Skip to main content
ODDS connects Elyra to decentralized prediction markets, scanning for mispriced events and structural arbitrage opportunities that manual browsing cannot surface at scale. Rather than scrolling through hundreds of open markets looking for edges, you query Elyra and receive a ranked list of opportunities ordered by edge size, with the reasoning behind each one.

Supported platforms

Polymarket

The largest decentralized prediction market by volume. ODDS scans event lines, computes implied probabilities, and applies semantic clustering to identify related markets with divergent pricing.

DFlow

Solana-native prediction market with on-chain orderbooks. ODDS analyzes taker flow and available liquidity to surface opportunities where order flow and quoted odds are misaligned.

What ODDS detects

1

Cross-book line shopping

ODDS queries both venues simultaneously and identifies where the best available odds on the same event differ between books. You get the optimal entry point without comparing venues manually.
2

Semantic clustering and mispricing

On Polymarket, ODDS groups semantically related markets (events that are logically correlated) and flags cases where the implied probabilities are inconsistent across the cluster. A cluster where the sum of implied probabilities deviates materially from 100% is a structural signal.
3

Structural arbitrage identification

When correlated markets are priced divergently — for example, two markets on the same underlying outcome trading at odds that imply a riskless spread — ODDS surfaces that pair ranked by profit potential percentage.
4

Ranked opportunity output

All identified opportunities are ranked by edge size. You control how many results to surface and how many markets to scan using natural language or explicit flags.

Natural language examples

Scanning parameters

You can control the scope of an ODDS scan through natural language or direct flags when running the trade research pipeline.
Polymarket semantic clustering uses sentence-transformers for embedding-based grouping. On first run, a Hugging Face model is downloaded to .cache/huggingface/. If the library is unavailable, ODDS falls back to TF-IDF clustering automatically — no configuration required.
Implied probability calculations reflect current market prices only. Liquidity depth varies significantly between markets. An apparent edge in a low-liquidity market may not be executable at the quoted size.
Use the arbitrage table alongside the opportunities table. The arbitrage output surfaces correlated market pairs, while the opportunities table surfaces single-market mispricings. Both are generated in the same scan.