Chapter 09 · Sizing and Discipline

仓位与纪律：状态判断、资金管理与"不交易"Sizing and Discipline: Regimes, Risk Budgets, and "No Trade"

前八章讲的是"选什么"。这一章讲的是"下多少"，以及"什么都不下"。这是最后一章，也是唯一一章讲的东西完全在你的控制之内——你无法控制市场往哪走、IV 会不会塌、财报是好是坏，但你可以百分之百地控制那个数字。 The first eight chapters were about what to choose. This one is about how much — and about choosing nothing. It's the last chapter, and the only one whose subject is entirely within your control. You cannot control which way the market goes, whether IV collapses, or how earnings land. You can control that one number completely.

9.1 唯一你能完全控制的变量9.1 The One Variable You Fully Control

第 8 章的三个案例给出了一个不舒服的结论：他们死于仓位，不是死于判断。把这个结论推到底，你会得到本手册最重要的一句话——在长期结果里，仓位的权重高于选股，也高于择时。这不是修辞，它有严格的算术支撑，而这一节就是那个算术。 Chapter 8's three cases delivered an uncomfortable conclusion: they died of size, not of judgment. Push that all the way and you arrive at this handbook's most important sentence — over long horizons, position size matters more than selection and more than timing. That isn't rhetoric; it has strict arithmetic behind it, and this section is that arithmetic.

INTERACTIVE · 同一笔 CSP，拖动张数看形状如何改变the same CSP — drag the contract count

请注意拖动"张数"时发生的事：盈亏平衡点纹丝不动，曲线的形状也完全一样，只有纵轴的刻度在变。换句话说，仓位不改变你判断的对错，它只决定判断错的时候你还能不能继续玩下去。这就是它区别于其他一切决策的地方——选股影响的是期望值，仓位影响的是你能否活到期望值兑现。 Watch what happens as you drag the contract count: the breakeven never moves and the shape is identical — only the vertical scale changes. Size does not alter whether your view is right; it determines whether you're still in the game when the view is wrong. That is what separates it from every other decision — selection moves your expected value, size decides whether you survive to collect it.

9.2 破产的算术9.2 The Arithmetic of Ruin

回撤和回本之间的关系不是对称的，而且不对称的程度远超直觉。亏损 -50% 之后，你需要 +100% 才能回到原点；亏损 -80% 之后，需要 +400%。 The relationship between a drawdown and its recovery isn't symmetric, and the asymmetry is far worse than intuition suggests. After −50% you need +100% to get back to even; after −80%, you need +400%.

回撤 → 回本所需涨幅drawdown → gain required to recover

这条曲线在 -50% 之后近乎垂直。它意味着一件非常具体的事：存在一条你永远不该越过的线，因为线那边不存在"努力赚回来"这个选项。第 8 章那位全仓 XIV 的投资者需要 +1,900% 才能回本，而承载他这个判断的那只票据在三周后就已经不存在了。 The curve is nearly vertical past −50%. That means something very concrete: there is a line you must never cross, because on the other side of it "work your way back" is not an available option. Chapter 8's all-in XIV holder needed +1,900% to break even, and the instrument that carried their view ceased to exist three weeks later.

CASE FILE · 1998-09

LTCM：两位诺奖得主，26 倍杠杆，四个月LTCM: Two Nobel Laureates, 26× Leverage, Four Months

长期资本管理公司（LTCM）于 1994 年成立，合伙人中包括 Myron Scholes 与 Robert Merton——他们在 1997 年因期权定价理论共同获得诺贝尔经济学奖。这大概是金融史上智力密度最高的一间办公室。它的业绩也配得上这份阵容：1995 年约 +43%，1996 年约 +41%，1997 年约 +17%（均为费后）。 Long-Term Capital Management was founded in 1994 with partners including Myron Scholes and Robert Merton, who shared the 1997 Nobel Prize in Economics for option pricing theory. It was probably the highest concentration of intellect ever assembled in one financial office. The record matched: roughly +43% in 1995, +41% in 1996, +17% in 1997, all net of fees.

