Gasless, final, fair
Trade YES or NO on one deep book with no gas per order. Fills are final in one block, winning shares settle to 1.00 automatically — no claim transaction.
Layer 1 for prediction & outcome markets
Rain Chain runs a native order-book engine and a full EVM in the same 200 ms block, finalised in one slot. Merged YES/NO books, native resolution, randomness and bounded leverage are protocol primitives — not contracts bolted on top.
Status: Phase 0. The core matching engine is built and benchmarked; devnet, testnet and mainnet are not live yet. See the roadmap.
Built for
Every primitive an outcome business needs lives in the protocol: books, complete sets, resolution, randomness, settlement and builder fees.
Trade YES or NO on one deep book with no gas per order. Fills are final in one block, winning shares settle to 1.00 automatically — no claim transaction.
Cancels and post-only orders execute before takers in every block — a consensus rule, not proposer goodwill. Batches, dead-man switch, agent keys and 0 bps maker fee.
Builder codes pay up to 50 bps per fill. RainEVM contracts read books and outcomes via precompiles and use a native randomness beacon.
Seven genesis validators, growing to 26 by stake. Voting power is stake-weighted and capped at 10% per validator. Commission 5–20%, slashing enforced.
How it works
Consensus, the RainCore trading engine and RainEVM share a single state root per block. Funds move from Arbitrum One through a rate-limited bridge.
Key numbers
Protocol parameters from specification v0.2. Network latency and throughput figures are design targets until a public load test replaces them.
* Phase 0 benchmark of the RainCore engine alone: 296k–346k actions/s depending on block size, on one CPU thread, excluding signature verification and consensus. It is not a network throughput claim. The network design target is ≥ 20,000 order actions/s sustained.
The ordering rule
Within each block, RainCore actions run in three classes. A block that breaks the order is rejected by every validator.
A market maker who quotes and cancels is never picked off within a block by a taker that arrived in the same block. That is the single most important property for tight spreads on news-driven markets.
Read the rule in detailNative primitives
Buying NO at q is selling YES at 1 − q on the same book. Matches transfer shares, mint a complete set, or burn one — automatically.
Up to 64 outcomes with exactly one winner, kept arbitrage-consistent by atomic negRisk-style conversion.
Three tiers — AI proposal, AI dispute, stake-weighted oracle vote — with RAIN bonds and automatic settlement.
Threshold-BLS beacon every block. Unbiasable and verifiable; exposed to contracts via a commit-then-reveal precompile.
Isolated, long-only margin on outcome shares, funded by a lending vault, with a four-step liquidation waterfall.
A protocol liquidity vault that quotes new markets post-only, backstops liquidations and earns a share of fees.
Status & roadmap
Rain Chain is in development. No public network is running yet. About 9–11 months from start to a stable mainnet is planned.
RainCore state machine built: merged books, mint/burn-on-match, multi-outcome, ordering rule, order types. 70 tests passing; engine benchmarked. Next: consensus spike, 4-node devnet, signatures.
~6 weeks7-node testnet, resolution module, market-maker onboarding, bridge on Arbitrum Sepolia, RainEVM + precompiles, explorers, SDK.
~3 monthsAudits, 7 validators (≥ 4 independent), $25M deposit cap, 2× leverage on eligible markets, app migration.
~2–3 monthsOpen set to 26 validators, leverage to 3× then 5×, permissionless markets, spot listing auctions, caps lifted gradually.
6–12 monthsConsensus, RainCore, leverage, bridge, token and validator economics — every parameter, in one document.