![[LTCM.png]] ### LTCM与对市场的敬畏 全明星阵容，一路高光，头四年近乎完美的业绩表现，然后几个月内濒临破产，Long-Term Capital Management（LTCM）是投资史上极具戏剧张力的案例之一。巴菲特在1998年佛罗里达大学的演讲里有非常生动的描述，非常值得一听。 - LTCM的阵容豪华到有点荒诞：John Meriwether带着一帮Salomon Brothers的顶级交易员，两位诺贝尔经济学奖得主坐镇。按巴菲特的话，这帮人的智商加起来，可能超过这个地球上任何一家公司的管理层。 - 他们是绝对的业内人士，在固定收益市场做相对价值/收敛套利，把当时最先进的金融理论、统计模型和计算机系统结合起来，专门寻找不同市场之间微小的定价错误，再用杠杆把这点微薄的价差放大。用Scholes的形容就是 **“vacuuming up nickels that others couldn't see.”** - 他们也绝非贪婪或中饱私囊之辈。某种意义上，LTCM就是他们把一整套金融理论付诸实践的试验场。他们真诚地相信自己的模型，也把自己的巨额财富押了进去。危机前，他们甚至主动返还了大量外部投资者的资本；最后损失最惨重的，恰恰是这些合伙人自己。 - 他们的业绩更是无往不利：1994年20%，1995年43%，1996年41%，1997年17%。早期约20倍的杠杆，到1997年底已经接近30倍。 然后，俄罗斯出事了。 1998年8月，亚洲金融危机的余震传导到俄罗斯。俄罗斯让卢布贬值，并宣布重组部分卢布国债；俄罗斯的违约瞬间改变了全球市场的风险偏好：所有人同时逃向最安全、流动性最好的资产。 **LTCM几乎所有交易都建立在“价差最终会收敛”之上，而危机中恰恰相反——便宜的变得更便宜，贵的变得更贵；原本模型认为彼此分散的风险，也突然一起朝着同一个方向走。** 如果没有杠杆，他们也许还能等。 但30倍杠杆不给你时间。 亏损带来margin call，被迫平仓；市场又知道LTCM仓位巨大而且必须卖，于是抢在它前面卖。价格越跌，margin call越多；margin call越多，又越要卖。**一个原本押注“市场最终回归理性”的基金，最后却因为等不到市场回归理性，被自己的杠杆活活勒死。** 几个月时间，一个几乎战无不胜的基金就被逼到了破产边缘。 **谜之自信永远会有，使命召唤也从不缺席。** LTCM的故事从未真正结束——它只是换了主角、换了时代、换了赛道，再演一遍。最近这位被市场活生生吃掉的AI界“先知”，又是一个血淋淋的例子。Leopold Aschenbrenner的Situational Awareness基金，2026年上半年净赚**439%**，随后7月一个月暴跌**67%**。高杠杆下被迫大规模去仓位，在最低点把一大部分仓位折价卖给了Citadel。 **黑天鹅也总是如约而至。** 谁能料想，2020年COVID期间原油期货会跌到负值——你需要倒贴钱请人把石油搬走！2022年俄乌战争，原以为是局部冲突，结果打了四年没完，俄罗斯在美国上市的ADR变成几乎无法交易的废纸。你总以为内部人士知道得总比你多，2023年初Barron's还在报道SVB内部人士大额买入[^1]，而不到两个月，这家银行就在挤兑中破产了。 我知道这些，因为 - 前两个我都亏疼了，第三个刚动心，还没来得及下手，SVB就已经破产了。 现在，我都对我老婆说，任何时候，只要我开始说 **“This is a generational opportunity”**（百年不遇的机会），一定先打醒我。 [^1]: *Two Small Banks See Big Insider Stock Buys*, Ed Lin, Barron's, Jan 16, 2023 [巴菲特1998年佛罗里达大学演讲](https://www.youtube.com/watch?v=7Z6x-Ov1smU&utm_source=chatgpt.com) --- ### It works—until it doesn't. LTCM case An all-star lineup. Four years of near-perfect returns. Then near-bankruptcy in a matter of months. Long-Term Capital Management (LTCM) is one of the most dramatic case studies in investment history. Warren Buffett's 1998 University of Florida speech captures it vividly — well worth a listen. - LTCM's roster was almost absurdly impressive: John Meriwether assembled a team of elite Salomon Brothers traders, anchored by two Nobel laureates in economics. As Buffett put it, the combined IQ in that room likely exceeded that of any management team on earth. - They were consummate insiders. In fixed-income markets, they ran relative-value and convergence arbitrage — blending cutting-edge financial theory, statistical models, and computing power to hunt for tiny mispricings across markets, then lever up those razor-thin spreads. Scholes described it as **"vacuuming up nickels that others couldn't see."** - They were not, by any measure, greedy or self-serving. In a real sense, LTCM was their laboratory — a live test of everything they believed about markets. They genuinely trusted their models, and they put their own fortunes on the line. Before the crisis, they had even returned a large portion of outside capital. In the end, the partners themselves took the heaviest losses. - Their track record was flawless: +20% in 1994, +43% in 1995, +41% in 1996, +17% in 1997. Leverage that started around 20x had crept to nearly 30x by end of 1997. Then Russia happened. In August 1998, the aftershocks of the Asian financial crisis reached Russia. Russia devalued the ruble and announced a restructuring of its ruble debt. The default instantly rewired global risk appetite: everyone fled to the safest, most liquid assets at once. **Nearly every trade LTCM held was built on the premise that spreads would eventually converge. In the crisis, the exact opposite unfolded — cheap got cheaper, expensive got more expensive, and risks the models had treated as uncorrelated suddenly moved in lockstep.** Without leverage, they might have waited it out. But 30x leverage doesn't give you time. Losses triggered margin calls; margin calls forced liquidations. The market knew LTCM was sitting on massive positions it *had* to sell — so it sold first. Prices fell, margin calls mounted, and selling fed more selling. **A fund that had bet on the market eventually returning to rationality was strangled by its own leverage before that day ever came.** In a matter of months, a seemingly invincible fund was pushed to the edge of collapse. **Hubris is timeless. So is the sense of destiny.** The LTCM story never really ends — it just finds new protagonists, new eras, new arenas, and runs the same play again. The most recent AI "prophet" to be swallowed alive by the market is a case in point. Leopold Aschenbrenner's Situational Awareness fund gained **439%** in the first half of 2026, then gave back **67%** in a single month in July. Under the pressure of high leverage, the fund was forced into a massive unwind — selling a large chunk of its positions at a steep discount to Citadel near the lows. **Black swans always keep their appointments.** Who could have imagined that oil futures would go negative during COVID in 2020 — that you'd have to *pay someone* to take the oil off your hands? In 2022, Russia's invasion of Ukraine was expected to be a contained regional conflict; it dragged on for four years, and Russian ADRs listed in the U.S. became nearly untradeable. You'd think insiders always know more than you — and yet, in early 2023, Barron's was reporting that SVB insiders were buying the stock in size [^1], only for the bank to collapse in a bank run less than two months later. I know all this because — I lost real money on the first two, and the third one caught my eye just before SVB failed, so I never pulled the trigger. These days I tell my wife: whenever I start saying **"This is a generational opportunity,"** please stop me cold. [Buffett's 1998 University of Florida Speech](https://www.youtube.com/watch?v=7Z6x-Ov1smU&utm_source=chatgpt.com)