1998 年初，它的权益约 47 亿美元，而资产负债表规模约 1,250 亿美元——杠杆约 26.6 : 1。它的主力策略是"收敛交易"：押注两个高度相关的债券之间的异常价差会回归正常。这个判断在统计上极其可靠，只是每次的价差都很小，所以必须用巨大的杠杆才能把它放大成有意义的收益。 Entering 1998 it held about $4.7 billion of equity against a balance sheet of roughly $125 billion — leverage near 26.6 : 1. Its core strategy was convergence: betting that an abnormal spread between two highly correlated bonds would revert to normal. Statistically that view was extremely reliable — but each spread was tiny, so enormous leverage was required to turn it into meaningful return.

1998 年 8 月 17 日，俄罗斯对其国债违约。全球资金涌向最安全的资产，而 LTCM 的所有收敛交易同时朝相反方向张开。仅 8 月一个月，基金就损失了约 44%。到 9 月下旬，权益从 47 亿美元跌到约 4 亿美元——约 -92%。9 月 23 日，纽约联储召集 14 家机构注资约 36 亿美元完成接管。 On August 17, 1998, Russia defaulted on its domestic debt. Capital fled worldwide toward the safest assets, and every one of LTCM's convergence trades widened at once. The fund lost roughly 44% in August alone. By late September equity had fallen from $4.7 billion to about $400 million — some −92%. On September 23 the New York Fed convened fourteen institutions to inject about $3.6 billion and take it over.

LTCM · 权益（十亿美元，约值）· equity ($bn, approx.)

现在做这一章的算术。46 亿美元的亏损，摊到 1,250 亿美元的资产上，只相当于约 -3.7%。换句话说：LTCM 的投资组合在四个月里只走坏了不到 4 个百分点，而这不到 4 个百分点，经由 26.6 倍杠杆放大成了权益的 -92%。杠杆没有让他们判断错误，它只是把一个温和的、完全正常的市场波动，翻译成了破产。 Now the chapter's arithmetic. A $4.6 billion loss spread across $125 billion of assets is only about −3.7%. Put plainly: LTCM's portfolio moved against it by under four percentage points across four months, and 26.6× leverage translated those four points into −92% of equity. Leverage did not make them wrong. It merely converted a mild, entirely ordinary market move into insolvency.

Counterfactual：同一组交易、同一次俄罗斯违约、同一个 -3.7% 的资产端走势——如果杠杆是 5 : 1 而不是 26.6 : 1，权益的损失约为 -18%。这是一个难受但完全可以恢复的年份（按 9.2 的曲线，回本需要 +22%），而不是一场需要美联储出面的系统性事件。而且他们最终是对的：那些收敛交易在此后确实收敛了。他们只是没能活到那一天。 Counterfactual: same trades, same Russian default, same −3.7% on the asset side — at 5 : 1 leverage instead of 26.6 : 1, the equity loss is about −18%. That's a painful but entirely recoverable year (by §9.2's curve, +22% restores it), rather than a systemic event requiring the Federal Reserve. And they were ultimately right: those convergence trades did converge afterwards. They simply weren't there for it.

简化声明：LTCM 的权益规模、资产规模、各年收益率与接管事实为真实历史约值。权益曲线上的中间点位为示意插值，用以表现崩塌节奏，非逐月披露数据。反事实中的"5 : 1 杠杆"是纯粹的算术换算（资产端 -3.7% × 5），它假设投资组合构成不变——现实中更低的杠杆也会改变交易规模与市场冲击，因此这个数字用于说明量级，不是精确的历史重演。 Stated simplification: LTCM's equity, balance sheet, annual returns and the rescue are approximate real history. Intermediate points on the equity curve are illustrative interpolation conveying the pace of the collapse, not disclosed monthly data. The "5 : 1" counterfactual is pure arithmetic (−3.7% on assets × 5) and assumes an unchanged portfolio — in reality lower leverage would also have changed trade sizes and market impact, so the figure conveys magnitude rather than replaying history exactly.

