1 · 套保 Hedging（矿山卖出套保 Producer Hedge）
目的：不是赚钱，是消灭价格风险、锁定生产利润。核心公式只有一个：
套保后实现价格 = 建仓期货价 F₀ + 平仓日基差 B_T
所以套保是用价格风险换基差风险。
Goal: not profit — kill price risk and lock the production margin. There is exactly one formula:
Realized price = F₀ (entry futures) + B_T (final basis)
Hedging swaps price risk for basis risk.
建仓 Day 0 · Setup
山西矿山已产出 1,000 吨焦煤，生产成本 1,200，两个月后才有买家提货。今日 Spot 1,600，JM09 期货 1,550，基差 +50。操作：卖出 1,000 吨期货 @1,550，煤原地放在矿山煤场，继续持有现货。
A Shanxi mine has produced 1,000 t at cost 1,200; the buyer collects in two months. Today: Spot 1,600, JM09 futures 1,550, basis +50. Action: short 1,000 t futures @1,550; the coal stays in the mine's yard, still owned.
先定义"实现价" · First, what "realized price" actually means
"实现价"不是任何市场上的报价，而是一个会计结果：平仓那天，把两条腿的现金流加总，折算成"我这批煤每吨实际到手多少钱"。平仓日两笔现金流：①现货卖煤收到 S_T；②期货空头平仓盈亏 F₀ − F_T（1,550 开空、F_T 买回）。
"Realized price" is not a quote on any market — it is an accounting result: on exit day, sum the cash flows of both legs and express them as "how much did I actually pocket per ton of coal." Two cash flows on exit day: ① physical sale brings in S_T; ② the futures short, opened at 1,550 and bought back at F_T, pays F₀ − F_T.
实现价 Realized = S_T + (F₀ − F_T) = F₀ + (S_T − F_T) = F₀ + B_T
用情景 A1 的数字对账：现货卖 1,400 进账 + 期货赚 200 进账 = 每吨到手 1,600 —— "就好像"以 1,600 卖出了这批煤，虽然 1,600 这个价格当天在任何市场上都不存在。这就是"实现价"三个字的全部含义。
Check it with Scenario A1's numbers: 1,400 cash from the coal + 200 cash from futures = 1,600 pocketed per ton — "as if" the coal had been sold at 1,600, even though no market quoted 1,600 that day. That is the entire meaning of "realized price."
这个恒等式把实现价拆成两块，风险属性完全不同：F₀ = 1,550 在 Day 0 下单那一刻就永久锁死（它是你自己的成交价，不再随市场变动）；不确定的只剩 B_T。所以在 Day 0 你就能推算：若基差不变（B_T = B₀ = +50），实现价必然 = 1,550 + 50 = 1,600，恰好是今天的现货价 S₀——这不是巧合，因为 F₀ + B₀ ≡ S₀ 是恒等式。"锁定今天的 1,600"是有条件的（条件：基差不变）；无条件锁定的只有 F₀ 那 1,550。
The identity splits realized price into two parts with completely different risk: F₀ = 1,550 is permanently locked the instant you trade on Day 0 (it is your own fill price — the market can never change it); the only uncertain piece is B_T. So on Day 0 you can already deduce: if basis is unchanged (B_T = B₀ = +50), realized must be 1,550 + 50 = 1,600 — exactly today's spot S₀. No coincidence: F₀ + B₀ ≡ S₀ is an identity. "Locking in today's 1,600" is conditional (condition: basis unchanged); the only thing locked unconditionally is the F₀ of 1,550.
平仓 Day 60 · 实物怎么走 Physical execution flow
D0 · 煤场 Yard1,000 t 躺在矿山煤场
Coal sits at the mine
→
D0 · 期货 Futures卖出 @1,550（只付保证金）
Short @1,550 (margin only)
→
D60 · 卖现货 Sell spot按当日现货价卖给钢厂，卡车拉走
Sell to a mill at that day's spot; trucks haul it away
→
D60 · 同日平期货 Buy back futures按当日期货价买回平仓
Offset at that day's futures price
→
结果 Result煤：矿山→钢厂。交易所全程没见过这批煤。
Coal: mine → mill. The exchange never saw it.
情景分析 Scenario analysis（¥/t，Day 60 结算）
| 情景 Scenario | Spot₆₀ | Fut₆₀ | 现货卖价 Phys sale | 期货盈亏 Fut PnL | 实现价 Realized | B_T |
|---|
| A1 · 同跌200（基差不变）Both −200 | 1,400 | 1,350 | 1,400 | +200 | 1,600 | +50 |
| A2 · 同涨300（基差不变）Both +300 | 1,900 | 1,850 | 1,900 | −300 | 1,600 | +50 |
| B · 现货跌200、期货跌100（基差走弱） | 1,400 | 1,450 | 1,400 | +100 | 1,500 | −50 |
| C · 现货涨200、期货涨300（基差走弱） | 1,800 | 1,850 | 1,800 | −300 | 1,500 | −50 |
| D · 现货涨300、期货涨200（基差走强） | 1,900 | 1,750 | 1,900 | −200 | 1,700 | +150 |
看 B 和 C：一个大跌、一个大涨，方向完全相反，实现价却都是 1,500——因为最终基差都是 −50。这就是"套保后方向不再重要，只剩基差重要"。验证公式：实现价 = 1,550 + B_T，每一行都成立。
Look at B vs C: one crashing market, one rallying market — opposite directions, yet the realized price is 1,500 in both, because final basis is −50 in both. After hedging, direction is irrelevant; only basis matters. Check: Realized = 1,550 + B_T holds on every row.
D60 是随便选的 · The exit day is arbitrary — timing only changes basis risk
D60 不是期货到期日，只是"煤卖掉的那一天"。平仓日 = 现货成交日：哪天把煤卖给钢厂，就在哪天把期货空单买回，两条腿同日退出。期货任何交易日都可以反向平仓，完全不必等到期。但选哪天退出，决定了你承担多少基差不确定性：
D60 is not the futures expiry — it is simply "the day the coal gets sold." Exit day = physical sale day: whichever day you sell the coal to the mill, that same day you buy back the futures short; both legs exit together. Futures can be offset on any trading day — no need to wait for expiry. But which day you exit determines how much basis uncertainty you carry:
离到期越远退出 → B_T 是市场随机给的 Exit far from expiry → B_T is whatever the market gives
贴着到期日退出 → B_T 被收敛压向 ≈ 0 → 实现价 ≈ F₀ = 1,550（提前已知） Exit at expiry → Realized ≈ F₀, known in advance
所以实务规则是：用贴近（等于或略晚于）计划销售日的合约月份来套保。八月底卖煤 → 用 09 合约。退出日离该合约到期越近，B_T 的波动范围越小，实现价越接近事先已知的 F₀。用远月合约对冲近期销售，等于主动多背几个月的基差风险。
Hence the practical rule: hedge with the contract month at or just after your planned sale date. Selling coal in late August → use the September contract. The closer your exit sits to that contract's expiry, the narrower the range of B_T, and the closer the realized price sits to the F₀ you already know. Hedging a near-term sale with a far-month contract means volunteering for extra months of basis risk.
收敛钉死的是"价差"，不是"价格" · Convergence pins the spread, not the level
一个常见困惑："既然到期收敛强制 F = S，那期货价格不是还在波动吗？"——对，两者都一直在波动，收敛管不着价格水平，它只管两者的差。到期前 S 和 F 各自随市场涨跌；交割制度只保证最后一天 S_T − F_T ≈ 0。想象两条船被一根不断收短的绳子拴着：浪照样把它们一起抛上抛下（价格水平在波动），但船距被绳子越拉越近（基差→0）。
A common confusion: "if convergence forces F = S at expiry, isn't the futures price still fluctuating?" — Yes. Both keep fluctuating; convergence says nothing about the price level, only about the gap between the two. Before expiry, S and F each ride the market freely; the delivery mechanism only guarantees that on the final day S_T − F_T ≈ 0. Picture two boats tied by a rope that keeps shortening: the waves still toss them up and down together (the level fluctuates), but the rope pulls the gap to zero (basis → 0).
| 到期日水平 Final level (S_T = F_T) | 现货卖价 Phys sale S_T | 期货盈亏 Fut PnL F₀ − F_T | 实现价 Realized = F₀ + B_T |
|---|
| 市场崩到 Crashed to 1,400 | 1,400 | +150 | 1,550 |
| 横盘在 Flat at 1,600 | 1,600 | −50 | 1,550 |
| 暴涨到 Soared to 1,900 | 1,900 | −350 | 1,550 |
三种到期水平差 500 元，实现价纹丝不动 = F₀ = 1,550。因为水平的波动同时打在两条腿上、正负相消，最后只剩那个已被钉成 0 的价差。这就是"套保者不关心价格波动"的准确含义：波动仍然存在，只是与他无关。至于收敛为什么是"强制"的：到期日若 F > S，任何人都能买现货交割给期货赚差价（第 2 章正套）；若 F < S，持货者卖现货、期货接货赚差价（第 3 章反套）。套利者会吃掉任何缺口，缺口 → 0。
Three final levels spanning ¥500, yet the realized price does not move: F₀ = 1,550 every time. Level fluctuations hit both legs simultaneously and cancel; all that survives is the gap — which convergence has nailed to zero. This is the precise meaning of "a hedger doesn't care about price fluctuation": the fluctuation still exists, it just no longer touches him. And why is convergence forced? If F > S at expiry, anyone can buy spot and deliver into futures for riskless profit (Ch. 2, cash & carry); if F < S, inventory holders sell spot and take delivery (Ch. 3, reverse). Arbitrageurs eat any gap, so the gap → 0.
你问的问题：基差不变时矿山到底赚了什么钱？ Your question: if basis doesn't change, what does the mine actually earn?
赚的是生产利润：实现价 1,600 − 生产成本 1,200 = 400 元/吨 × 1,000 吨 = 40 万。套保不创造这 400，它只是不让市场把这 400 抢走。矿山的"货最终出掉"就在 D60 卖给钢厂那一步（铁律①）——不是通过交割，期货那边纯现金结算。
It earns its production margin: realized 1,600 − cost 1,200 = ¥400/t × 1,000 t = ¥400k. The hedge doesn't create the 400 — it stops the market from taking it away. The coal "exits" in the D60 sale to the mill (Rule ①); the futures side settles purely in cash, no delivery.
镜像：钢厂买入套保 · Mirror image — the steel mill's long hedge
先解决一个概念陷阱：钢厂手里什么货都没有，那它对冲的是什么？答案："三个月后必须买 1,000 吨"这个需求本身，就是一个天然空头（natural short）。虽然仓库是空的，但煤价每涨 1 块，它未来的采购就多花 1 块——敞口和"做空了 1,000 吨煤"一模一样。矿山的天然多头是挖出来的煤，钢厂的天然空头是躲不掉的采购义务；两者都不是自己选的头寸，是生意本身自带的。套保永远是给这个"生来就有"的敞口配一条反向的期货腿：
First, a conceptual trap: the mill owns no coal at all — so what is it hedging? Answer: the obligation "must buy 1,000 t in three months" is itself a natural short position. The warehouse is empty, yet every ¥1 rise in coal costs its future purchase ¥1 more — the exposure is identical to being short 1,000 t. The mine's natural long is the coal it dug up; the mill's natural short is the purchase it cannot avoid. Neither position was chosen — the business itself creates it. Hedging always means attaching an opposite futures leg to this exposure you were born with:
矿山 Mine： 天然多头现货 Natural LONG physical + 卖出期货 SHORT futures
钢厂 Mill： 天然空头现货 Natural SHORT physical（未来必须买 must buy later） + 买入期货 LONG futures
建仓 Day 0：Spot 1,600，JM12 期货 1,550。钢厂买入 1,000 吨期货 @1,550。现货腿什么都不做——本来就没有货，"天然空头"就是那份三个月后的采购需求本身。Day 90 平仓：在现货市场按当日价格正常买煤（煤：贸易商/矿山 → 钢厂），同日把期货多单卖出平掉。和矿山一样：实物走现货市场，期货纯现金了结（铁律①）。
Day 0 setup: Spot 1,600, JM12 futures 1,550. The mill goes long 1,000 t futures @1,550. On the physical side it does nothing — there is no coal to touch; the "natural short" is simply the purchase requirement itself. Day 90 exit: buy coal normally in the spot market at that day's price (coal: trader/mine → mill), and sell the futures long the same day. Exactly like the mine: physical moves through the spot market, futures settle purely in cash (Rule ①).
