How DEX Perps Differ From CEX
// Onchain settlement, oracle dependence, MEV exposure, and why funding rates can diverge wildly between venues. The wasteland runs different rules.
A perpetual futures contract on a centralised exchange runs on private servers, settles in a private database, and depends on the exchange's word that the dollars are still there. A DEX perp runs on a public chain, settles to a public ledger, and depends on smart contracts and oracles you can read. Same product, radically different threat model.
// SETTLEMENT — WHO HOLDS THE LEDGER
CEX: the venue holds the ledger. Your account balance is a row in their database. When you deposit, your dollars become an IOU on their books. When the venue dies (FTX), the IOU dies with it — there's no chain you can appeal to.
DEX: the chain holds the ledger. Your collateral lives in a smart contract you can audit. When the venue's frontend goes down, your position still exists on-chain — you can interact with the contract directly through any RPC. The venue can't run off with the money in the same way; they can only freeze access to their UI.
The tradeoff: chain throughput is slower than database throughput. Hyperliquid's L1 settles roughly every 0.07 seconds; that's blazing for a chain, glacial for a CEX matching engine. Most DEX perp venues live in this gap — fast enough for human trading, slow enough to be exploited by sophisticated MEV.
// ORACLES — WHERE PRICE COMES FROM
Every perp needs an index price. CEX builds the index from its own internal book and a few external feeds. DEX builds the index from oracles — Pyth, Chainlink, custom TWAP feeds, sometimes the venue's own pool.
Oracle dependence is the DEX perp's core risk surface. Manipulate the oracle and you can manipulate liquidations. Mango Markets in 2022 lost $115M to an oracle manipulation attack — the attacker pumped the venue's underlying spot pool to inflate their MNGO position's mark, then borrowed against the inflated mark and walked. The vulnerability lived in the oracle, not the perp contract.
When picking a venue, ask: which oracle does it use, how is the oracle aggregated, what TWAP windows does the contract apply, and what happens if the oracle falters? Hyperliquid runs its own validator-attested feed. Drift uses a Pyth consortium. GMX v2 uses a hybrid Chainlink + custom signed oracle. Each has different blow-up modes.
// FUNDING RATES — WHY THEY DIVERGE
Centralised perp markets converge fast. Same retail flow, same arb desks, same maker-taker rebates. Funding rates across Binance, Bybit, OKX rarely diverge by more than a few basis points.
DEX perp markets diverge wildly. The reasons:
- Different liquidity pools. Hyperliquid's HLP, Drift's DLP, GMX's GLP — each is a discrete liquidity layer, isolated from the others. Imbalances on one don't auto-arb against another.
- Bridge friction. Capital that wants to arbitrage HL ↔ Paradex has to bridge between Hyperliquid L1 and StarkWare. That's not free, and it's not instant.
- Different fee structures and tier discounts. The same trader on the same trade pays different bps at different venues. The trader who's optimal on Lighter (zero taker) is suboptimal on Hyperliquid (3.5 bps), and vice versa.
- Different maker dynamics. Some venues run market-maker incentive programs that subsidise specific sides. The MM is paid to provide depth, which often correlates with which side gets the favourable funding.
The wider lesson: DEX funding gaps are real because the markets are not unified. The Forge exists because the gaps are real, persistent, and exploitable.
// MEV — THE TAX YOU DON'T SEE
On a CEX, your order hits the engine atomically. On a DEX, your order hits the public mempool first. Searchers and MEV bots see it sitting there. They can front-run, sandwich, or back-run it depending on the opportunity.
Most DEX perp venues mitigate MEV through:
- Private orderflow. Orders go through the venue's API, never into the public mempool. Hyperliquid does this; you submit orders to their RPC, they sequence privately.
- Frequent batch auctions. Orders within a window settle at one price together, removing the "first-in" advantage.
- TWAP oracles. Even if MEV manipulates spot for one block, a TWAP feed averages it out before liquidations fire.
When MEV mitigation isn't strong, the visible cost shows up as wider effective spreads. The taker bps you see on the venue's published fee schedule isn't the full picture — slippage on a thin DEX perp market is often 5–15 bps on top of the headline taker. The Forge's cost ledger bakes a slippage estimate into every venue.
// CUSTODY — WHO CAN FREEZE YOU
CEX: the venue can freeze your account by policy decision. KYC failure, compliance flag, sanctions screening, internal investigation — any of those locks your funds for hours, days, or forever.
DEX: nobody can freeze your wallet. They can freeze the venue (deny you access to the UI), but the wallet itself stays under your control. Connect to a different frontend, hit the contract directly through Etherscan or a custom script, route through a different aggregator.
The asterisk: stablecoins. USDC and USDT can freeze any wallet at the issuer level. If your DEX perp collateral is in a freezable stablecoin and you somehow land on a sanctions list, the chain stays open but your collateral becomes inert. Decentralised collateral (ETH, BTC via wrapped variants, native stablecoins like crvUSD) reduces this risk; it doesn't eliminate it.
// LIQUIDATION — WHO PULLS THE TRIGGER
CEX: the venue's risk engine pulls the trigger. Internal logic, often opaque, sometimes generous (some venues add small grace windows for whales), sometimes brutal (some venues liquidate the entire position the moment maintenance margin breaches).
DEX: liquidations are public. Anyone can call the contract's liquidate(account) method when the position is underwater, and the protocol pays them a bounty. This is good (eliminates favouritism) and bad (during volatility, a flood of liquidator bots compete to close your position the millisecond it's underwater, leaving zero grace).
The practical implication: DEX perp positions need wider buffers than CEX positions. Where 5% above liquidation might be safe on a CEX with a slow risk engine, the same buffer on a DEX gets blown by the first liquidator bot that wakes up. Set wider stops. Watch the buffer.
// THE FORGE'S ANSWER
The DEX perp landscape is messy by design — many venues, many oracles, many funding regimes, many failure modes. That mess is also the opportunity. Funding rates diverge precisely because the markets aren't unified. The Forge's whole job is to read the divergence and tell you where the spread lives.
Read the venue. Read the oracle. Read the funding rate. Read the spread. Run the play.
// READ THE BOARD // READ: DELTA-NEUTRAL CARRY// THE CLOSING DISPATCH
CEX is opaque infrastructure with smooth UX. DEX is transparent infrastructure with rough edges. The Forge bets on transparency winning over decade-scale time horizons — and on the rough edges meaning real opportunity for traders who learn to read them.
Welcome to the wasteland.
// READ THE RISK BRIEFING