亏损减少：同一组交易、同一次违约，资产端约 -3.7% 在 26.6:1 杠杆下变成权益 -92%；若杠杆为 5:1，同样的走势只对应约 -18%Loss reduced: same trades, same default — a −3.7% move on assets became −92% of equity at 26.6:1 leverage; at 5:1 the identical move costs about −18%

9.3 凯利公式：一个数学答案，和它为什么必须打折9.3 Kelly: A Mathematical Answer, and Why You Must Discount It

凯利公式回答的正是"下多少"这个问题：给定优势和赔率，什么下注比例能让长期复合增长率最大？对于一个胜率 55%、赔率 1:1 的赌局，答案是 10%。 The Kelly criterion answers exactly the "how much" question: given an edge and payoff odds, what betting fraction maximizes long-run compound growth? For a 55%-win, 1:1-payoff bet, the answer is 10%.

下注比例 → 长期增长率（胜率 55%，赔率 1:1）bet size → long-run growth (55% win rate, 1:1 payoff)

这条曲线里有两个数字值得记一辈子。第一，它是一座山，不是一条上升的直线。下注比例超过最优点之后，增长率下降——即使你的优势完全真实、一点没变。第二，也是更刺眼的：在下注 20%（也就是两倍凯利）时，长期增长率已经变成负数。你有真实的、可验证的 55% 胜率优势，你却在长期中稳定亏钱——原因仅仅是下注太大。 Two numbers on this curve are worth remembering for life. First, it is a hill, not a rising line. Past the optimum, growth falls — even though your edge is entirely real and entirely unchanged. Second, and more startling: at a 20% bet — twice Kelly — long-run growth is already negative. You possess a genuine, verifiable 55% edge and you lose money steadily over time, for no reason other than betting too large.

但请不要把凯利当成一个可以直接套用的公式，原因很实际：它要求你准确知道自己的胜率和赔率，而在真实市场里你并不知道。你只有一个估计，而估计通常偏乐观。凯利对高估优势的惩罚是不对称的——高估一倍就直接把你推到那条曲线的负值区。 But don't apply Kelly literally, for a very practical reason: it requires knowing your win rate and payoff odds exactly, and in real markets you don't. You have an estimate, and estimates skew optimistic. Kelly punishes overestimation asymmetrically — being wrong by a factor of two drops you straight into the negative region of that curve.

所以实践中的标准做法是半凯利：按公式算出的一半下注。代价很小而好处很大——在上面的例子里，半凯利保留了满凯利 75% 的增长率，同时把波动和回撤大幅降低，并且为"我高估了自己的优势"留出了安全边际。半凯利真正买到的不是收益，而是"我可能算错了"这件事的保险。 So the standard practice is half Kelly: bet half of what the formula says. The cost is small and the benefit large — in the example above, half Kelly retains 75% of full-Kelly growth while sharply reducing volatility and drawdown, and it leaves a margin of safety for "I overestimated my edge." What half Kelly actually buys isn't return, it's insurance against your own estimate being wrong.

满凯利Full Kelly: 增长最快，回撤剧烈Fastest growth, brutal drawdowns
半凯利Half Kelly: 保留约 75% 增长Retains about 75% of growth
两倍凯利Double Kelly: 增长率为负Negative growth
优势估错时If the edge is wrong: 惩罚不对称The penalty is asymmetric

给期权交易者的可执行版本The executable version for option traders

你不需要真的去算凯利。把它的精神翻译成两条可以直接执行的规则就够了：① 单笔交易的最大亏损不超过组合净值的 1%–2%。注意是最大亏损而不是持仓市值——对定义风险的结构，这个数字在开仓时就是已知的（价差的宽度减权利金）；对没有上限的结构，这个数字不存在，所以你根本不该开它。② 相关头寸要合并计算。五个不同科技股的 CSP 不是五笔独立交易，它们在同一天出事，应当按一笔来计仓位。第 1 章那个"保险公司 vs 期权卖方"的思考题问的就是这件事。 You never need to compute Kelly. Translating its spirit into two directly executable rules is enough. ① No single trade risks more than 1–2% of portfolio equity. Note that's maximum loss, not position value — for a defined-risk structure that number is known at entry (width minus credit); for a structure with no ceiling the number doesn't exist, which is why you shouldn't be opening it. ② Correlated positions count as one. Five CSPs on five different tech names are not five independent trades; they go bad on the same day and must be sized as one. Chapter 1's "insurer versus option seller" question was asking exactly this.