实际采购成本 Effective cost = S_T − (F_T − F₀) = F₀ + (S_T − F_T) = F₀ + B_T （同一个公式 same formula!）
| 情景 Scenario | Spot₉₀ | Fut₉₀ | 现货买价 Buy coal @ | 期货盈亏 Fut PnL | 实际成本 Effective cost | B_T |
|---|
| A1 · 同涨200（基差不变）Both +200 | 1,800 | 1,750 | −1,800 | +200 | 1,600 | +50 |
| A2 · 同跌200（基差不变）Both −200 | 1,400 | 1,350 | −1,400 | −200 | 1,600 | +50 |
| B · 基差走强 Basis strengthens (+300/+200) | 1,900 | 1,750 | −1,900 | +200 | 1,700 | +150 |
| C · 基差走弱 Basis weakens (+100/+200) | 1,700 | 1,750 | −1,700 | +200 | 1,500 | −50 |
对照矿山的表格逐行看：公式一模一样（成本 = F₀ + B_T），输赢方向正好反过来。A1 里煤涨了 200，钢厂现货多花 200、期货赚回 200，成本仍是 1,600——套保替它挡住了涨价。基差走强（情景 B，即矿山的情景 D）钢厂多花 100：让矿山高兴的基差走强，恰好就是让钢厂难受的基差走强，同一个数字，两端感受相反。这也是为什么研报里常说的"套保盘压制盘面"主要指矿山：现货一涨，矿山排队卖期货锁利润，卖压把盘面摁住；钢厂的买套保是买盘，方向相反但在那轮行情里力量小得多。
Read this table against the mine's table row by row: identical formula (cost = F₀ + B_T), winners and losers exactly reversed. In A1 coal rose 200: the mill pays 200 more for physical, earns 200 back on futures — cost still 1,600; the hedge absorbed the price rise. When basis strengthens (Scenario B here = the mine's Scenario D) the mill pays 100 extra: the same basis strengthening that delights the mine is precisely what hurts the mill — one number, opposite experiences at each end. This is also why the research-report phrase "hedging pressure caps the board" refers mainly to mines: when spot rallies, mines queue up to sell futures and lock margins, and that selling pins the board; mills' long hedges are buying flow — opposite direction, but far smaller in that particular episode.
数学之外：套保的操作现实 · Beyond the math — the operational realities of hedging
你的直觉是对的：套保的数学只有一行公式，真正的难度全在运营层面。以下五点是实务中让套保做砸的地方，每一条都不需要新数学：
Your instinct is right: the math of hedging is one line; all the real difficulty is operational. These five are where hedges actually blow up in practice — none of them needs new math:
① 现金流不对称——套保最大的隐形杀手。两条腿的盈亏结算节奏完全不同：期货亏损是每天都要交的真现金（保证金追缴），现货盈利却要等到几周后煤真卖掉才兑现。情景 D 里矿山"净赚 100"，但过程是：先被期货追缴 200/吨 = 20 万现金今天就要打进期货账户，那 300/吨的现货浮盈躺在煤堆里一分钱掏不出来。方向完全正确的套保，可以在中途把公司的流动性拖死——1993 年德国金属公司（Metallgesellschaft）就是这样亏掉十几亿美元的：套保逻辑对，现金流先断。所以套保规模必须以"最坏情景下追保现金付得起"为上限，而不是以敞口为上限。
① Cash-flow asymmetry — hedging's biggest silent killer. The two legs settle on completely different clocks: futures losses are real cash due daily (margin calls), while the physical gain only becomes cash weeks later when the coal actually sells. In Scenario D the mine "nets +100" — but the path is: pay 200/t of margin calls first = ¥200k of cash into the futures account today, while the 300/t physical gain sits in a coal pile, untouchable. A perfectly correct hedge can strangle a company's liquidity mid-journey — this is how Metallgesellschaft lost over a billion dollars in 1993: right hedge, cash ran out first. Size the hedge to the worst-case margin call you can fund, not to the exposure.
② "期货亏了"不等于套保失败。按设计，期货腿单独看大约一半时间是亏的（价格上涨时）——那正是现货赚钱的时候。企业内部最经典的灾难：老板只看期货账户对账单，把情景 D 的"期货亏 20 万"骂成交易员失职，于是下一轮不套了，然后价格崩盘。评价套保只有一个合法口径：两条腿加总后的实现价，对比当初要锁定的目标价。单看任何一条腿都是耍流氓。
② "The futures lost money" ≠ the hedge failed. By design, the futures leg alone loses roughly half the time (whenever prices rise) — which is exactly when the physical is winning. The classic internal disaster: the boss reads only the futures statement, scolds the trader for Scenario D's "¥200k futures loss," the next cycle goes unhedged, and then the market crashes. The only legitimate scorecard is the combined realized price of both legs versus the target you set out to lock. Judging either leg alone is meaningless.
③ 套保是保险，保费是放弃的上涨。锁定 1,600 意味着涨到 1,900 也只拿 1,600。牛市里没套保的同行看起来像天才，董事会会问"我们为什么少赚了 300"。答案必须在做之前就写进制度：我们锁的是生产利润 400/吨，锁定之后价格波动与我们无关——这是商业模式的选择，不是每次行情的赌注。没有这份事前共识，套保制度撑不过第一轮牛市。
③ A hedge is insurance, and the premium is the upside you surrender. Locking 1,600 means collecting 1,600 even when the market hits 1,900. In a bull market the unhedged competitor looks like a genius, and the board asks "why did we leave 300 on the table?" The answer must be written into policy before the trade: we lock a 400/t production margin, and once locked, price moves are none of our business — a business-model choice, not a per-rally bet. Without that pre-agreed charter, no hedging program survives its first bull market.
④ 套多少：对冲比例与"超套即投机"。矿山计划产 1,000 吨就最多套 1,000 吨；套 1,500 吨，多出的 500 吨没有现货腿对应，就是裸空投机——历史上多家企业的"套保巨亏"其实都是超套。反过来，产量本身不确定（矿难、限产）时要打折套（如套 80%），否则减产 + 价格上涨会让你变成事实上的超套。比例的锚永远是"确定会发生的现货量"。
④ How much to hedge: the hedge ratio, and "over-hedging = speculation." A mine planning 1,000 t of output hedges at most 1,000 t; hedge 1,500 t and the extra 500 t has no physical leg — it is a naked short. Most corporate "hedging disasters" in history were actually over-hedges. Conversely, when output itself is uncertain (accidents, production curbs), hedge at a haircut (say 80%), or a production shortfall plus a rally turns you into a de-facto over-hedger. The ratio is always anchored to the physical volume that will certainly happen.
⑤ 交易所不是中性管道。涨跌停板会让你想平仓时平不掉（连续锁死几个板，追保照收）；持仓限额限制大企业的套保规模（需申请套保额度）；行情过热时交易所临时提高保证金、扩大涨跌停，直接改变你的现金流压力；主力合约不连续（黑色系只有 1/5/9 月活跃）迫使你必然承担第 5 章的移仓成本。这些规则风险和价格风险一样真实。
⑤ The exchange is not a neutral pipe. Price limits can lock you out of exiting for days (limit-down after limit-down — margin calls still due); position limits cap large hedgers' size (hedge quotas must be applied for); in overheated markets the exchange raises margins and widens limits overnight, directly changing your cash-flow burden; and discontinuous liquidity (only Jan/May/Sep are active in Chinese ferrous futures) makes Chapter 5's roll costs unavoidable. These rule risks are as real as price risk.
2 · 正套 Cash & Carry（Contango 时：买现货 + 卖期货 + 真能交割）
赌的不是涨跌，是"基差到期必收敛为零"——因为到期日期货=现货（交割强制画等号）。只要期货升水 > 持有成本，利润在 Day 0 就锁死了。
The bet is not direction — it's that basis must converge to zero at expiry, because delivery forces Futures = Spot on the last day. If the contango exceeds the cost of carry, the profit is locked on Day 0.
建仓 Day 0 · Setup
Spot 1,500，两个月后到期的期货 1,600（Contango，基差 −100）。两个月持有成本：仓储 20 + 资金利息 25 + 交割检验杂费 15 = 60。锁定毛利 100 − 60 = 净赚 40/吨，与未来价格无关。操作：买入 1,000 吨现货 @1,500 + 卖出期货 @1,600。
Spot 1,500; futures expiring in two months 1,600 (contango, basis −100). Two-month carry: storage 20 + financing 25 + delivery/inspection fees 15 = 60. Locked profit: 100 − 60 = ¥40/t regardless of where prices go. Action: buy 1,000 t spot @1,500 + short futures @1,600.
实物怎么走（这是唯一"煤真进交易所"的策略之一）· Physical flow — one of the only strategies where coal actually enters the exchange system
① 港口买煤 Buy at port@1,500，运往交割库
Truck/rail to delivery warehouse
→
② 入交割库 Enter warehouse交易所指定库，质检合格
Exchange-approved, passes inspection
→
③ 注册仓单 Register warrant煤变成一张电子仓单
Coal becomes an e-warrant
→
④ 到期交割 Deliver at expiry仓单划给期货买方，收 1,600 货款
Warrant transfers to the long; collect 1,600
→
⑤ 煤没动 Coal didn't move过户瞬间煤还在原库位，新主人日后提货
Same pile, new owner; pickup is his problem
情景分析 Scenarios（¥/t）
| 情景 Scenario | 到期价 Final (S=F) | 现货腿 Phys | 期货腿 Fut | 成本 Carry | 净利 Net |
|---|
| 市场暴跌 Market crashes | 1,400 | −100 | +200 | −60 | +40 |
| 市场暴涨 Market rallies | 1,750 | +250 | −150 | −60 | +40 |
| 提前解锁 Early unwind (D30)：S 1,580 / F 1,615，基差收敛到 −35 | — | +80 | −15 | −30 | +35 |
前两行：无论到期是 1,400 还是 1,750，净利永远是 +40——价格水平被完全对冲，赚的只是"基差 −100 → 0"的收敛。第三行是现实中 95% 的做法（铁律②）：基差收敛到一半（−100→−35）就提前平掉——卖掉现货、买回期货，捕获 65 的收敛、省下后半程成本，煤根本不用注册仓单。只有当没人接现货、收敛不发生时，才走到底交割——交割能力是让基差收敛的威慑，不是目的。
Rows 1–2: whether the market ends at 1,400 or 1,750, net is always +40 — price level is fully hedged; the profit is purely the −100 → 0 basis convergence. Row 3 is what happens 95% of the time (Rule ②): once basis converges halfway (−100 → −35), unwind early — sell the coal spot, buy back the futures, pocket 65 of convergence, save the remaining carry; the coal never needs a warrant. You only ride to actual delivery if convergence refuses to happen — the ability to deliver is the threat that forces convergence, not the goal.
