PaperHedge 白皮书
v1 · 2026 年 10 月 · 开源 · paperhedge.xyz
1. 一句话
PaperHedge 是一个托管工具：在 Papertrade 上同时开多和空，让亏损那一腿铸 PAPER，赢的那一腿把钱赚回来，用大约 1/6 的成本拿到同样多的 PAPER。服务器 24 小时执行，你只授权开平仓、只放愿意烧掉的钱。
2. PAPER 是什么
Papertrade 是 HyperEVM 上的合成永续合约交易所：交易者和一个资金池（LP）对赌。交易者每亏 1 美元，协议按当前"铸币率"铸出 PAPER 给他。PAPER 不能转账，质押后按比例分平台手续费（USDC）。
铸币率不是固定的：池子低于 200 万美元时每美元铸 100 枚；超过后按尾部进度衰减——10 月 13 日 23 枚，10 月 15 日已降到 14 枚。越早挖越便宜。
3. 为什么对冲便宜
同一个账户里同时开等量的多和空。价格走到 ±x% 时两腿一起平：
- 亏的那腿：亏
名义 × x，全额铸 PAPER（协议扣 2% 后作为铸币基数）。
- 赢的那腿：赢
名义 × x，但协议会对盈利收一笔"冲击费"（haircut）再加 2% 手续费。
净花的只是那笔冲击费和手续费。冲击费的公式来自 Papertrade 文档：
scale = (1 − 0.10) / (1 + 1/(x·15000) + 100000/(1e6·x·M))
M = 814.6 (BTC) / 484.0 (ETH)
赢家实收 = 名义·x·scale·0.98
按 BTC 算，不同走动幅度下赢家能留住的比例：
| 走动 x | 赢家留住 | 净花 / 亏损额 | 每 1000 PAPER（铸币率 14） |
|---|
| 0.2% | 80.6% | 19.4% | $14.2 |
| 0.3% | 83.0% | 17.0% | $12.4 |
| 0.5% | 85.0% | 15.0% | $10.9 |
| 1.0% | 86.6% | 13.4% | $9.8 |
| 直接亏着挖 | — | 100% | $72.9 |
走动越小一轮越快，但每枚越贵；±0.3%–0.5% 是速度和成本的平衡点。
4. 它怎么跑
- 读协议状态：没暂停、没公告、池子没欠款、haircut 参数没被改。
- 在 BTC/ETH 里选每 1000 PAPER 最便宜、且两边 OI 余量够的市场；按余额算每腿保证金和"目标能放下"的最高杠杆。
- 先开 OI 余量少的那边，再开另一边；第二腿没开成立刻平第一腿。
- 盯 Hyperliquid 中间价；到 ±x 两腿一起平；到紧急线（强平距离 × 0.6）强制平。
- 从成交记录读实际盈亏和铸出的 PAPER，记账、结算服务费，歇 20 秒再开。
- 只剩一腿时立刻平掉；断线重连后自动接管已有仓位；预算烧完自动停。
5. 费用
现在免费。以后如果开始收费，会提前在网站和 X 上说，模型是这样的：服务费 = 策略净已实现亏损的 3%。预付（开跑前按预算 × 3%，最低 1 USDC），每轮结算：赢的那腿把净亏损压小时，多收的退回余额；停了余额留着下次用；余额为负则全平暂停。手动开的仓不收。
网站上第 3 步有个"给作者一点鼓励"，是自愿的，转多少都行，不转功能一样。转的 USDC 直接从你钱包到我的地址，链上可查。这个工具我会一直改下去，有反馈就能更快。
6. 安全边界
| 动作 | 授权的密钥 |
|---|
| 开仓 / 平仓 | 能 |
| 提现 / 转账 / 动钱包其它资产 | 不能（协议层面只认钱包本人签名） |
| 质押 / 解押 / 领分红 | 不能（每次都要你钱包签） |
| 到期后 | 不能（最长 7 天） |
密钥在服务器上 AES-256-GCM 加密存储；钱包私钥从不经过本站。最坏情况（服务器被攻破）：攻击者能乱开仓亏掉保证金，拿不走钱。代码开源：github.com/borispie/paperhedge。细节见安全说明。
7. 风险
- 这是故意亏钱换 PAPER；PAPER 现在不能转账，值多少取决于平台后续。
- Papertrade 规则和接口随时改；参数一变本站自动停开新仓，但已有仓位按原规则处理。
- relayer 排队会让成交价偏离触发价；极端行情可能单腿残留、甚至强平一腿。
- 铸币率持续下降，同样的钱换到的 PAPER 越来越少。
8. 路线
- 已上线：自动选币、自动规模、抄底模式、断线接管、Telegram 告警、账单、一键质押/领分红。
- 下一步：策略历史导出、多策略并行、"亏损铸币"类项目的通用模板。
本文不构成投资建议。PaperHedge 是个人项目，与 Papertrade 无任何关系。作者：@siyiluo2
PaperHedge Whitepaper
v1 · October 2026 · open source · paperhedge.xyz
1. In one sentence
PaperHedge is a hosted tool that opens a long and a short on Papertrade at the same time: the losing leg mints PAPER, the winning leg earns the money back, and you end up with the same PAPER for roughly 1/6 of the cost. A server runs it around the clock; you only authorize open/close and only deposit money you are willing to burn.