9.4 Regime：把前八章接起来9.4 Regime: Connecting the First Eight Chapters

这本书从头到尾只有一条主线：策略没有优劣，只有与 Regime 的匹配与错配。现在可以把它整理成一张表——它不是交易信号，而是一份"什么时候该拿哪把工具"的提醒。 One thread runs through this entire book: no strategy is superior or inferior; there is only a match or mismatch with the regime. It can now be laid out as a table — not as trade signals, but as a reminder of which tool belongs to which weather.

单边急涨Melt-up: 买方的世界（2023 NVDA）Buyers' world (NVDA 2023)
震荡 / 缓涨Chop / grind: 卖方的世界（2017、Wheel）Sellers' world (2017, the Wheel)
趋势熊Trending bear: Put 与价差（2022 ARKK）Puts and spreads (ARKK 2022)
恐慌Panic: 保险最贵，别在此时买（2024-08）Insurance costs most — don't buy here (2024-08)

这张表最有用的地方不是告诉你该做什么，而是提醒你同一个策略在不同状态下是完全不同的两笔交易。2022 年的 Covered Call 把 AAPL 的 -27% 垫成 -13%，同样的操作用在 2023 年的 NVDA 上则是灾难性的机会成本。2017 年做空波动率的人整年顺风顺水，2018 年 2 月 5 日一天还清。没有任何一个策略需要被"戒掉"，需要被戒掉的是"不看天气就出门"这个习惯。 The table's real use isn't telling you what to do; it's reminding you that the same strategy in two different regimes is two different trades. Covered calls padded AAPL's −27% into −13% in 2022, and the identical mechanic on NVDA in 2023 was a catastrophic opportunity cost. Short-volatility traders sailed through all of 2017 and repaid it on a single day in February 2018. No strategy needs to be sworn off. The habit that needs to be sworn off is going out without checking the weather.

9.5 "不交易"是一个合法输出9.5 "No Trade" Is a Legitimate Output

这本手册反复出现过这句话，现在给它一个正式的位置。"不交易"不是没有做决定，它就是那个决定。而它之所以困难，不是因为它需要判断力，而是因为它需要你对抗一种结构性的压力：账户里的现金看起来像是闲置的资源，而一个没有持仓的月份看起来像是失职。 This sentence has recurred throughout the handbook; here it gets a formal place. "No trade" is not the absence of a decision — it is the decision. What makes it hard isn't the judgment it requires but the structural pressure it must resist: cash in an account looks like an idle resource, and a month with no positions looks like negligence.

CASE FILE · 1969

巴菲特关掉了自己的基金Buffett Closed His Own Fund

1969 年，巴菲特合伙企业已经连续十二年跑赢道琼斯指数，年化收益率约 29.5%，而同期道指约 7.4%。按任何标准，这都是一个正处在巅峰的基金经理。 By 1969, Buffett Partnership had beaten the Dow for twelve consecutive years, compounding at roughly 29.5% a year against the index's 7.4%. By any measure, this was a manager at his peak.

然后他给合伙人写信，宣布解散这家合伙企业。理由不是他累了，也不是他判断市场会崩盘——而是更朴素的一句话：按他的标准，市场上已经找不到足够便宜的东西了。他把钱还给了合伙人。 Then he wrote to his partners announcing that he was winding the partnership up. Not because he was tired, and not because he forecast a crash — for a plainer reason: by his standards, there was no longer anything cheap enough to buy. He returned the money to his partners.

DJIA · 1969–1975 （年末收盘与关键点位，约值）(year-end closes and key levels, approx.)