3 · 反套 Reverse Cash & Carry（Back 时：卖现货 + 买期货，本质是库存置换）
这是你最困惑的一个，先回答你的原话："1600 卖掉现货、1500 买期货，不是白赚 100？之后货没有价格风险吗？"
不是白赚，也没有裸露风险。关键：卖掉现货的那一刻你失去了库存，买入期货则把库存敞口原封不动接了回来（纸面形式）。你赚的 100 是市场付给你的"两个月不持有煤"的租金——即放弃便利收益（convenience yield）的补偿。
This was your biggest confusion, so let's answer your exact words: "sell spot at 1,600, buy futures at 1,500 — isn't that a free 100? And isn't the coal then exposed to price risk?"
It's not free, and there's no naked exposure. The instant you sell the coal you lose the inventory; the long futures re-acquires that exposure one-for-one in paper form. The 100 is rent the market pays you for going two months without holding coal — compensation for surrendering convenience yield.
建仓 Day 0 · Setup（前提：你手里必须真有煤）
你是持有 1,000 吨库存的贸易商。Spot 1,600，期货 1,500（Back，基差 +100）。操作：今天把库存 @1,600 卖给钢厂（煤真的拉走了，仓库空了） + 买入期货 @1,500。交割杂费预算 15。
You are a trader holding 1,000 t. Spot 1,600, futures 1,500 (backwardation, basis +100). Action: sell the inventory @1,600 to a mill today — trucks come, warehouse now empty + buy futures @1,500. Budget 15 for delivery fees.
实物怎么走：煤出去一次、回来一次 · Physical flow: coal leaves once, comes back once
① D0 卖现货 Sell inventory@1,600，煤：你的仓库 → 钢厂。收现金 160 万
Coal: your warehouse → the mill. Cash in: 1.6 M
→
② D0 买期货 Long futures@1,500，敞口从库存变成纸面
Exposure: inventory → paper
→
③ D60 到期接货 Stand for delivery按交割结算付款，收到仓单
Pay ≈1,500 settlement, receive warrant
→
④ 提货 Pick up从交割库把煤拉回自己仓库
Haul coal from delivery warehouse back home
→
⑤ 复位 Reset库存回到 1,000 吨，现金多了 ≈85/吨
Inventory restored; ≈¥85/t richer
情景分析 Scenarios（¥/t）
| 情景 Scenario | 到期价 Final (S=F) | 卖现货 Sold @ | 买回成本 Reacquire via delivery | 费用 Fees | 相对"傻拿着" vs just holding |
|---|
| Back 收敛于中间 Converges mid | 1,550 | 1,600 | −1,500 | −15 | +85 |
| 价格崩了 Market collapses | 1,350 | 1,600 | −1,500 | −15 | +85 |
| 价格暴涨 Market soars | 1,800 | 1,600 | −1,500 | −15 | +85 |
| 提前解锁 D20：Back 收窄至 +30（S 1,650 / F 1,620） | — | 1,600（少赚50） | 期货 +120 | 0 | +70 |
注意基准：正确的比较对象不是"什么都没有"，而是"傻拿着库存不动"。无论到期 1,350 还是 1,800，反套都让你比傻拿着多赚 85——因为 1,600 卖、经交割 1,500 买回，中间的 100（扣费 85）在 Day 0 就锁死了。价格暴涨你也不吃亏：现货"卖飞"少赚的，期货多单一分不少地补回来。真正的风险是：①这两个月里有客户急要煤，你仓库是空的（便利收益没了）；②接回来的仓单可能是符合交割标准的蒙煤，而你卖出去的是山西主焦煤——质量基差风险；③交割前 Back 先走扩（如 D20 现货冲到 1,700），账面难受、可能被追保。
Mind the benchmark: the right comparison is not "having nothing" but "dumbly holding the inventory". Whether the market ends at 1,350 or 1,800, the reverse arb beats holding by 85 — sold at 1,600, reacquired via delivery at 1,500, the 100 spread (85 after fees) was locked on Day 0. A rally doesn't hurt you: what the physical sale "missed", the long futures earns back exactly. The real risks: ① a customer needs coal urgently during those two months and your warehouse is empty (lost convenience yield); ② the warrant you receive may be delivery-grade Mongolian coal while you sold Shanxi prime — quality basis risk; ③ the back can widen first (spot spikes to 1,700 at D20), causing painful mark-to-market and margin calls before convergence.
为什么 Back 难以维持，但也不会瞬间消失 · Why backwardation erodes but doesn't vanish instantly
每个持有库存的人看到 +100 的 Back 都想做上面这套：卖现货（压现货价）＋买期货（抬期货价），基差被两头挤压——这就是研报说的"反套资金令 Back 难以维持"。但反套只有持货者能做——煤没法像股票那样融券做空，所以反套力量有天花板；当现货真的极度紧缺（山西矿难）、没人舍得卖库存时，Back 就能顽固地存在一阵子。
Everyone holding inventory who sees a +100 back wants to run this trade: selling spot (pushes spot down) + buying futures (pushes futures up) squeezes the basis from both ends — this is the report's "reverse-arb money makes the back hard to sustain." But only inventory holders can do it — you can't borrow coal to short it like a stock — so reverse-arb firepower is capped. When spot is genuinely scarce (Shanxi mine accidents) and nobody will part with inventory, the back can stubbornly persist for a while.
4 · 基差交易 Basis Trading（期现套利，赌基差变化方向，不赌收敛）
和正套/反套的区别一句话说清：正套/反套赌收敛（到期必然发生），基差交易赌走向（可能对可能错）。做多基差 Long Basis = 买现货 + 卖期货，赌基差走强；做空基差反之。这是国内"期现商"（蒙煤贸易商）的日常。
One sentence separates this from carry arb: carry arb bets on convergence (guaranteed at expiry); basis trading bets on the direction of basis change (can be wrong). Long basis = buy physical + short futures, betting basis strengthens; short basis is the mirror. This is the daily bread of Chinese cash-futures merchants (e.g., Mongolian coal traders).
建仓 Day 0 · Setup
贸易商在甘其毛都口岸买入 1,000 吨蒙 5 精煤 @1,500，同时卖出 JM09 @1,550。基差 = −50。判断：矿难导致现货将领涨，基差会走强。持有成本约 15/月。
A trader at the Ganqimaodu border buys 1,000 t Mongolian #5 coal @1,500 and shorts JM09 @1,550. Basis = −50. Thesis: mine accidents will make spot lead the rally, so basis strengthens. Carry ≈ 15/month.
实物怎么走 · Physical flow — 注意"解期现"这个出口 The exit is everything
① 口岸买煤 Buy at border@1,500，入库堆存，占用资金/授信
Into a bonded yard; ties up capital
→
② 同时卖期货 Short futures@1,550，锁住 flat price
Flat price locked
→
③ 等基差走强 Wait煤在堆场躺着，每天付仓储利息
Coal sits; storage & interest tick daily
→
④ 解期现 Unwind把煤卖给下游钢厂/焦化厂
Sell the coal to a downstream mill
→
⑤ 同日平期货 Lift the short买回期货，头寸清零
Buy back futures; book flat
情景分析 Scenarios（¥/t，持有 1 个月，成本 15 已计入）
| 情景 Scenario | Spot | Fut | 现货腿 Phys | 期货腿 Fut | 基差 Basis | 净利 Net |
|---|
| ① 同涨100 Both +100 | 1,600 | 1,650 | +100 | −100 | −50 → −50 | −15 |
| ② 现货领涨 Spot leads (+200/+100) | 1,700 | 1,650 | +200 | −100 | −50 → +50 | +85 |
| ③ 期货领涨 Futures lead (+100/+200) | 1,600 | 1,750 | +100 | −200 | −50 → −150 | −115 |
| ④ 双双下跌但现货抗跌 (−100/−200) | 1,400 | 1,350 | −100 | +200 | −50 → +50 | +85 |
②和④揭示本质：一个涨市、一个跌市，只要基差同样走强 100，净利同样是 +85。基差交易完全方向中性。情景③就是研报里蒙煤贸易商的噩梦——"期现解不出来"：期货涨得比手里的蒙煤快，基差走弱到 −150，此刻解期现（卖煤+平空）就是锁定 −115 的亏损；不解，仓储利息每天流血，还赌不到基差回来。这就是"现货越强、盘面难有起色"背离行情里被夹住的那批人。
Rows ② and ④ reveal the essence: one rising market, one falling market — but the basis strengthened by 100 in both, so both net +85. Basis trading is fully direction-neutral. Row ③ is the Mongolian trader's nightmare from the report — "can't unwind the cash-futures position": futures rose faster than his Mongolian coal, basis weakened to −150, and unwinding now (sell coal + lift short) locks in −115; not unwinding means bleeding carry daily while praying basis recovers. These are the people trapped inside the "spot ever stronger, board going nowhere" divergence.
基差为什么会走强、走弱 · What actually moves the basis — and where the profit comes from
先记住一个不对称：现货价反映"今天真实的吨数"，期货价反映"市场对未来的预期"。期货是快钱，消息一出秒动；现货是慢钱，要等真实的货、真实的买家跟上。所以基差变动的本质几乎总是：现实与预期赛跑，谁跑得快。做多基差 = 赌现实追上/超过预期；做空基差 = 赌预期跑过头了会回来。
First, one asymmetry to memorize: spot reflects "real tonnes today"; futures reflect "the market's expectation of the future." Futures are fast money — they jump the second news breaks; spot is slow money — it waits for real cargoes and real buyers. So a basis move is almost always the same race: reality vs expectation — which one runs faster. Long basis = betting reality catches up with or overtakes expectation; short basis = betting expectation has overrun and will come back.
| 方向 Direction | 驱动因素 Driver | 机制 Mechanism |
|---|
基差走强 Strengthens
（现货领跑 spot leads） | 现货突发紧缺 Sudden physical tightness | 矿难、安监、口岸关闭：今天的煤真的少了，但期货因交割品宽松（蒙煤）跟不动 Mine accidents, safety inspections, border closures: today's coal is genuinely scarce, but futures lag because the deliverable basket (Mongolian coal) stays loose |
| 下游被迫补库 Forced restocking | 钢厂库存见底必须现在买，现货有真实买盘、期货没有 Mills at rock-bottom inventory must buy now — real bids hit spot, not paper |
| 套保卖压封住盘面 Hedge selling caps futures | 第 1、5 章：现货越涨矿山越卖期货，期货被人为摁住 Ch. 1 & 5: the more spot rallies, the more mines sell futures — the board is artificially pinned |
| 临近交割的收敛 Convergence into expiry | 若基差为负（Contango），到期日基差必然向 0 走强——这是唯一"必然"的驱动 If basis is negative, expiry forces it up toward 0 — the only guaranteed driver |
基差走弱 Weakens
（期货领跑 futures lead） | 预期先行 Expectations run ahead | 宏观刺激、限产传闻：政策消息先打进期货，现货要等真实成交跟进 Stimulus or production-cut rumors hit futures instantly; spot waits for actual deals |
| 供应宽松兑现 Supply loosening materializes | 进口放量、复产：现货有真实卖压，期货早已计价 Imports surge, mines restart: real selling hits spot while futures priced it in long ago |
| 反套资金入场 Reverse-arb money arrives | 第 3 章：Back 太深 → 持货者卖现货买期货，两头挤压基差 Ch. 3: back too deep → inventory holders sell spot / buy futures, squeezing from both ends |
| 临近交割的收敛 Convergence into expiry | 若基差为正（Back），到期日基差必然向 0 走弱 If basis is positive, expiry forces it down toward 0 |
利润从哪来？两种完全不同的钱：①判断的钱——比市场更早看出上表哪个驱动正在发生（如：口岸数据显示蒙煤通关暴增 → 做空基差）；②渠道的钱——期现商真正的护城河：他能以比市场便宜 20 的价格从口岸拿到货、能把货卖给别人够不着的下游，这段"渠道价差"叠加在基差变动之上，别人复制不了。研报里的期现商亏钱，往往不是渠道没了，而是①的判断被打脸（情景③）。
Where does the profit come from? Two entirely different kinds of money: ① judgment money — spotting a driver in the table before the market does (e.g., border data shows Mongolian customs clearance surging → short basis); ② channel money — the basis merchant's real moat: he can source at the border 20 below market and sell to downstream buyers others can't reach; that channel edge stacks on top of the basis move and cannot be copied. When the report's Mongolian traders lose, it's usually not the channel failing — it's judgment ① being run over (Row ③).
