth, resolution sources.
• On-chain Polygon trades — for historical depth, we backfill via Dune Analytics' indexed copy of Polymarket CTF fills (covers everything from the protocol's launch in 2022).
• Polymarket event metadata — for context that disambiguates short or jargon-heavy market questions.

We don't scrape private data, don't use insider sources, and don't have any privileged feed. The advantage is in aggregating and scoring what's already public, not in seeing what others can't.

How we score it

Every trade runs through a composite scoring pipeline built from roughly twenty engineered features about the wallet, the market, the trade itself, and the surrounding flow. The composite is augmented by a machine-learning classifier (gradient-boosted trees, retrained nightly on the full corpus of resolved Polymarket trades) that estimates win probability per trade.

The whole thing is tuned against a backtest mart of every alert we've ever dispatched and its eventual market resolution. As more outcomes accumulate, the model refits — every alert we send becomes training data for the next version. The signal engine is genuinely self-improving.

Which features matter most, how they combine, and where the thresholds sit is the part we keep private. The moat isn't that the building blocks are exotic — it's the months of tuning against real, resolved outcomes.

How alerts reach you

• Free public channel (@Dibs_Daily_Alerts) — daily post at 07:05 UTC: 7-day pipeline hit-rate plus a sample of the prior day's alerts.
• Paid private channel (€5/mo) — every flagged trade as it fires (~15-min cron cadence between batches). Cluster-aware framing when multiple wallets coordinate. Plus the same Daily Digest as the Free channel.
• Audit trail — every dispatched alert is logged against its eventual market outcome, feeding the backtest mart that drives weight-tuning. The signal engine is genuinely self-improving.

What we don't do

• We don't give investment advice. Alerts are observational signals; you make your own trading decisions. (See Terms.)
• We don't trade for you. There's no managed-money product here.
• We don't sell or rent subscriber data. (See Privacy.)
• We don't run analytics or behavioural tracking on visitors to this site. No cookies beyond what Stripe needs to process payments.
• We don't do drip-marketing emails. The welcome email and the Stripe receipts are it.

In development

Selected roadmap items in active or imminent development. The backlog is broader; this is what's most likely to ship next.

Contact

Questions, feedback, partnership / API enquiries: hello@dibs.horse.