此后发生的事：道指在 1973 年 1 月见顶于约 1,052 点，随后进入战后最严重的熊市之一，1974 年 12 月跌至约 578 点——自高点约 -45%。而选择"不交易"的人，在这段时间里的回撤是 0，并且在 1974 年底手握现金面对一个便宜得多的市场。1974 年 11 月，巴菲特在接受《福布斯》采访时表示，此时的他终于又可以放手买入了。 What followed: the Dow peaked near 1,052 in January 1973, then entered one of the worst post-war bear markets, falling to about 578 by December 1974 — roughly −45% from the high. Whoever chose "no trade" experienced a drawdown of 0 across that stretch, and stood at the end of 1974 holding cash against a far cheaper market. In a November 1974 Forbes interview, Buffett indicated he could finally buy freely again.

这个案例的意义不在于"巴菲特预测到了熊市"——他并没有，他只是说找不到符合标准的东西。这才是可以被普通人复制的部分："不交易"的触发条件不是你对未来的预测，而是当下的机会是否满足你事先写下的标准。前者需要预知能力，后者只需要诚实。 The point isn't that Buffett predicted the bear market — he didn't; he said only that nothing met his standards. And that is the replicable part: the trigger for "no trade" isn't a forecast about the future, it's whether today's opportunities meet a standard you wrote down beforehand. The first requires prescience. The second requires only honesty.

简化声明：道指点位、合伙企业的年化收益与解散事实为真实历史约值。"回撤为 0"指持有现金本身的名义回撤——1973–74 年的高通胀会显著侵蚀现金的实际购买力，此处未计入这一影响。 Stated simplification: the Dow levels, the partnership's compound return and its wind-up are approximate real history. "A drawdown of 0" refers to the nominal drawdown of holding cash — the high inflation of 1973–74 substantially eroded cash's real purchasing power, which is not accounted for here.

亏损减少：1969 年选择"不交易" → 道指自 1973 年 1 月高点回撤约 -45%，而持有现金的名义回撤为 0Loss avoided: choosing "no trade" in 1969 → the Dow fell about −45% from its January 1973 peak, while cash's nominal drawdown was 0

思考Think 如果"不交易"是一个合法输出，那么一个交易系统的完整性，应该用什么来衡量？是它在有信号时的胜率，还是它在没有信号时保持沉默的能力？把这个问题具体化：过去十二个月里，你有多少次因为"这个月还没做过交易"而开了一个本来不该开的仓？ If "no trade" is a legitimate output, then by what should a trading system's completeness be measured? Its win rate when it fires, or its ability to stay silent when it shouldn't? Make it concrete: over the past twelve months, how many positions did you open only because you hadn't traded yet that month?

9.6 开仓前的一页纸9.6 The One Page Before You Enter

把这本书能给的所有东西压缩成六个问题。它们全部可以在下单前一分钟回答完，而且任何一个答不出来，就是"不交易"。 Everything this book can give you, compressed into six questions. All of them can be answered in the minute before you submit an order, and any one you can't answer means "no trade".

① 观点① The view: 方向？幅度？时间窗？Direction? Magnitude? Window?
② 定价② The price: 市场定了多少？我在它哪一边？What's priced? Which side am I on?
③ 结构③ The structure: 这个观点最省钱的表达是什么？Cheapest expression of that view?
④ 最坏情况④ The worst case: 跳空 30%，我亏多少？（要数字）Gap 30% — how much? (a number)
⑤ 仓位⑤ The size: ④ 是否 ≤ 净值的 1–2%？Is ④ ≤ 1–2% of equity?
⑥ 相关性⑥ Correlation: 它和我已有的持仓同一天出事吗？Does it fail on the same day as my book?

第 ④ 条是这张清单的核心，也是第 8 章三块墓碑上共同缺失的那一条。请注意它要求的是一个具体的数字，而不是一种感觉。"风险可控"不是答案，"跌到 $180 我最多亏 $740"才是。而当你的结构没有上限、这个数字根本不存在时——那本身就是答案。 Question ④ is the heart of the list, and the one missing from all three of Chapter 8's headstones. Note that it demands a specific number, not a feeling. "The risk is manageable" is not an answer; "if it falls to $180 I lose at most $740" is. And when your structure has no ceiling so the number doesn't exist — that is itself the answer.

