We deployed 3 oracle engines on
@SuiNetwork mainnet. Each one does something the old betting model thought was impossible.
⚡ WARP - Weighted Atomic Resolution Protocol
512 bets. One PTB. One gas fee. Zero partial states.
When the oracle fires, it doesn't loop through bets one-by-one. It assembles a Programmable Transaction Block with 512 instant_settle_bet calls and drops one atomic transaction. Either all 512 settle - or none do. Rollback is free.
Gas amortised across the batch = ~97% cheaper per bet vs sequential settlement.
This is what Sui's PTB architecture was built for.
🌊 FLUX - Fractional Liquidity Under eXecution
P2P odds markets, but with infinite depth.
A maker posts an offer. Takers don't have to fill the whole thing - they take FluxShard<T> slices. Each shard is an independent owned object. Oracle settles shards in parallel across PTBs. No shared-object contention between shards.
Sui's owned-object fastpath = shard settlement at single-validator speed (~50ms). No consensus overhead for what's already yours.
🔵 PULSE - Pari-mutuel Under-Liquidity Shifting Engine
No fixed odds. No house. Just a pool.
Side A vs Side B. Everyone's stake goes in. Oracle locks the pool pre-match, settles post-result. Winners claim proportional share. Fee goes to the protocol.
pulse_lock_pool → pulse_settle_pool → pulse_claim_winnings. Three calls. Fully on-chain. No off-chain payout logic.
WARP stats: 17 batches, 25 bets settled.
FLUX stats: 5 batches, 5 shards.
PULSE stats: 2 batches, 5 pools.
Live. Mainnet. Now.
@Sui_Bets - P2P sports betting. Just Sui.
@SuiNetwork @EmanAbio @Mysten_Labs @EvanWeb3
Most betting protocols use Sui like Ethereum with faster blocks.
We didn't.
Three on-chain engines. Each one exploits a different primitive that only exists on Sui.
WARP exploits PTBs as a settlement bus.
Not composability. Not chaining. Mass atomic settlement.
One PTB. 512 independent P2PMatchedBet objects. 512 oracle calls. One signature. One gas fee.
If input 301 fails — all 512 roll back. No partial settlement. Ever.
No other chain can do this. EVM processes one state transition per tx. Move on Sui processes N with the same atomicity guarantee.
We're running 97% cheaper per bet at batch size 100. Not an estimate — live on mainnet.
FLUX exploits the owned-object fastpath.
FluxShard<T> objects are owned, not shared. When the oracle settles a shard — no consensus. Single-validator execution. ~50ms.
Shared-object bets need validators to agree on ordering. Owned-object shards don't. That's a 8x latency gap baked into the data model.
The fractional liquidity is the product feature. The owned-object fastpath is the engineering reason it works at scale.
PULSE exploits cross-package capability auth.
PULSE is a separate deployed package. It calls p2p_betting::OracleCap from the original package - no upgrade, no migration, no re-deployment.
One cap object. Three packages. One oracle wallet.
That's Move's type system doing access control across package boundaries without a proxy contract, a multisig, a governance vote.
@SuiBets - the first betting protocol that was designed around Sui's execution model instead of ported onto it.