![[TLT_analysis_2026-08-28.png]] ### 国债波动收益 **谁想赌下一次美联储开会说什么，把我的TLT拿去赌吧。会开完把份额还给我，再留下$50。** 这个策略通过不断低买高卖，为国债增加一层波动收益，使其总收益有机会匹配股权类资产。收益主要取决于国债产品本身的波动性以及设置的交易频度。波动性大体可以用IV30来观察 (期权市场隐含的未来30天波动率，年化). 交易频度如果设置在1天到1星期完成一个买卖来回，对于利率这种大部分时间在一个宽区间里波动的产品，波动本身可以贡献相当可观的收益 (~3x IV30)。 TLT（20+年美国国债ETF）平静期的IV30通常在12%。我的目标是，在合适的市场环境下，通过票息 + 波动收益做到20–30%左右的年化回报。 实测也大体如此。2025/2026年使用这个策略交易TLT，期间偶尔手动调仓，年化收益分别约21%和34%。期间30年国债利率大致在4.3–5.3%之间波动，TLT在$82–93/share之间波动。 总共131 + 148笔交易，主要集中于2025/04–2026/08这16个月，大约每两天一个买卖来回。 策略组成 1） 超长久期国债 主要TLT。因为长久期(约15)，对利率非常敏感，粗略说1%利率变化对应15%的反向价格变化。 Zero-coupon的ZROZ、EDV对利率更加敏感，但流动性和季度分红等让实际操作稍微麻烦一些。 2）单笔交易收益与手续费 每单Slot $4000-8000，网格步宽0.6-1.2%。一个买卖来回赚约$50，手续费约$1。 3）低融资利息的证券账户 这个策略叠加在账户其他持仓之上，所以动用资金需要考虑融资成本，目前约5%。如果波动收益连融资成本都cover不了，也就没有折腾的必要。 4）网格化交易自动填单 跌下来买一点，涨回去卖一点。大部分自动挂单完成，不需要猜CPI和美联储。 补充考虑 1）为什么把它当作压舱石 Druckenmiller访谈提过，如果想做空市场，他会首选买国债，因为有carry, 成本最划算。 我自己的组合有不少风险资产。如果风险仓位表现不好，资金涌向安全资产，国债可以提供一定缓冲。所以这部分仓位平时收利息、做波动，risk-off时又可能起到对冲作用。 2）如果三十年利率破区间去7% 2025/2026年30年国债利率虽然逐渐抬升，但总体还在区间波动，没有出现极端单边行情。这其实是这个策略很舒服的环境。 如果30年利率一路涨到7%，TLT会产生~30%的浮亏，而且网格会一路下跌一路加仓。之前反复赚的$50小票，并不能完全消灭这种尾部风险。 不过我自己有30年房贷固定在3.25%。如果真有6–7%的长期国债可以买，我倒也愿意慢慢接。某种意义上，就是拿3.25%的长期固定负债去匹配更高收益率的长期资产。 3）如果美国政府真不还国债呢？ 那美元大概也要出大问题。我的账户偏实物资产/商品——黄金、能源——应该可以cushion一部分。 但真到了这种末日场景，要担心的可能已经不只是证券账户了。 ------------ 巴菲特不喜欢黄金：不增长、不产生现金流，还要倒贴保管费。但有人把黄金借给首饰商赚利息，让原本躺着不动的资产多产生一层收益。 我觉得国债也有点类似。 放着不动，一年5%已经不错。但市场每天都有人赌CPI、美联储和Powell下一句话，制造出大量利率噪音。 我不一定需要知道他们谁对。 利率噪音多，又经常出不了一个大区间，倒是很适合长期做一把流动性供应商。 --- ### Bond Churn *Whoever wants to bet on what the Fed will say at its next meeting — go ahead and use my TLT. Give the shares back when it's over, and keep the $50.* This strategy adds a layer of volatility return on top of a bond's coupon, giving the total return a fighting chance to rival equities. How much you earn depends mainly on the bond's volatility and how frequently you trade. Volatility can broadly be tracked with IV30 — the 30-day implied volatility from the options market, annualized. At a cadence of roughly one round trip every day to one week, interest rates — which spend most of their time bouncing within a wide range — can contribute meaningfully to returns through volatility alone (~3x IV30). In quiet markets, TLT (the 20+ year U.S. Treasury ETF) typically has an IV30 around 12%. My target, in the right environment, is 20–30% annualized from coupon plus volatility combined. Real-world results have broadly matched that. Trading TLT with this strategy in 2025 and 2026 — with occasional manual adjustments — produced annualized returns of roughly 21% and 34%, respectively. Over that stretch, the 30-year Treasury yield oscillated between about 4.3% and 5.3%, and TLT moved between $82 and $93 per share. All told, 131 + 148 trades over the 16 months from April 2025 to August 2026 — roughly one round trip every two days. **How the strategy works** 1. **Long-duration Treasuries.** Primarily TLT. Its long duration (~15) makes it highly sensitive to rate moves: a rough rule of thumb is that a 1% change in rates translates to about a 15% move in the opposite direction. Zero-coupon alternatives like ZROZ and EDV carry even more rate sensitivity, but their liquidity and quarterly dividend mechanics add some operational friction. 2. **Trade sizing and commissions.** Each slot is $4,000–8,000, with a grid step of 0.6–1.2%. One round trip earns roughly $50, with about $1 in commissions. 3. **Low-cost margin.** This strategy runs on top of the rest of the portfolio, so borrowed capital means financing costs — currently around 5%. If the volatility return can't even cover the cost of margin, there's no point running it. 4. **Grid auto-orders.** Buy a little on the way down; sell a little on the way up. Most of it executes automatically, with no need to guess the CPI print or parse Powell's next sentence. **A few additional thoughts** *Why I treat this as ballast.* Druckenmiller has said in interviews that when he wants to short the market, his first choice is buying Treasuries — the carry makes it the most cost-efficient hedge. My own portfolio carries meaningful risk exposure. When risk assets sell off, capital tends to flow into safety assets, and Treasuries can cushion that. So this position earns interest and volatility income in normal times, and can act as a hedge when risk appetite dries up. *What if the 30-year rate breaks higher and hits 7%?* In 2025–2026, the 30-year yield has been drifting gradually higher but still oscillating within a range — no dramatic one-way move. That's actually the sweet spot for this strategy. If the 30-year runs to 7%, TLT would be sitting on roughly a 30% unrealized loss, and the grid would keep buying on the way down. The $50 per round trip can't fully absorb that kind of tail risk. That said, I have a 30-year mortgage fixed at 3.25%. If I can buy long-dated Treasuries yielding 6–7%, I'm actually willing to accumulate slowly — in effect, funding higher-yielding long bonds with a mortgage I'm already paying at 3.25%. *What if the U.S. government actually defaults?* Then the dollar probably has bigger problems. My portfolio skews toward real assets and commodities — gold, energy — which should absorb some of the blow. But in a scenario that extreme, a brokerage account is probably the least of your worries. --- Buffett famously dislikes gold: it doesn't grow, generates no cash flow, and costs money to store. But some people lend gold to jewelry makers for interest — squeezing an extra layer of return from an asset that would otherwise just sit there. I think about Treasuries the same way. Sitting still, 5% a year is already respectable. But every day, the market is full of people betting on CPI, the Fed, and Powell's next word — generating a constant stream of rate noise. I don't need to know who's right. Heavy rate noise, contained within a wide range — that's a pretty good setup for being a long-term liquidity provider.