9.7 本章测验9.7 Chapter Quiz

点击选项即可作答，答案与解析即时显示。Click an option to answer; the correct answer and explanation appear immediately.

你的账户回撤了 50%。需要多少涨幅才能回到原点？Your account is down 50%. What gain restores it?

C。100 跌 50% 剩 50，从 50 回到 100 需要翻倍。回撤与回本之间的关系是 d ÷ (1−d)，而不是对称的。这个不对称在 -50% 之后急剧恶化：-80% 需要 +400%，-90% 需要 +900%。A 是最直觉也最危险的答案，它正是很多人低估回撤代价的原因。 C. 100 down 50% leaves 50, and 50 back to 100 is a double. The relationship is d ÷ (1−d), not symmetry. The asymmetry worsens sharply past −50%: −80% requires +400%, −90% requires +900%. A is the intuitive and most dangerous answer, and it's precisely why people underestimate what a drawdown costs.

LTCM 的投资组合在四个月里资产端只走坏了约 3.7%。为什么权益亏了约 92%？LTCM's portfolio moved against it by only about 3.7% on the asset side. Why did equity fall about 92%?

B。3.7% × 26.6 ≈ 98%——这就是全部的机制，一行乘法。A、C 与事实不符：那些收敛交易此后确实收敛了，标的和模型都没问题。D 是最诱人的错误答案——它把责任推给"黑天鹅"，但真正的问题不是违约无法预测，而是他们的仓位配置成"任何 4% 的意外都不能发生"。在 5:1 杠杆下，同一次违约只是一个 -18% 的难受年份。 B. 3.7% × 26.6 ≈ 98% — that is the entire mechanism, one line of multiplication. A and C are false: those convergence trades did converge later; neither the assets nor the models were the problem. D is the most seductive wrong answer, assigning blame to a black swan — but the real issue isn't that the default was unforeseeable, it's that they were sized so that no 4% surprise could be allowed to happen. At 5:1 the same default is a painful −18% year.

你有一个真实的 55% 胜率优势（赔率 1:1），最优下注比例是 10%。如果你下注 20%，长期会怎样？You have a genuine 55% edge at 1:1 odds, so the optimal bet is 10%. What happens long-run if you bet 20%?

D。这是凯利曲线最反直觉也最重要的一点：它是一座山。越过顶点后增长率下降，而在两倍凯利处它已经跌破零——你拥有一个真实、可验证的优势，却在长期中稳定亏钱，唯一的原因是下注太大。这也解释了为什么实践中要用半凯利：它保留约 75% 的增长，同时为"我高估了自己的优势"留出安全边际。 D. This is the Kelly curve's least intuitive and most important feature: it's a hill. Past the peak, growth declines, and by double Kelly it has already crossed below zero — you hold a real, verifiable edge and still lose money over time, for no reason other than betting too large. It's also why practitioners use half Kelly: about 75% of the growth, with a margin of safety for having overestimated the edge.

你在五只不同的科技股上各卖出一张 CSP。按仓位管理，这算几笔交易？You sell one CSP on each of five different tech stocks. For sizing purposes, how many trades is that?

A。分散的前提是不相关，而五只科技股在市场下跌时高度相关——它们会同时被指派。这正是第 1 章"保险公司 vs 期权卖方"那道思考题的答案：保险公司能卖出十万份互不相关的车险，而你的五张 CSP 是同一场车祸的五份保单。B 是最常见的自我安慰：持仓数量不等于分散度，相关性才是。 A. Diversification requires independence, and five tech names are highly correlated in a selloff — they get assigned together. This is the answer to Chapter 1's "insurer versus option seller" question: an insurer writes a hundred thousand uncorrelated auto policies, while your five CSPs are five policies on the same collision. B is the commonest self-comfort: the number of positions is not diversification; correlation is.