你问的另一个问题：两头都用 Benchmark + 固定升贴水，再用期货对冲，是不是就零风险了？ · Your other question: benchmark-priced purchase AND sale plus a futures hedge — zero risk?
不是。例：Trafigura 采购 CCI+20、销售 CCI+50，账面锁定 30。但①定价日不同（采购月底点价、销售下月底点价，中间 CCI 波动 100 就是裸敞口）→ 所以还要用期货盖住时间差；②期货标的是 DCE 焦煤，不是 CCI → 对冲后剩下 Benchmark 基差风险；③还有质量、港口、运费、汇率、违约风险。套保消灭的只是 flat price 这一层，剩下的十几种残余风险才是贸易商每天真正管理的东西——这也正是他们赚 30 块/吨毛利的理由。
No. Example: Trafigura buys at CCI+20, sells at CCI+50 — 30 locked on paper. But ① the pricing dates differ (purchase prices month-end, sale a month later; CCI moving 100 in between is naked exposure) → futures are used to bridge the timing gap; ② the futures contract is DCE coking coal, not CCI → hedging leaves benchmark basis risk; ③ plus quality, port, freight, FX, and counterparty risk. A hedge kills only the flat-price layer; the dozen residual risks are what a merchant actually manages daily — and are precisely why the ¥30/t margin exists at all.
锁定两条现货腿之后：期现对冲的完整解剖 · After both physical legs are locked — the full anatomy of the hedge
把上面那张卡片展开算一遍。设定：Trafigura 签了采购 CCI+20（7月31日点价）和销售 CCI+50（8月31日点价），各 1,000 吨。注意：+20 和 +50 这两个升贴水是现货腿的基差，已经锁死——你控制得了它们。整笔交易的敞口随时间分三个阶段：
Let's expand that card into a full calculation. Setup: Trafigura signs a purchase at CCI+20 (prices July 31) and a sale at CCI+50 (prices Aug 31), 1,000 t each. Note: the +20 and +50 premia are the physical legs' basis, already locked — those you control. The trade's exposure then lives in three phases:
| 阶段 Phase | 成本 Cost | 收入 Revenue | 价格敞口 Flat-price exposure | 该做什么 Action |
|---|
| ① 签约日 → 7/31 Signing → Jul 31 | 浮动 floats with CCI | 浮动 floats with CCI | 零（两边同浮，互相抵消）Zero — both float, they cancel | 什么都不做。此时卖期货反而会制造敞口 Do nothing — hedging now would create exposure |
| ② 7/31 → 8/31（危险的一个月 the dangerous month） | 已锁 fixed | 浮动 floats | 做多 CCI 一个月 LONG the benchmark for one month | 7/31 点价当刻卖出期货 Sell futures the instant the purchase prices |
| ③ 8/31 之后 After Aug 31 | 已锁 fixed | 已锁 fixed | 零 Zero | 8/31 点价当刻买回期货 Buy back futures the instant the sale prices |
逐笔对账（数字版）。7月31日：CCI 定格 1,000 → 成本锁定 1,020；同日卖出 DCE 期货 @1,050。8月31日：CCI 定格 1,100 → 收入锁定 1,150；同日买回期货 @1,130。结算：现货利润 = 1,150 − 1,020 = +130（承诺的 30 + 敞口月里 CCI 涨的 100）；期货 = 1,050 卖、1,130 买回 = −80。净利 = 50 = 30 + 20。那个 20 是哪来的？CCI 涨了 100，而 DCE 只涨了 80——你的敞口跟着 CCI 走，你的对冲工具跟着 DCE 走，两个指数走不齐的部分就漏出来了。它这次是 +20，下次完全可以是 −20。
The ledger, in numbers. July 31: CCI prints 1,000 → cost locked at 1,020; same day, sell DCE futures @1,050. Aug 31: CCI prints 1,100 → revenue locked at 1,150; same day, buy back futures @1,130. Settlement: physical margin = 1,150 − 1,020 = +130 (the promised 30, plus the 100 CCI rally you were long); futures = sold 1,050, bought back 1,130 = −80. Net = 50 = 30 + 20. Where did the 20 come from? CCI rose 100 while DCE rose only 80 — your exposure tracks CCI, your hedge tracks DCE, and whatever the two indexes don't share leaks through. This time +20; next time it can just as easily be −20.
净利 Net profit = 锁定价差 Locked margin 30 + （ΔCCI − ΔDCE）敞口期内两指数的走差 the two indexes' divergence during the exposure window
| 敞口月情景 Scenario (Jul 31 → Aug 31) | ΔCCI | ΔDCE | 现货利润 Phys | 期货 Fut | 净利 Net |
|---|
| 两指数完全同步 Perfect sync | +100 | +100 | +130 | −100 | 30 |
| 现货指数领涨 CCI leads（本例 this example） | +100 | +80 | +130 | −80 | 50 |
| 期货领涨 DCE leads（蒙煤贸易商剧本 the Mongolian trader’s script） | +100 | +130 | +130 | −130 | 0 |
| 同跌但期货跌更多 Both fall, DCE falls harder | −100 | −150 | −70 | +150 | 80 |
这就是你悟到的那件事，明确写出来：你能锁住现货两端的升贴水（+20、+50），但你对对冲工具那个指数的走势没有任何控制权。CCI 是京唐港现货调研价，反映今天真实成交；DCE 期货价 = 预期 + 交割品篮子（含蒙煤）+ 资金流（第 5 章）。它们是两个不同的东西，凭什么必须同步？所以：只要"合同定价基准"和"对冲工具"不是同一个东西，基差风险就永远存在——套保从来不是消灭风险，而是把一个大的敞口（flat price，动辄 ±200）换成一个小的敞口（两指数走差，通常 ±30 以内）。研报里山西 +280 / 盘面 +208 / 蒙煤 +174 的三轨分化，就是这个"走差"在现实中张开到最大的样子。
Here is your realization, stated explicitly: you can lock the premia on both physical ends (+20, +50), but you have zero control over how the hedge instrument's index moves. CCI is a Jingtang-port physical survey price reflecting today's real deals; the DCE futures price = expectations + the deliverable basket (Mongolian coal included) + fund flows (Ch. 5). They are two different objects — nothing obliges them to move together. Therefore: whenever the contract's pricing benchmark and the hedge instrument are not the same thing, basis risk exists — always. Hedging never eliminates risk; it swaps a large exposure (flat price, routinely ±200) for a small one (the two indexes' divergence, usually within ±30). The report's three-track split — Shanxi +280 / board +208 / Mongolian +174 — is exactly this divergence stretched to its widest in real life.
对冲质量的四个等级。残余风险的大小完全取决于"基准"和"工具"离得多近：
The four grades of hedge quality. The residual risk depends entirely on how close the benchmark sits to the instrument:
| 等级 Grade | 合同基准 Contract benchmark | 对冲工具 Hedge instrument | 残余风险 Residual |
|---|
| ★★★★ 完美 Perfect | 盘面点价，两腿同日定价 DCE-priced, both legs same day | 无需对冲 no hedge needed | 零 Zero — 价差生来就是锁死的 the margin is born locked |
| ★★★ 精确 Exact | 盘面点价，定价日不同 DCE-priced, different dates | DCE 期货，点价当刻成交 DCE futures at the pricing instant | ≈零（同一工具，只剩执行滑点）≈ zero: same instrument, execution slippage only |
| ★★ 交叉 Cross | CCI 等现货指数 spot index like CCI | DCE 期货 DCE futures | ΔCCI − ΔDCE 基准基差风险 benchmark basis risk（本节主角 this section's subject） |
| ★ 裸奔 Naked | 固定价或无基准 fixed price / no benchmark | 无 none | 全部 flat price 风险 full flat-price risk |
两个推论。①这解释了盘面点价合同为什么统治了现代贸易（第 6.1 章）：它让每一本贸易账册都能升级到 ★★★ 精确档。②国际市场对 ★★ 档的解法是找一个按合同基准本身结算的掉期（第 7.2 章：API2 合同用 API2 掉期对冲，走差归零）；国内没有流动的 CCI 掉期，所以国内的贸易商们只能停在 ★★ 档，靠盈亏归因（第 8.2 章）天天盯着这条走差。另注意：点价若是月均价，单笔对冲复刻不了平均值，要在整个窗口内每天分批卖——这也是掉期天然更合身的原因。
Two corollaries. ① This explains why board-priced (盘面点价) contracts conquered modern trade (Ch. 6.1): they upgrade any merchant book to the ★★★ exact grade. ② The international fix for grade ★★ is a swap settling on the contract's own benchmark (Ch. 7.2: an API2 contract hedged with an API2 swap — divergence collapses to zero); China has no liquid CCI swap, so Chinese merchants stay at ★★ and watch this divergence line daily through PnL explain (Ch. 8.2). Also note: when pricing is a monthly average, one hedge trade can't replicate an average — you sell in daily slices across the window, another reason swaps fit more naturally.
价格层死透之后，还剩什么。就算做到 ★★★ 档、走差为零，这笔"锁定 30"仍然悬在下面这些风险上——注意第一条的恶毒之处：对手方违约的动机，恰好在你的对冲赚最多的时候最强；而对手方一违约，你的平账瞬间变成"裸头寸 + 还挂着的对冲"，套保当场变投机。
What's left once the price layer is fully dead. Even at grade ★★★ with zero divergence, the "locked 30" still hangs on the risks below — and note the malice of the first one: the counterparty's incentive to default is strongest exactly when your hedge is winning most; the moment they walk, your flat book becomes "a naked position with a live hedge attached," and the hedge turns into speculation on the spot.
| 残余风险 Residual risk | 一句话 In one sentence |
|---|
| ① 违约 Counterparty | CCI 涨 200，矿山"检修"不发货；跌 200，钢厂挑质量拒收 CCI +200 and the mine finds "maintenance issues"; −200 and the mill rejects on quality pretexts |
| ② 数量 Volume（溢短装 tolerance） | 合同 ±5% 由对方选择——对冲了 10,000 吨，来了 9,300 吨，700 吨裸奔，且对方永远选对你不利的方向 ±5% at the counterparty's option — they always exercise it against you |
| ③ 质量 Quality | 灰分超标触发扣价公式，对冲不会跟着扣；最坏被拒收，煤砸手里 assay penalties the hedge doesn't mirror; worst case, rejected cargo lands unhedged in your lap |
| ④ 时间/QP Timing | 船晚了，点价窗口从 8 月滑到 9 月，对冲挂错月份 → 月差敞口（第 5 章）vessel delay slides the pricing window a month — your hedge sits in the wrong contract |
| ⑤ 运费 Freight / ⑥ 汇率 FX | 第 7.4 章的另外两层皮，各自要各自的纸 the other two layers, each needing its own paper |
| ⑦ 资金 Financing | 付矿山在先、收钢厂在后的利息缺口 + 对冲的每日追保现金（第 1 章①）the payment gap's interest + the hedge's daily margin calls |
| ⑧ 政策 Policy | 口岸关闭、限价令：2021 年 10 月发改委干预，盘面三周腰斩而现货合同冻结，指数本身失真 border closures, price caps: Oct 2021 — futures halved in three weeks while physical contracts froze |
| ⑨ 操作/欺诈 Ops & fraud | 录错吨数、假仓单、重复质押（青岛 2014）——历史上杀死贸易商最多的不是价格，是这一条 wrong entries, fake receipts, double pledging — historically the top killer of trading houses, not price |
本节收尾一句话：那 30 块/吨不是"承担价格风险"的报酬——价格已经被对冲杀死了。它是运营上面这台残余风险机器的工钱：审对手方、写溢短装和质量条款、对齐 QP、排船、融资、盯走差。对冲基金赚不到这 30，因为它没有卡车、没有授信关系——这就是护城河。
This section in one sentence: the ¥30/t is not payment for carrying price risk — the hedge killed price. It is the wage for operating the residual-risk machine above: vetting counterparties, drafting tolerance and quality clauses, aligning QPs, chartering, financing, watching the divergence line. A hedge fund can't earn that 30 — it has no trucks and no credit relationships. That is the moat.