2. What PAPER is
Papertrade is a synthetic perpetuals exchange on HyperEVM where traders trade against a liquidity pool. For every dollar a trader loses, the protocol mints PAPER to that trader at the current mint rate. PAPER is non-transferable; staked PAPER earns a share of protocol fees in USDC.
The mint rate is not fixed: 100 PAPER per dollar while the pool is under $2M, then decaying with "tail progress" — 23 per dollar on Oct 13, 14 per dollar by Oct 15. Earlier is cheaper.
3. Why hedging is cheaper
Open equal long and short in the same account and close both when price moves ±x%:
- Losing leg: loses
notional × x, all of it mints PAPER (98% after the protocol's 2% is carved out).
- Winning leg: wins
notional × x, minus the protocol's asymmetric-impact haircut and a 2% win fee.
Your net cost is only the haircut plus fee. The haircut formula is from the Papertrade docs:
scale = (1 − 0.10) / (1 + 1/(x·15000) + 100000/(1e6·x·M))
M = 814.6 (BTC) / 484.0 (ETH)
winner receives = notional·x·scale·0.98
On BTC, what the winner keeps at different move sizes:
| Move x | Winner keeps | Net cost / loss | Per 1,000 PAPER (rate 14) |
|---|
| 0.2% | 80.6% | 19.4% | $14.2 |
| 0.3% | 83.0% | 17.0% | $12.4 |
| 0.5% | 85.0% | 15.0% | $10.9 |
| 1.0% | 86.6% | 13.4% | $9.8 |
| Plain loss mining | — | 100% | $72.9 |
Smaller moves complete faster but cost more per PAPER; ±0.3%–0.5% is the sweet spot between speed and cost.
4. How it runs
- Read protocol state: not paused, no notices, no queue debt, haircut parameters unchanged.
- Pick the market (BTC/ETH) with the lowest cost per 1,000 PAPER and enough open-interest headroom on both sides; size each leg from your balance and use the highest leverage the target still fits under.
- Open the side with less headroom first, then the other; if the second leg fails, close the first immediately.
- Watch the Hyperliquid mid price; close both legs at ±x; force-close at the emergency line (0.6 × liquidation distance).
- Read realized PnL and minted PAPER from the trade history, record the round, settle the fee, pause 20 s, repeat.
- Close a stray single leg at once; re-adopt open positions after a disconnect; stop when the budget is spent.
5. Fees
Free for now. If fees are ever switched on, it will be announced here and on X first. The model: service fee = 3% of the strategy's net realized loss. Prepaid (budget × 3% before start, minimum 1 USDC), settled every round: when a winning close shrinks the net loss, the excess is refunded to your balance; the balance stays for next time when you stop; a negative balance flattens and pauses. Manual positions are not charged.
Step 3 on the site is an optional "buy the author a coffee": any amount, or none, and the tool works the same. USDC goes straight from your wallet to my address, verifiable on-chain. I keep improving this tool; feedback makes it faster.
6. Security boundary
| Action | The authorized key |
|---|
| Open / close positions | Can |
| Withdraw / transfer / touch other wallet assets | Cannot (the protocol only accepts the wallet's own signature) |
| Stake / unstake / claim | Cannot (your wallet signs each time) |
| After expiry | Cannot (7 days max) |
Keys are stored AES-256-GCM encrypted on the server; your wallet private key never touches this site. Worst case (server compromise): an attacker could open bad positions and burn margin, but cannot take funds out. Source: github.com/borispie/paperhedge. Details in the security notes.
7. Risks
- This deliberately loses money for PAPER; PAPER is non-transferable today and its value depends on the platform.
- Papertrade's rules and interfaces change often; a parameter change halts new positions site-wide, but open positions settle under the old rules.
- Relayer queuing moves fills away from the trigger price; extreme moves can leave a single leg or liquidate one.
- The mint rate keeps falling, so the same money buys less PAPER over time.
8. Roadmap
- Live: auto market selection, auto sizing, dip mode, disconnect takeover, Telegram alerts, billing, one-click stake/claim.
- Next: history export, parallel strategies, a reusable template for other loss-minting protocols.
Not investment advice. PaperHedge is a one-person project, not affiliated with Papertrade. Author: @siyiluo2