巴菲特 1969 年解散合伙企业，最主要的理由是什么？What was Buffett's principal reason for winding up the partnership in 1969?

B。这个区别是这一章的关键。A 需要预知能力，B 只需要诚实。巴菲特没有预测熊市，他只是把"符合标准的机会"这个条件如实地评估为"不存在"。这才是普通人能复制的部分："不交易"的触发条件是当下机会与事先标准的比对，不是对未来的预测。C 与事实相反——他此前十二年年化约 29.5%。 B. The distinction is this chapter's crux. A requires prescience; B requires only honesty. Buffett did not forecast the bear market; he simply assessed "opportunities meeting my standard" and honestly found none. That is the replicable part: the trigger for "no trade" is comparing today's opportunities against a standard set in advance, not a prediction about the future. C is the opposite of the truth — he had compounded at about 29.5% for twelve years.

本章要点 · 全书收束Key Takeaways · Closing the Book

仓位是这本书里唯一完全属于你的变量。市场往哪走、IV 会不会塌、财报是好是坏，都不归你管；而那个数字百分之百归你管——并且它决定的不是你判断的对错，而是判断错的时候你还能不能继续玩下去。破产的算术是不对称的：-50% 需要 +100% 才能回本，-90% 需要 +900%，而 LTCM 的资产端只走坏了 3.7%，26.6 倍杠杆就把它翻译成了权益的 -92%——那些收敛交易后来确实收敛了，他们只是没能活到那一天。凯利公式说得更狠：即使你拥有真实的优势，下注到两倍最优值，长期增长率也会变成负数。所以规则很朴素——单笔最大亏损不超过净值的 1%–2%，相关头寸合并计算，而"最大亏损"必须是一个你在下单前就能说出口的具体数字。 Size is the only variable in this book that fully belongs to you. Which way the market goes, whether IV collapses, how earnings land — none of that is yours; that one number is entirely yours, and what it decides isn't whether you're right but whether you're still playing when you're wrong. The arithmetic of ruin is asymmetric: −50% needs +100% back, −90% needs +900%, and LTCM's assets moved just 3.7% before 26.6× leverage translated it into −92% of equity — those convergence trades did converge, the firm simply wasn't there. Kelly puts it harder still: even with a genuine edge, betting twice the optimum turns long-run growth negative. So the rules are plain — no single trade risks more than 1–2% of equity, correlated positions count as one, and "maximum loss" must be a specific number you can say out loud before you enter. 而贯穿这九章的那条线，到这里可以完整地说出来了。策略没有优劣，只有与 Regime 的匹配与错配：2020 年属于 Put 买方，2022 年属于 Call 卖方，2023 年属于 Call 买方，而同一个 Covered Call 在这三年里分别是救命的、赚钱的和灾难性的。入场纪律高于管理技巧：Wheel 在微软和英特尔上走了完全相同的四个步骤，结果相差七十个百分点，差别全部发生在第一步。而"不交易"是一个和买入、卖出完全平等的合法输出——巴菲特在 1969 年用它跳过了一场 -45% 的熊市，靠的不是预测，只是诚实地承认没有机会符合他事先写下的标准。期权给了你一套精确表达观点的工具，但它从不告诉你什么时候你其实没有观点。那一部分，永远只能由你自己填。 And the thread running through all nine chapters can now be stated whole. No strategy is better or worse; there is only match or mismatch with the regime: 2020 belonged to put buyers, 2022 to call sellers, 2023 to call buyers — and the same covered call was, across those three years, a lifesaver, a moneymaker, and a disaster. Entry discipline outranks management technique: the Wheel ran identical four-step procedures on Microsoft and Intel to outcomes seventy points apart, and the entire difference occurred at step one. And "no trade" is an output exactly as legitimate as buy or sell — Buffett used it in 1969 to sit out a −45% bear market, not by forecasting it, but by honestly admitting that nothing met a standard he had written down in advance. Options give you a precise vocabulary for expressing a view. They never tell you when you don't actually have one. That part is always yours to supply.