5 · 移仓换月 vs 套利资金：谁在拉宽期限结构，谁在压扁它 · Hedge Rolls vs Arbitrage: Who Stretches the Term Structure, Who Squeezes It
期限结构（近月与远月的价差）不只由供需决定，还由两股性格完全相反的资金流塑造：套保者的移仓换月是不看价格、到点必做的机械流，它把结构越推越开；套利资金是只看价格、越界才出手的弹簧，它把结构拉回持有成本的走廊里。搞懂这两股力，你就能看懂研报里几乎所有关于"月差"的段落。
The term structure (the spread between nearby and deferred months) isn't shaped by supply and demand alone, but by two flows with opposite personalities: hedgers' contract rolls are price-blind, calendar-driven mechanical flow that pushes the structure further apart; arbitrage capital is price-sensitive, threshold-triggered — a spring that pulls the structure back inside the cost-of-carry corridor. Understand these two forces and nearly every "calendar spread" paragraph in a research report becomes readable.
5.1 什么是移仓换月，为什么必须做 · What rolling is, and why it's compulsory
矿山的煤是持续生产的，敞口没有到期日；但期货合约有。空单挂在 JM09 上，而 09 快到期了（且到期前流动性会提前枯竭），套保却还得继续——怎么办？移仓：同一天，买回 09 空单 + 在 01 合约上重新卖出。两笔都是期货交易，实物端一克煤都没动（铁律①：没有卡车在跑，只有仓单都没有——纯纸面）。钢厂镜像：卖掉近月多单、买入远月多单。
A mine produces continuously — its exposure has no expiry date; futures contracts do. The short sits in JM09, which is about to expire (and liquidity dries up well before expiry), yet the hedge must live on. Solution: roll — same day, buy back the JM09 short + re-sell in JM01. Both trades are futures trades; not one gram of coal moves (Rule ①: no truck driving, not even a warrant — pure paper). The mill mirrors it: sell the nearby long, buy the deferred long.
移仓不是免费的 —— 滚动成本。例：移仓日 JM09 = 1,500，JM01 = 1,440（Back 60）。矿山 1,500 买回、1,440 重新卖出：新空单比旧空单低了 60。若现货维持 1,500，01 合约会向 1,500 收敛（第 1 章），这张新空单注定亏掉这 60。也就是说：Back 市场里，空头套保者每滚动一次就付一次 Back；Contango 里则每滚一次赚一次升水。钢厂正好反过来。
Rolling isn't free — the roll cost. Example: on roll day JM09 = 1,500, JM01 = 1,440 (back of 60). The mine buys back at 1,500 and re-sells at 1,440: the new short sits 60 lower than the old one. If spot stays at 1,500, JM01 will converge up toward it (Ch. 1), and this new short is destined to lose that 60. In other words: in backwardation, a short hedger pays the back every time he rolls; in contango he collects the premium every roll. The mill is the exact mirror.
| 期限结构 Structure（月差 60 spread of 60） | 矿山滚动空头 Mine rolling shorts | 钢厂滚动多头 Mill rolling longs |
|---|
| Contango（远月贵 60 deferred +60） | 每滚赚 +60 collects 60/roll（高卖远月 re-sells higher） | 每滚付 −60 pays 60/roll（高买远月 re-buys higher） |
| Back（远月便宜 60 deferred −60） | 每滚付 −60 pays 60/roll（低卖远月 re-sells lower） | 每滚赚 +60 collects 60/roll（低买远月 re-buys lower） |
5.2 移仓流：把结构越推越开的力 · Roll flow — the force that stretches the structure
看移仓那两笔单子砸向市场的方向：矿山移仓 = 买近月 + 卖远月 → 近月被买高、远月被卖低 → 月差走向 Back（或 Contango 收窄）。钢厂移仓 = 卖近月 + 买远月 → 推向 Contango。净效果看谁的量大：焦煤市场里矿山的卖套保远大于钢厂的买套保（研报里"套保盘压制盘面"说的就是他们），于是每个移仓窗口，远月合约都挨一轮集中卖压——远月被系统性压低，结构被推向 Back。
Look at the direction the two roll orders hit the market: a mine's roll = buy nearby + sell deferred → nearby bid up, deferred sold down → the spread stretches toward backwardation (or contango narrows). A mill's roll = sell nearby + buy deferred → pushes toward contango. The net effect depends on whose volume dominates: in coking coal, mines' short hedging far outweighs mills' long hedging (they are the research reports' "hedging pressure"), so every roll window the deferred contract absorbs a concentrated wave of selling — deferred months get systematically pressed down, stretching the structure toward back.
这股流最关键的性格：它不问"这个月差合不合理"。敞口在、合约要到期，就必须滚——价差再难看也得成交。所以移仓流可以把结构推到超过公允水平（这正是凯恩斯"正常贴水/normal backwardation"理论的现代版：净空头套保压力使远月系统性低于预期现货价，愿意接盘的投机/套利多头因此获得一份风险溢价作为报酬）。
The defining trait of this flow: it never asks "is this spread fair?" The exposure exists and the contract expires, so the roll must trade — however ugly the spread. Roll flow can therefore push the structure beyond fair value (this is Keynes' "normal backwardation" in modern dress: net short hedging pressure keeps deferred prices systematically below the expected spot price, and the speculators/arbitrageurs willing to take the other side collect that gap as a risk premium).
5.3 套利资金：把结构压回走廊的弹簧 · Arbitrage — the spring that squeezes it back
套利资金性格相反：平时睡觉，价差越过持有成本边界才出手，而且出手方向永远是把结构压扁。Contango 太宽（远月升水 > 全部持有成本）→ 第 2 章正套进场：买近月/现货 + 卖远月 → 近月抬、远月压 → Contango 被压回持有成本以内。Back 太深（现货升水 > 持货的便利收益）→ 第 3 章反套进场：卖现货 + 买期货 → Back 被压窄。
Arbitrage capital has the opposite personality: it sleeps until the spread crosses a carry boundary, and when it acts, it always compresses the structure. Contango too wide (deferred premium > full cost of carry) → Chapter 2's cash & carry wakes up: buy nearby/spot + sell deferred → nearby lifted, deferred pressed → contango squeezed back inside carry. Back too deep (spot premium > the convenience yield of holding) → Chapter 3's reverse arb wakes up: sell spot + buy futures → back squeezed narrower.
无套利走廊 No-arbitrage corridor： −持有成本 Full carry ≤ 月差/基差 Spread ≤ 便利收益 Convenience yield
Contango 一侧 = 硬墙 hard wall（有钱有仓库就能做正套 anyone with cash + warehouse can C&C）
Back 一侧 = 软墙 soft wall（反套必须先有货，煤不能融券 reverse arb needs inventory — coal can't be borrowed）
这就是为什么现实里 Contango 很少显著超过持有成本（越界秒被正套吃掉），而 Back 却能在现货紧张时顽固地深下去：软墙那侧，弹簧的力量受制于"市面上有多少人真的持有可卖库存"。
This is why in practice contango rarely exceeds full carry for long (any breach gets eaten instantly by cash & carry), while backwardation can stay stubbornly deep when spot is tight: on the soft-wall side, the spring's strength is limited by how many players actually hold sellable inventory.
5.4 两股力的拔河 = 你看到的期限结构 · The tug of war = the structure you see on screen
把 5.2 和 5.3 叠起来，就是完整画面：移仓流持续把结构往外推（焦煤里主要推向 Back），套利资金在走廊边界往回拉。盘面上的月差，就是这场拔河的实时比分。回到你的研报："阶段性走出 Back 结构"= 矿难 + 矿山滚动卖压把结构推入 Back；"持续性受反套资金影响难以维持"= 一旦 Back 深到值得持货者卖现货买期货，软墙弹簧就开始回拉；"盘面很难给到大幅升水"= Contango 方向还没等走出来，正套的硬墙和源源不断的套保卖压就把它封死了。一句研报，三股力量，全在这一章里。
Stack 5.2 on 5.3 and you have the full picture: roll flow keeps pushing the structure outward (in coking coal, mainly toward back), and arbitrage capital pulls it back at the corridor's edges. The calendar spread on your screen is the live score of this tug of war. Back to your report: "a phase of backwardation emerged" = mine accidents + mines' rolling sell pressure pushed the structure into back; "its persistence is threatened by reverse-arb money" = once the back got deep enough to reward inventory holders for selling spot and buying futures, the soft-wall spring started pulling; "the board can hardly offer a large premium" = before any contango could even form, the hard wall of cash & carry plus the endless hedge selling sealed it off. One report sentence, three forces — all in this chapter.
6 · 期货的其他用法 · Other Types of Futures Uses
原理层面，期货只做三件原始的事：①转移风险（把不想要的波动卖给愿意承担的人）、②传输价格信息（把全市场的预期压缩成一个公开数字）、③把商品变成"纸"（可秒速买卖、可抵押融资的标准化凭证）。前五章讲的套保、套利、基差交易都是①和③的组合。下面这些用法把②和③也用满了——每一个都是真实世界里天天发生的大生意。
At the level of principle, futures do only three primitive things: ① transfer risk (sell unwanted volatility to someone willing to carry it), ② transmit price information (compress the whole market's expectations into one public number), and ③ turn the commodity into "paper" (a standardized claim that trades in seconds and can be pledged for financing). Chapters 1–5 — hedging, arbitrage, basis trading — are all combinations of ① and ③. The uses below exploit ② and ③ to the full, and every one is a large, everyday real-world business.
