资金费率、手续费与风险载费

Dexter 中每一项手续费或载费组件都在引擎内计算——会计层稍后对总额对账,但下表数字都源自结算底层成交的同一次有序状态推进。

#资金费率如何进入账本

Dexter 的资金费率按每个市场每 8 小时结算。引擎持续测量方向性不平衡——多空之间的未平仓偏斜——按一个钳制带计算资金费率,并在每个 8h 边界为该市场上持有库存的每个账户推进 fundingAccrued。付款方是拥挤方:多头在未平仓中占优时,多头付空头;空头占优时,空头付多头。费率钳制使得单一窗口不会在瞬时偏斜激增时产生病态数字。

账户在单一 8h 窗口内支付的净资金费率超过 $1K 部分,1% 管理员费用进入协议拥有的储备;较小窗口按总额通过。这一切片的存在,让大部分载费仍是订单簿两侧交易者之间的点对点转账,而集中且额度过大的付款者向吸纳其压力的保险与国库层贡献小份额。由于资金费率是运行时状态推进而非报告产物,账户净值在 8h 边界关闭时即反映载费——费率发布与扣除之间没有滞后。

#手续费如何变为协议拥有的价值

maker 与 taker 手续费在成交被接受的那一刻产生。0.020% maker / 0.060% taker 费率在与移动持仓规模、入场基差、已实现 PnL 与账户现金相同的状态转换内对名义价值应用——不存在可能与成交不一致的独立"手续费记账"步骤。挂单 maker 贡献相对于穿越它的 taker 获得 -0.005% 返利;如果你在已发布档位之一上质押 DXTR,-10% / -20% / -30% taker 折扣在出纳对账时落地,而非在每一笔单独成交上。

稍后的对账路径将引擎的手续费、资金费率管理员费用与清算费用总额转化为链上移动:协议拥有的手续费余额、保险补充与国库分配。撮合引擎与财务层之间的分割是刻意的。引擎拥有载费与手续费何时如何产生;国库路径拥有这些协议拥有的余额稍后如何结算进金库。两层看到相同的数字,因为两层都从相同的有序状态读取。

 交易执行
   -> taker 支付名义价值 0.060%,maker 在挂单贡献上返利 -0.005%
   -> 基于簿不平衡应用偏斜调整 taker 附加费
   -> fundingAccrued 在下一个 8h 边界推进;单窗口超过 $1K 部分 1% 管理员切片
   -> DXTR 质押档位折扣每晚在出纳处应用
   -> 对账器将协议增量转化为金库手续费、保险与国库移动

#Why carry belongs in the engine

If funding, the skew-adjusted taker surcharge, and the liquidation fee were treated as reporting artifacts settled out of band, a venue could publish a clean dashboard while its actual carry drifted underneath. Dexter writes all three into the same ordered state advance as the underlying fill, so the market's visible posture, an account's equity, and protocol-owned reserve balances always derive from the same transition. The funded trader sees the same fee an auditor reconstructs from the chain — and the rate that hits the leaderboard is the rate that hit the wallet.

#What this costs a funded trader

• Fee schedule. 0.020% maker / 0.060% taker on notional. A $50K round-trip at taker on both sides costs $60. The fee lands inside the same fill transition, so a position that goes through and reverses inside one block pays both legs without any out-of-band reconciliation.
• Funding cadence. Every 8 hours per market. The rate publishes ahead of the boundary; the deduction lands on fundingAccrued at the boundary itself. Above $1K of funding paid in a single window, a 1% admin cut routes to insurance — smaller windows pass through gross.
• Skew-adjusted taker. The published 0.060% rate is the floor. A long taker into a 90/10 long-heavy book pays an explicit surcharge on top; a long taker that flattens that skew pays at or near the floor. Fighting the skew compounds: higher taker, higher funding paid by the crowded side, lower fill quality at the inventory cap.
• Maker rebate. Resting maker contribution earns -0.005% on the notional it crossed. On a thin attempt, designing entries that rest rather than take saves 0.065% per round-trip — a meaningful edge for higher-frequency strategies running into the same +10% target.
• DXTR stake discount. Staked DXTR moves the taker rate down by 10% / 20% / 30% at the 1K / 5K / 25K tiers. The discount is computed nightly from the on-chain stake snapshot and applied at the cashier reconciliation — it is not retroactive on intraday fills, but compounds across the 30-day window.
• Liquidation fee. A forced close attaches an explicit liquidation fee on top of the realized loss. The fee splits 50% to the keeper that triggered the close and 50% to insurance after any bad-debt offset. Size positions so liquidation fee + realized loss never crosses the -4% daily or -8% total drawdown floor — once liquidated, the rule break is automatic.

Concrete cost model: a 5x notional trader doing 4 round-trips per day on a $10K allocation pays roughly $24/day in taker fees ($10K × 5 × 4 trips × 0.0006 × 2 sides). Across 30 days that is about $720 of pure fee drag — 7.2% of allocation — before funding, before skew surcharges, and before any liquidation fee. A 1K DXTR stake brings that to $648, a 5K stake to $576, a 25K stake to $504. The +10% target assumes those costs are inside the strategy, not stacked on top of it; the discount tiers are how a high-frequency challenge run claws back its edge.