Papertrade Dossier

Papertrade is a synthetic perp exchange on HyperEVM, Hyperliquid's EVM layer. It has no order book, no matching engine and no oracle of its own. The protocol's pool takes the other side of every trade, and the contract reads prices straight from Hyperliquid's order book. According to the founder, two or three people built it. It lists two markets: BTC and ETH.

It looks like it was built in a day, and technically that's close to the truth: there is very little to build. What's being sold is the mechanism, not the engineering.

Why it took off

• "You get paid for losing." PAPER is minted only from losses and liquidations. Win and you take the profit; lose and you get tokens.
• Casino settings. Up to 1000x leverage, no trading fee, no funding, "no slippage".
• The "next HYPE" narrative. Supply starts at zero: no team or VC allocation, no airdrop, no vesting.
• FOMO built into emissions. While the pool is under $2M, every $1 lost earns 100 PAPER, and the rate falls after that. Early users get several times more.
• Credibility and marketing. Media identify co-founder Jez as Colin Hong (MIT, Morgan Stanley, Standard Crypto). The launch video passed 1M views on X.

Where the price comes from

The PriceOracle contract reads Hyperliquid's best bid and ask through the system precompile 0x…080e and takes the midpoint. That is the entry and exit price. There is no TWAP, no median and no deviation check. The precompile returns HyperCore's state as of the moment the HyperEVM block is built, so you get the price of the block your transaction lands in.

The relayer

The relayer is an off-chain service run by the team. It takes signed orders and submits them to the BatchExecutor, paying the gas. It doesn't set prices, but it decides when and in what order a trade lands in a block. Liquidations go first, then closes, then opens, with large trades ahead of small ones. Opening positions, registering session keys and staking only work through the relayer. Withdrawals work without it.

The winners' queue

Winnings are paid only out of other traders' losses. If the pool is empty, unpaid profit joins a FIFO queue and waits for future losers. Winners get their margin back immediately. As of Oct 16 the queue is empty.

Fees and haircut

There is no trading fee and no funding. Winnings pay three things: a 0.2 bp deadband, an asymmetric haircut (larger on small moves) and a 2% win fee. On a losing close, the 2% is paid by the pool, not the trader. Share of a BTC win you keep at launch parameters:

Liquidations

Margin is isolated. You can't add margin, change leverage or partially close. Liquidation fires about 5 bp before the margin runs out: at 1000x, after a move of roughly 0.052%. The whole margin is lost, even though a manual close would have returned about half.

One wallet held 30 ETH shorts of about $9.9M each at 660x leverage, roughly $297M in notional. The $10M cap applied per position, not per wallet. At the same time the wallet moved the top of Hyperliquid's ETH book: it bought about 6,060 ETH (price up about $3), then sold about 5,210 ETH (down about $3.50). The result was about $1.28M in profit within a few hours.

• Moving a single quote for a single block was enough; there was no need to hold the price.
• At 1000x, even a fraction of a percent pays well: roughly 0.1% of $297M ≈ $300k.
• Hyperliquid itself wasn't hacked. The weakness is in how Papertrade uses its quote.

After the incident the team added the deadband, impact haircut and win fee, and cut OI caps. There is still no deviation check on the quote. The team said the funds were almost entirely returned through new accounts. The Block could not confirm this.

• Emissions. 100 PAPER per $1 while the tracked pool is under $2M. After that the rate decays along a high-watermark curve: about 30.8 by Oct 12, 16.6 by Oct 16.
• Staker income is paid in USDC. Half of the 2% fee plus everything in the pool above the cap. The cap is $5M now and rises to $10M on Oct 17. Until the pool gains another ~$5M, staker income will be close to zero.
• Transfers are disabled. PAPER can only be staked and has no market price. Supply is projected to reach 14–18B.
• Contract roles. The minter can burn PAPER from any address, and a privileged spender has an unlimited allowance over everyone's PAPER. On-chain these are PaperTokenomics and PaperStaking, both behind the Timelock.

The $5.06M daily average is heavily inflated by the first day, which brought in $24M. Today 1M PAPER earns about $190 a day, and the PAPER earned from $1 of losses earns about $0.003 a day.

The idea: open a position on Papertrade and hedge it on Hyperliquid. The loss on Papertrade earns PAPER, and the hedge's gain offsets it. Hyperliquid works better than other venues because that's where the price comes from. The real cost of the strategy is Hyperliquid fees, funding and execution mismatch.

Costs with a taker round-trip fee of $90 per $100k notional and emissions of 16.6 PAPER per $1:

Funding adds to the cost the longer you hold. Maker orders on Hyperliquid cut costs roughly threefold but add fill risk.

Pitfalls

• Leverage. No more than 100x: liquidation takes the whole margin while the hedge recovers only half. Aim to close manually after a 0.5–1.5% move.
• The price moves in your favor. Don't close: the profit gets a haircut and you get no tokens. You have to wait for a reversal and pay funding.
• The relayer. Small trades wait, so the two legs can close at different prices.
• Yield. On Oct 17 the staker cap doubles, and income may drop close to zero for a while.
• The token. PAPER can't be sold. While the pool is above the cap, the dollar you lose is shared among all stakers, mostly early farmers.

The addresses come from the official docs and were checked through the HyperEVM RPC (rpc.hyperliquid.xyz/evm) and the HyperCore API. Method: reading EIP-1967 proxy slots, calling view functions, simulating calls from different addresses with eth_call, and extracting function selectors from the bytecode.

Who controls it

The Timelock is also the owner() of every implementation. The multisig doubles as pause guardian: it can halt opens or retire a market, but only the Timelock can lift a global pause.

Attempts to bypass the Timelock (simulation)

The multisig also holds about $144k of the team's USDC. User funds are not there; they sit in the Exchange's HyperCore account.

Roles in Exchange

Withdrawals

The bytecode includes withdrawToCore(uint256). Simulating it from an arbitrary address and from the BatchExecutor returned the same errors: "zero amount" and "no balance". So there's no caller restriction, and you can withdraw straight from your own wallet.

There is also withdrawToCoreBySig(user, amount, nonce, deadline, v, r, s): you sign a message and anyone can submit it. This isn't a hole. There's no recipient parameter, so funds go to the signer's own HyperCore account, and the nonce and deadline stop the signature from being reused. If the relayer starts censoring, any other wallet can submit the signature.

Addresses

There is no official Papertrade source code on GitHub, neither a current nor an older version. Neither the docs nor the site link to GitHub.

aa-matthew/papertrade-contract-research third-party

A reverse-engineering study of the deployed contracts, dated Oct 15. The author states: "Original Papertrade Solidity remains unavailable". The useful part is the ABIs pulled from the official frontend: they confirmed withdrawToCoreBySig, setMintParams and unpauseMarket.

The other repositories from the past week are also third-party: farming bots, SDKs, dashboards and a clone of the project on another network.

Verdict