6.1 点价：期货成为现货合同的语言 · Basis pricing — futures as the language of physical contracts
这是原理②最大的应用，也是现代大宗贸易合同的主流形态：合同不写死价格，而写 价格 = JM09 盘面价 + 升贴水 30，并约定买方在某个窗口内自选一天"点价"（以那天的盘面价定格）。好处：双方提前锁定的是相对关系（升贴水，即基差），把最难谈的绝对价格外包给期货市场这个公开裁判。卖方签完合同后通常立刻用期货把敞口盖住，等买方点价时再解开——你会发现这就是第 4 章基差贸易的合同化形式。点价制度是"期货价格反过来统治现货贸易"的机制，也是为什么盘面一动，全产业链的现货报价秒动。
This is principle ②'s biggest application and the dominant form of modern bulk-commodity contracts: the contract fixes no price, but states Price = JM09 board price + premium 30, and the buyer may pick any day within a window to "price in" (freeze that day's board price). The beauty: both sides pre-agree only the relative part (the premium — i.e., the basis) and outsource the impossible negotiation — the absolute price — to the public referee that is the futures market. The seller typically hedges the moment the contract is signed and unwinds when the buyer prices — which you'll recognize as Chapter 4's basis trade wearing a contract. Basis pricing is the mechanism by which futures prices come to govern physical trade — and why every physical quote in the chain moves the second the board does.
6.2 虚拟库存与虚拟产能 · Virtual inventory & virtual capacity
钢厂看涨煤价想囤三个月的货，但仓库不够、资金占用巨大、煤还会风化损耗——于是不买煤，买等量期货：价格涨了期货利润正好抵消未来采购涨价，效果如同囤了货，成本只有保证金。这叫虚拟库存。反过来，矿山看跌想"提前卖掉未来半年的产量"，物理上做不到（煤还没挖出来），盘面上卖半年的期货即可——虚拟销售。原理：期货让"持有库存"和"拥有产能"这两件重资产的事，变成一张纸的多空。第 3 章反套里"卖掉真煤、用期货多单接回敞口"就是虚拟库存替代真实库存的实例。
A mill bullish on coal wants three months of stock — but the yard is too small, the capital tie-up huge, and coal degrades in storage. So it buys no coal and goes long the equivalent futures: if prices rise, futures profits offset the pricier future purchases — the effect of a stockpile at the cost of margin only. This is virtual inventory. Symmetrically, a bearish mine wanting to "pre-sell six months of output" can't do it physically (the coal isn't dug yet) — but can sell six months of futures: virtual sales. The principle: futures turn the heavy-asset acts of "holding inventory" and "owning capacity" into a piece of paper. Chapter 3's reverse arb — sell the real coal, hold the exposure back via long futures — is exactly virtual inventory replacing real inventory.
6.3 锁加工利润：对冲的是"差"，不是"价" · Locking processing margins — hedging a spread, not a price
焦化厂的利润 = 焦炭卖价 − 焦煤成本。它可以不理会两个绝对价格，直接买入焦煤期货 JM + 卖出焦炭期货 J，把"盘面焦化利润"一次锁死；钢厂同理：买入 JM + 铁矿 I、卖出螺纹 RB，锁"盘面钢厂利润"（国际市场对应原油的 crack spread）。这在概念上是套保的升维：企业真正的敞口从来不是某个价格，而是买价与卖价之间的差——期货市场允许你把整条利润公式直接搬到盘面上对冲。研报里常说的"盘面利润压缩/扩张"，看的就是这些价差上的资金行为。
A coking plant's margin = coke revenue − coking-coal cost. It can ignore both absolute prices and directly buy coking-coal futures (JM) + sell coke futures (J), locking the "board coking margin" in one stroke; a steel mill likewise: long JM + iron ore (I), short rebar (RB) — the "board steel margin" (internationally, oil's crack spread). Conceptually this is hedging one dimension up: a firm's true exposure is never a price — it is the difference between its buy price and its sell price — and the futures market lets you hedge the entire margin formula directly on the board. When reports discuss "board margins compressing/expanding," they are reading capital flows in exactly these spreads.
6.4 仓单融资：商品变成抵押品 · Warrant financing — the commodity becomes collateral
原理③的纯粹形态。一堆煤很难拿去银行借钱（真伪难验、价值难估、处置困难——"融资骗贷"重灾区）；但注册成交易所标准仓单后，它变成质量已认证、价格公开跳动、随时可通过交割变现的标准资产——银行乐意按仓单价值七八成放款，利率还低。贸易商的经典循环：买煤 → 注册仓单 → 质押融资 → 用融来的钱买下一批煤，同时用期货空单锁住仓单价值（否则煤价跌穿质押率会被银行追保）。期货市场在这里的角色不是交易场所，而是给商品发"身份证"的公证处。
Principle ③ in its purest form. A pile of coal is terrible loan collateral (authenticity unverifiable, value uncertain, disposal painful — the epicenter of commodity-financing fraud); but registered as an exchange standard warrant, it becomes a quality-certified, publicly-marked, instantly-liquidatable standard asset — banks happily lend 70–80% of warrant value at low rates. The merchant's classic cycle: buy coal → register warrant → pledge for financing → buy the next cargo with the proceeds, while shorting futures to lock the warrant's value (otherwise a price drop through the pledge ratio triggers the bank's margin call). Here the futures market acts not as a trading venue but as the notary that issues the commodity an identity card.
6.5 投机：生态系统的对手盘 · Speculation — the ecosystem's counterparty
最后别忘了最简单的用法：纯粹表达观点。投机者没有煤、不想要煤、永远不碰交割——但第 1 章矿山甩出来的风险总得有人接。第 5 章讲过：净空头的套保压力使远月系统性偏低，愿意做多的投机者收取这份风险溢价作为报酬——这是投机在系统中的"工资"，不是运气。没有投机盘，套保者互相之间凑不齐对手方，买卖价差会宽到没法用。研报里的"多头资金""产业空头"博弈叙事，说的就是这两个物种的相互依存。
Finally, the simplest use: pure expression of a view. Speculators own no coal, want no coal, never touch delivery — but the risk Chapter 1's mine sheds must land on someone. As Chapter 5 showed, net short hedging pressure keeps deferred prices systematically low, and the speculators willing to go long collect that risk premium as their wage — payment for service, not luck. Without speculative flow, hedgers can't find enough counterparties among themselves and bid-ask spreads widen beyond use. The "long funds vs industrial shorts" narrative in research reports is precisely the symbiosis of these two species.
| 用法 Use | 动用的原理 Primitive used | 谁在用 Who | 实物动不动 Does coal move? |
|---|
| 套保 Hedging（第 1 章） | ①转移风险 Risk transfer | 矿山、钢厂 Mines, mills | 动，走现货市场 Yes, via spot market |
| 正套/反套 Carry arbs（第 2/3 章） | ①+③（交割强制收敛 delivery enforces convergence） | 有仓储/库存的资金 Capital with warehouses/inventory | 动，可进出交割库 Yes, may enter delivery system |
| 基差交易 Basis trading（第 4 章） | ①+②（现实 vs 预期 reality vs expectation） | 期现商 Cash-futures merchants | 动，口岸→下游 Yes, border → downstream |
| 点价 Basis pricing | ②价格信息 Price information | 整条产业链的贸易合同 Every physical contract in the chain | 动（合同货照常走）Yes (contract cargo flows normally) |
| 虚拟库存/产能 Virtual inventory | ③商品变纸 Commodity → paper | 不想占用仓库/资金的产业方 Firms avoiding storage/capital | 不动 — 这正是意义所在 No — that's the whole point |
| 锁加工利润 Margin locking | ①升维：对冲价差 Risk transfer on a spread | 焦化厂、钢厂 Coking plants, mills | 不动（纯盘面组合 pure paper combo）No |
| 仓单融资 Warrant financing | ③商品变抵押品 Commodity → collateral | 贸易商 + 银行 Merchants + banks | 不动，躺在交割库里生钱 No — it sits in the warehouse earning credit |
| 投机 Speculation | ②对预期下注 Betting on expectations | 基金、散户 Funds, retail | 永不 Never |
7 · 期货之外：完整的商贸衍生品工具箱 · Beyond Futures — the Complete Merchant Derivatives Toolkit
Trafigura、Glencore 这类公司的对冲工具箱远不止期货。本章补齐剩下的五件工具：期权（重点）、场外掉期（重点）、交叉套保、运费与汇率、期转现。原理仍然只有第 6 章那三条——变的只是工具的形状。
The hedging toolkit at Trafigura or Glencore goes well beyond futures. This chapter adds the remaining five tools: options (in depth), OTC swaps (in depth), cross-hedging, freight & FX, and EFP. The principles are still Chapter 6's three primitives — only the shape of the tool changes.
7.1 期权与含权贸易 · Options & embedded-option trades（含权贸易）
期货锁定价格；期权是给价格买（或卖）保险。矿山用期货套保锁死 1,550——踏实，但 1,900 的牛市也与它无关（第 1 章⑤的董事会问题）。第三条路：买入看跌期权（Put），行权价 1,550，付权利金 30。跌破 1,550 时期权赔付差额（= 有底），涨上去时期权作废、现货照赚（= 上不封顶），代价是那 30 保费。再进一步：嫌保费贵，可以同时卖出一个 1,750 的看涨期权（Call）收回 20——用放弃 1,750 以上的涨幅换保费,这叫领口（Collar），净成本只剩 10。
Futures lock a price; options buy (or sell) insurance on a price. The futures hedge locks the mine at 1,550 — safe, but a 1,900 bull market passes it by (the boardroom problem of Ch. 1 ⑤). The third path: buy a put, strike 1,550, premium 30. Below 1,550 the put pays the difference (= a floor); above, it expires worthless and the physical keeps winning (= open upside), at the cost of the 30 premium. One step further: to cheapen the insurance, simultaneously sell a 1,750 call for 20 — trading away the upside beyond 1,750 to fund the floor. That is a collar, net cost 10.
| 矿山的四种选择 Mine's four choices（假设到期 S=F assume S=F at expiry） | 跌到 1,300 | 停在 1,550 | 涨到 1,900 | 性格 Character |
|---|
| 不对冲 Unhedged | 1,300 | 1,550 | 1,900 | 裸奔 naked |
| 卖期货 Short futures @1,550 | 1,550 | 1,550 | 1,550 | 全锁死 fully locked |
| 买 Put 1,550（−30） | 1,520 | 1,520 | 1,870 | 有底、上不封顶 floor + upside |
| Collar：买 Put 1,550 卖 Call 1,750（净 −10） | 1,540 | 1,540 | 1,740 | 有底有顶 floor + cap |
含权贸易 = 把期权缝进现货合同里卖给产业客户。这是 Trafigura 这类贸易商增长最快的业务之一。例：钢厂怕煤价暴涨但又不会用期权，Trafigura 给它签"封顶采购合同"——点价采购，但最高不超过 1,700，作为交换钢厂多付 15 的升水。拆开看：Trafigura 等于卖给了钢厂一个 1,700 的 Call（收 15），转手在交易所买入同样的 Call 花 12——风险完全对冲，赚 3 的中介差价 + 客户粘性。同理还有给矿山的"保底销售合同"（嵌入 Put）、"二次点价"（嵌入重新定价权）。原理：产业客户要的不是衍生品，是确定性；贸易商把交易所的标准期权翻译成客户能签的现货合同，赚翻译费。
含权贸易 (embedded-option trade) = sewing an option into a physical contract and selling it to an industrial client. One of the fastest-growing businesses at merchant firms like Trafigura. Example: a mill fears a coal spike but won't trade options itself, so Trafigura signs a "capped purchase contract" — price on 点价, but never above 1,700 — in exchange for a 15 higher premium. Decomposed: Trafigura has sold the mill a 1,700 call (collecting 15) and buys the same call on the exchange for 12 — fully hedged, earning 3 of intermediation spread plus client stickiness. Likewise "floor-price sales contracts" for mines (embedded put) and "re-pricing rights" (embedded repricing options). The principle: industrial clients don't want derivatives, they want certainty; the merchant translates exchange-standard options into signable physical contracts and charges for the translation.
