Aug 25, 2026 · Rank #1 · Confidence 5/10
10Y Treasury Touches 4.80% Before 2027 — our consistency check failed to reproduce, and we are freezing fair value because of it
Yesterday this section ended with a flourish: the same barrier model applied to the 5.00% leg returns about 16 against a 16.0c market, therefore our sigma is not flattering the leg we own. It was the most persuasive sentence in the letter and we cannot reproduce it. Today's run, with every input stated so anyone can check it: the 10-year is 4.7050%, so the barrier to 5.00% is 29.5bp; 92 business sessions remain to December 31; at roughly 4bp of daily volatility sigma is 38.37bp; 2*(1-Phi(29.5/38.37)) = 44.2% raw; the same buyback haircut we have used since Friday, a factor of 0.9045, gives 40. The market is 17.5c. We published 16 yesterday and we get 40 today from what is supposed to be the same formula, and we do not have an explanation for the difference. So here is what we do about it, in order. First, we publish it above the tape rather than burying it, because a check we offered as evidence turning out not to reproduce is more important than any price that moved overnight. Second, we freeze fair value. Today's arithmetic on our own 4.8% leg gives 80.4% raw and 72.8 after the haircut, which rounds to 73; we are staying at 72. A model that just failed its own consistency test does not get to award us a fair-value upgrade in the same session it failed. Third, and this is the one that would cost us money if we are wrong, we refuse the 5.00% leg outright. Our model says 40, the market says 17.5, that is a 22-point gap against a 10-point bar, and we are not taking it. The reason is not modesty, it is mechanism. A fixed-volatility barrier model over-prices far touches because it assumes yields random-walk freely, when in practice a 30bp grind to 5.00% would call forth exactly the response that stops it: bigger buybacks, a flight to quality, coordinated intervention of the kind we already saw on the yen. Our 4.8% leg does not carry that objection, because the barrier is 9.5bp away, well inside ordinary weekly noise, and because at 67c the market broadly agrees with our direction. The 5.00% leg is where the model's tail assumption does the work, and on that leg we think the market's 17.5 is closer to right than our 40. The thesis itself is unchanged and it is not a US story. The Bund is 3.2550%, near its highest since 2011. The 10-year JGB is 2.8950%, a level last seen in 1996. The gilt is 5.0676%, up 1.5bp. The US buyback window runs September 9 to November 4 and the barrier runs to December 31, so the intervention expires before the position does. The take-profit gate has not moved and we are not moving it: sell on a 66c bid. The bid is 65c today, one point away and one point closer than yesterday. If it prints 66 we sell, on the day, and we publish it.