7.2 场外掉期与远期 · OTC swaps & forwards
国际市场上，大量对冲根本不走交易所期货，而是场外掉期（Swap）：两家机构双边签约，到期按某个指数的月均价现金结算，没有实物、没有交割库。为什么商贸公司偏爱它？关键在指数匹配：Trafigura 的煤炭销售合同大多按 API2 九月月均价 定价——如果用期货对冲，期货结算的是某一天的收盘价，与"月均价"之间存在天然的时间错配；而卖出一张同样按 API2 九月月均价结算的 Swap，两边分毫不差，基差风险为零。这就是 Swap 的本质：可以定制结算指数、均价窗口、数量和月份的"裁缝版期货"。
Internationally, a large share of hedging never touches exchange futures at all, but uses OTC swaps: two institutions contract bilaterally and cash-settle against a monthly average of an index — no physical, no delivery warehouse. Why do merchants love them? Index matching. Most of Trafigura's coal sales price at the API2 September monthly average — hedge that with futures and you settle on one specific day's close, a built-in timing mismatch against a monthly average; sell a swap settling on the very same API2 September average and the two sides match to the cent — zero basis risk. That is a swap's essence: a tailor-made futures contract whose settlement index, averaging window, size, and month you choose.
| 交易所期货 Exchange futures | 场外掉期 OTC swap |
|---|
| 条款 Terms | 完全标准化 fully standardized | 指数/窗口/数量全可定制 fully customizable |
| 对手方风险 Counterparty risk | 清算所担保 clearinghouse guaranteed | 双边信用（需 ISDA 协议/授信）或转入清算 bilateral credit (ISDA, credit lines) or cleared |
| 保证金 Margin | 每日追缴 daily margin calls | 可用授信额度替代现金 credit lines can replace cash |
| 匹配现货合同 Fit to physical contract | 有时间/指数错配 timing & index mismatch | 可做到零错配 can be exact |
| 透明度 Transparency | 公开行情 public prices | 私下成交（对手看不到你的书）private (rivals can't see your book) |
注意演化方向：2008 年后大量掉期转入集中清算（如 SGX 铁矿掉期、CME 煤炭掉期），既保留定制结算又消掉双边信用风险——铁矿石市场今天的国际定价基准，正是从场外掉期一步步长成的。远期（Forward）则是掉期的实物版：双边、定制、到期真交货——你在第 4 章见过的那些 CCI+20 采购合同，本质上就是嵌了基准定价的远期合同。
Note the direction of evolution: after 2008, swaps moved massively into central clearing (SGX iron-ore swaps, CME coal swaps) — keeping customized settlement while killing bilateral credit risk; today's international iron-ore pricing benchmark literally grew out of OTC swaps. A forward is the physical cousin of a swap: bilateral, customized, and actually delivering — those CCI+20 purchase contracts from Chapter 4 are, at bottom, forwards with benchmark pricing sewn in.
7.2b 拆开一张掉期：机器怎么转 · A swap, opened up — how the machine actually runs
先把"基准（Benchmark）"说死：它就是合同里指向的那个公开参考价——可以是期货价（盘面点价）、也可以是报价机构评估的现货指数（CCI、Platts IODEX、API2）。掉期是围绕基准造出来的最简单的机器：一方付固定价，一方付浮动价（= 基准在约定窗口的均值），到期只轧差付现金，永远没有实物。
First, let's nail down "benchmark": it is simply the published reference price a contract points at — a futures price (board pricing), or an agency-assessed spot index (CCI, Platts IODEX, API2). A swap is the simplest machine ever built around a benchmark: one side pays a fixed price, the other pays the floating price (= the benchmark's average over an agreed window); at expiry only the difference changes hands in cash — never any physical.
掉期结算 Swap settlement = （固定价 Fixed − 基准均价 Index average）× 吨数 Tonnage
卖出掉期 Selling a swap = 收固定付浮动 receive fixed, pay floating（= 做空基准 short the benchmark）
实数走一遍（铁矿石）。Trafigura 有一船 10 万吨铁矿，销售合同定价 = Platts 62% 九月月均价。它怕九月指数跌，于是卖出一张 SGX 铁矿掉期：固定价 105 美元，浮动腿 = 同一个 Platts 九月月均价，10 万吨。到期两种世界：
Real numbers (iron ore). Trafigura holds a 100 kt cargo whose sales contract prices at the Platts 62% September monthly average. Fearing a September slide, it sells an SGX iron-ore swap: fixed price $105, floating leg = that same Platts September average, 100 kt. Two possible worlds at expiry:
| 九月均价结果 Sept average prints | 现货收入 Physical revenue | 掉期结算 Swap = (105 − avg) × 100kt | 合计 Total per ton |
|---|
| 跌到 Falls to 95 | 95 | +10 × 100kt = +$1.0M | 105 |
| 涨到 Rises to 118 | 118 | −13 × 100kt = −$1.3M | 105 |
无论指数怎么走，每吨到手都是 105——注意这不是"对冲得好"，而是代数：收入含 +均价，掉期含 −均价，均价这一项在合计里直接消掉了。还有一个漂亮的细节：月均价掉期天然等于22 张"单日迷你对冲"的捆绑——指数每公布一天，就有 1/22 的结算被定格；而现货 QP 窗口也是每天定价一部分货。两边逐日咬合，这就是为什么交易员说掉期"像手套一样合身"，而期货只能一刀切在某个收盘价上。执行方式：通过经纪商双边成交，再提交 SGX/CME 集中清算——从此像期货一样每日保证金，但结算条款保持定制。
Whichever way the index goes, the take-home is 105/t — and note this is not "good hedging," it is algebra: revenue contains +average, the swap contains −average, and the average simply cancels out of the total. One more elegant detail: a monthly-average swap is natively a bundle of ~22 "one-day mini-hedges" — each daily index print freezes 1/22 of the settlement, exactly as each day of the physical QP window prices a slice of the cargo. The two sides interlock day by day; that's why traders say a swap "fits like a glove" while a futures contract can only cut once, at one closing price. Execution: dealt bilaterally through a broker, then submitted to SGX/CME clearing — daily margin like futures thereafter, but the customized settlement terms survive.
7.2c 工具＝基准：基差风险能被彻底消灭吗？ · Instrument = benchmark — can basis risk ever be fully eliminated?
上面的例子已经回答了你的问题：能——当且仅当对冲工具的结算基准就是合同的定价基准本身。写成代数一眼就懂：
The example above already answers your question: yes — if and only if the hedge instrument settles on the very benchmark the contract prices on. Written as algebra it's self-evident:
现货收入 Revenue = 基准 Index + 升贴水 premium 2
掉期盈亏 Swap PnL = 固定 Fixed 105 − 基准 Index
合计 Total = 105 + 2 = 107，基准项相消 the Index term cancels — 零基差风险，是构造出来的，不是运气 zero basis risk by construction, not by luck
这正是铁矿石市场的历史：2008 年前铁矿靠年度长协谈判定价；长协崩溃后，现货合同转向 Platts/TSI 指数定价，SGX 顺势推出按同一指数现金结算的掉期——工具与基准天生同体，套保精确到分。掉期市场因此爆炸式增长，反过来又巩固了指数的地位。对照第 4 章的四级对冲质量表：这就是把 ★★ 档升到 ★★★ 档的那部工具——DCE 盘面点价合同用 DCE 期货对冲是同一原理（工具＝基准），只是"基准"恰好是期货本身。
This is literally the history of iron ore: before 2008 it priced through annual benchmark negotiations; when that system collapsed, physical contracts moved to Platts/TSI index pricing, and SGX launched swaps cash-settling on that same index — instrument and benchmark born identical, hedges exact to the cent. The swap market exploded, which in turn entrenched the index. Against Chapter 4's four-grade table: this is the machine that promotes grade ★★ to ★★★ — and DCE board-priced contracts hedged with DCE futures are the same principle (instrument = benchmark), just with a benchmark that happens to be the futures itself.
但要说完整：基差风险的守恒定律。基差风险不是自然法则，它是错配的残渣——对冲的每一个条款（指数、窗口、月份、数量）与现货合同每对齐一项，对应那一块风险就归零。所以"指数层面的零基差"完全可以做到；可风险会从没对齐的缝里钻回来：
But the complete answer needs a conservation law: basis risk is not a law of nature — it is the residue of mismatch. For every term of the hedge (index, window, month, volume) that exactly matches the physical contract, that component of risk goes to zero. So "zero basis at the index level" is fully achievable; risk only crawls back through whatever seams were left unmatched:
| 没对齐的缝 The unmatched seam | 钻回来的风险 The risk that crawls back |
|---|
| 你的货 ≠ 指数的货 Your cargo ≠ the index cargo | 指数是 62% Fe 青岛 CFR；你的船是 61% Fe、高铝 → 合同价 = 指数 ± 品质升贴水，而这个升贴水自己会浮动——基差风险换了个名字叫"差价风险" The index describes 62% Fe CFR Qingdao; your vessel is 61% Fe, high alumina → priced at index ± quality differentials, and the differentials themselves float — basis risk reborn as "differential risk" |
| 窗口滑动 The window slips | 船迟到，QP 从九月滑到十月，掉期还挂在九月 → 月差敞口（第 5 章）Vessel delay slides the QP into October while the swap sits in September — calendar exposure |
| 数量 Volume | ±10% 溢短装由对方选择，差额裸奔 tolerance at the counterparty's option; the difference runs naked |
| 指数本身 The index itself | 评估样本稀薄、方法论修改、极端行情下停报（2021 年 10 月）——基准是人编的，不是重力 thin assessments, methodology changes, suspended publication in extremes — a benchmark is man-made, not gravity |
| 现金时序 Cash timing | 掉期/期货保证金是今天的现金，现货货款是下月的现金（第 1 章①）margin cash today vs physical cash next month |
收尾的思维方式：贸易商的工作不是"承受基差风险"，而是风险错配工程——凡是能对齐的条款，用对的工具（同指数掉期、盘面点价、切片对冲）便宜地对齐掉；剩下对不齐的残渣（品质差价、QP 滑动、违约），才是这门生意真正定价并收钱的东西。也想通一层经济学：如果所有风险都能对齐归零，30 块的毛利早被竞争打没了——护城河恰恰长在"对不齐"的那几条缝上，因为只有拥有实物机器（车队、堆场、化验室、授信）的人才有资格管理它们。
The closing mental model: a merchant's job is not "bearing basis risk" — it is mismatch engineering: align every alignable term cheaply with the right tool (same-index swaps, board pricing, sliced hedges), and the unalignable residue (quality differentials, QP slips, defaults) is what the business actually prices and gets paid for. And note the economics: if every risk could be aligned to zero, competition would have ground the ¥30 margin to nothing long ago — the moat grows precisely in the unalignable seams, because only those who own the physical machine (trucks, yards, assay labs, credit lines) are qualified to manage them.
7.3 交叉套保 · Cross-hedging（当你的货没有期货时 when your commodity has no futures）
Trafigura 做的是铜精矿贸易——世界上不存在铜精矿期货。怎么办？用最相关的合约对冲最大的那块风险：一船 10,000 吨、品位 28% 的精矿含铜 2,800 吨，按合同可计价（payable）96.5% ≈ 2,702 吨——就在 LME 卖出 2,702 吨铜期货。铜价这个最大敞口被盖住了；盖不住的残余留在书上：加工费 TC/RC 的波动、金银副产品计价、水分品位争议——这些没有任何期货能碰，只能靠合同条款和组合分散管理。原理：交叉套保是"用一个小的相关性风险，换掉一个大的绝对价格风险"——和第 1 章"用基差风险换价格风险"是同一个动作，只是相关性更松。同理：没有动力煤期货可用的年代，电厂曾用原油/天然气合约做代理对冲；蒙煤商用 DCE 焦煤期货对冲蒙 5，本质也是一次品质维度的交叉套保（第 4 章情景③的教训）。
Trafigura trades copper concentrates — and no concentrates futures exist anywhere. The answer: hedge the biggest risk with the most correlated contract. A 10,000 t cargo at 28% grade contains 2,800 t of copper, payable at 96.5% ≈ 2,702 t — so sell 2,702 t of LME copper futures. The dominant exposure (copper price) is covered; what can't be covered stays on the book: TC/RC (treatment-charge) swings, gold/silver by-product credits, moisture and assay disputes — no futures contract touches these; they're managed through contract clauses and portfolio diversification. The principle: cross-hedging swaps a large flat-price risk for a small correlation risk — the same move as Chapter 1's "swap price risk for basis risk," just with a looser correlation. Similarly: Mongolian traders hedging Mongolian #5 with DCE coking-coal futures are doing a quality-dimension cross-hedge — exactly the lesson of Chapter 4, Row ③.
7.4 运费与汇率：一船货的另外两层皮 · Freight & FX — the other two layers of a cargo
一船煤的到岸利润 = 商品价差 + 运费 + 汇率，三层每层都会动、每层都有自己的纸面市场。运费：FOB 纽卡斯尔买入、CFR 中国卖出的贸易商，天然做空运费（运费涨、利润薄）——用 FFA（远期运费协议，Forward Freight Agreement）对冲：按波罗的海航线指数现金结算的掉期，买入对应航线、对应月份的 FFA，运费涨了纸面赚回来。汇率：美元买货、人民币卖货，签合同到收款之间的汇率波动直接吃利润——用银行远期购汇/NDF 锁死。原理不变：每一层敞口找到结算逻辑最匹配的纸面工具，逐层盖住；盖完之后剩下的，才是贸易商真正想赚的那段价差。
A cargo's landed margin = commodity spread + freight + FX — three layers, each moving, each with its own paper market. Freight: a trader buying FOB Newcastle and selling CFR China is naturally short freight (rates rise, margin shrinks) — hedge with an FFA (Forward Freight Agreement): a swap cash-settled on Baltic route indexes; buy the FFA for the right route and month and a freight spike comes back on paper. FX: buying in dollars, selling in RMB — the drift between contract date and payment date eats margin directly; lock it with bank forwards/NDFs. Same principle throughout: cover each layer with the paper instrument whose settlement logic best matches it; what remains uncovered is exactly the spread the merchant set out to earn.
7.5 期转现 EFP · Exchange for Physical
最后一件小而常用的工具。场景：矿山持有期货空单套保，钢厂持有期货多单套保，两家现在要做一笔现货买卖。各自去盘面平仓再签现货合同？两次滑点、两次手续费，而且平仓价和合同价对不齐。期转现（EFP）允许他们打包处理：向交易所申报，把两边的期货头寸按双方私下商定的价格互相划转注销，同时完成现货过户——期货退出价格与现货合同价格精确咬合，零滑点，不排队交割。原理：EFP 是铁律①和铁律②之间的旁门——既不走盘面平仓、也不走标准交割，让两个套保者把"纸对纸、货对货"一次结清。大宗国际贸易（LME 金属尤甚）里极为常用。
One last small but everyday tool. Scene: a mine holds short hedges, a mill holds long hedges, and the two now want to trade physical with each other. Each closing on the board and then signing a physical contract means two rounds of slippage and fees — and the closing prices won't line up with the contract price. EFP (Exchange for Physical) lets them do it as one package: they report to the exchange, and the two futures positions are cancelled against each other at a privately negotiated price, simultaneously with the physical title transfer — futures exit price and physical contract price lock together exactly, zero slippage, no delivery queue. The principle: EFP is the side door between Rule ① and Rule ② — neither an on-screen close-out nor standard delivery, letting two hedgers settle "paper against paper, goods against goods" in one stroke. Ubiquitous in international bulk trade, especially LME metals.
8 · 头寸簿与风控组织：这一切怎么管 · The Position Book & the Risk Organization
前七章的每一笔交易，最终都汇入同一个地方：头寸簿（Position Book）。商贸公司的赚钱模式是"无数个被锁住的小价差"（30 一单、40 一单），风控组织存在的唯一理由是：确保没有任何一个漏掉的敞口，大到能一口吃掉全年攒下的价差。
Every trade in the previous seven chapters flows into one place: the position book. A merchant's business model is "countless small locked spreads" (30 here, 40 there), and the risk organization exists for exactly one reason: to ensure no single overlooked exposure ever grows big enough to swallow a year of accumulated spreads in one bite.
8.1 头寸簿长什么样 · What the book actually looks like
头寸簿不是流水账，而是按品种、按月份归集的净敞口表。每一行回答同一个问题：这一格里，价格动 1 块钱，我赚亏多少？关键洞察：已签但未点价的合同也是敞口（第 6 章点价：价格还没定，就等于还在裸奔）；期权按 Delta 折算成等效吨数。下面是一本焦煤书的一天：
The book is not a transaction log — it is a net-exposure grid by commodity and by month. Every row answers one question: if the price moves ¥1, how much does this cell make or lose? Key insight: signed-but-unpriced contracts are exposure too (Ch. 6's 点价: price not yet fixed = still naked); options enter as delta-equivalent tonnes. One day of a coking-coal book:
| 敞口来源 Exposure source | 吨数 Tonnes | 方向 Direction | 备注 Note |
|---|
| 港口库存 Port inventory | +30,000 | 多 Long | 成本 1,500，已折仓单 20,000 吨 cost 1,500; 20,000 t as warrants |
| 采购合同·已点价 Purchases, priced | +15,000 | 多 Long | CCI+20，8 月交货 Aug delivery |
| 销售合同·未点价 Sales, unpriced | 0 | — | 价格未定 = 尚无价格敞口 unpriced = no flat-price exposure yet（但有基差敞口 basis exposure exists） |
| 销售合同·已点价 Sales, priced | −25,000 | 空 Short | @1,640，9 月交货 Sep delivery |
| 期货 JM09 Futures | −18,000 | 空 Short | @1,552 套保盘 hedges |
| 期权 Delta Options (delta-equiv.) | −1,500 | 空 Short | 卖给钢厂的封顶合同对冲余量 residual of capped contracts |
| 净敞口 NET flat-price exposure | +500 | 微多 barely long | 价格每动 100，盈亏 ±5 万 a 100-point move = ±¥50k |
看最后一行：账面上百万吨级的业务规模，净价格敞口只有 500 吨——这就是"对冲后的商贸公司"的正常形态。但净敞口≈0 不等于没风险：这本书里还叠着基差敞口（蒙煤 vs 盘面，第 4 章）、月差敞口（8 月多 vs 9 月空，第 5 章）、运费、汇率、对手方——每一层都要单独一张表单独限额。风控的第一课：净敞口是按"维度"算的，不是一个数。
Look at the bottom line: a business handling hundreds of thousands of tonnes carries a net flat-price exposure of just 500 t — the normal shape of a hedged merchant. But net ≈ 0 is not riskless: stacked inside this book are basis exposure (Mongolian vs board, Ch. 4), calendar exposure (long Aug vs short Sep, Ch. 5), freight, FX, and counterparty — each gets its own grid and its own limit. Risk lesson one: net exposure is computed per dimension, not as one number.
8.2 每日循环：MTM 与 PnL 归因 · The daily cycle — mark-to-market & PnL explain
每天收盘后，整本书按市价重估（Mark-to-Market）：库存按现货价、期货按结算价、未点价合同按当前盘面折算。得出今日盈亏后，还要做盈亏归因（PnL explain）——把总数拆回"价格动了多少 / 基差动了多少 / 月差动了多少 / 新签了什么合同 / 仓储利息花了多少"。归因对不上的那一块（unexplained PnL）是风控最警觉的信号：说明书里有没被正确记录的敞口。第 1 章①的现金流问题也在这里管理：财务每天根据 MTM 预测明天的追保现金，提前备妥授信。
After every close, the entire book is marked to market: inventory at spot, futures at settlement, unpriced contracts at the current board. Today's PnL is then decomposed — the PnL explain: how much came from flat price, from basis, from calendar spreads, from new deals signed, from carry costs. The piece that won't reconcile (unexplained PnL) is risk control's loudest alarm: it means the book contains an exposure recorded wrongly or not at all. Chapter 1's cash-flow problem ① is managed here too: treasury projects tomorrow's margin calls from today's MTM and lines up credit in advance.
8.3 组织：前中后台的三道防线 · The organization — three lines of defense
前台（交易员）做交易；中台（风控）独立计量和限制风险；后台（运营/结算）核对每一张合同、每一笔货权和资金。三条铁纪律：①风控直接汇报给 CRO/管理层，绝不汇报给交易主管——给自己打分的考卷没有意义；②交易员不碰确认、结算、货权文件——历史上几乎每一起"魔鬼交易员"事件（中航油 2004 年卖期权巨亏、青岛港重复质押骗贷）都始于同一人既做交易又管单据；③限额是硬的：净头寸限额（如 ±20,000 吨）、止损限额、基差限额、单一对手方授信限额，触线必须减仓，没有"再等等行情"。
Front office (traders) trade; middle office (risk) independently measures and limits risk; back office (operations/settlement) verifies every contract, title document, and payment. Three iron disciplines: ① risk reports to the CRO/management, never to the head of trading — an exam you grade yourself is worthless; ② traders never touch confirmations, settlement, or title documents — nearly every rogue-trader disaster (China Aviation Oil's 2004 options blow-up, the Qingdao port double-pledging fraud) began with one person doing both; ③ limits are hard: net position limits (say ±20,000 t), stop-loss limits, basis limits, single-counterparty credit limits — a breach forces reduction, with no "let's wait for the market to come back."
风险的通用标尺：VaR。一句话定义：95% 单日 VaR = "正常情况下，20 个交易日里最多只有 1 天亏得比这个数多"。粗算演示：净多 10,000 吨、煤价 1,600、日波动率 2% → 单日 95% VaR ≈ 1.65 × 2% × 1,600 × 10,000 ≈ 528 万元。管理层给整本书设 VaR 上限（比如 800 万），交易员在限额内自由发挥——这就是"授权但受控"的现代商贸风控范式。VaR 的著名缺陷也要记住：它对第 4 章情景③那种基差断裂、以及流动性枯竭时的极端行情习惯性失明，所以永远要配合压力测试（"如果重演 2021 年 10 月焦煤三周跌 50%，这本书亏多少？"）。
The universal yardstick: VaR. One-sentence definition: 95% one-day VaR = "in normal conditions, only 1 trading day in 20 should lose more than this." Rough demo: net long 10,000 t, coal at 1,600, daily volatility 2% → one-day 95% VaR ≈ 1.65 × 2% × 1,600 × 10,000 ≈ ¥5.28 M. Management caps the book's VaR (say ¥8 M) and traders run free inside it — the modern merchant paradigm of "empowered but bounded." Remember VaR's famous blind spots: it habitually misses basis breaks like Chapter 4 Row ③ and liquidity-vacuum crashes, which is why it is always paired with stress tests ("if October 2021's three-week 50% coking-coal crash replayed, what does this book lose?").
Part II 的一句话总结：交易员的技艺在第 1–7 章；公司能活二十年，靠的是第 8 章。价差是赚出来的，寿命是管出来的。
Part II in one sentence: chapters 1–7 are the trader's craft; chapter 8 is why the firm is still alive twenty years later. Spreads are earned; longevity is managed.