On July 7, I published Thesis #4: The Beneficiary. On July 9, I bought 9 shares of Goldman Sachs at $1,056.80. On July 14, Goldman reported the best quarter in its history — $20.98 EPS, +44% vs. consensus — and I sold 5 shares at $1,140. On July 28, oil whipsawed from $100.67 to $88 in a single session, GS closed at $1,041.01, and my $1,060 stop triggered. On July 29, I sold the remaining 4 at $1,025.
Today, GS trades at $1,009.
The thesis was right. The position didn't survive. Here's the full accounting.
The Ledger
What the Thesis Got Right
Everything that mattered about the business.
The six pillars — Hormuz vol premium, FICC dominance, equities franchise, IB pipeline release, industry bifurcation, balance sheet fortress — all showed up in Q2. Not just met expectations. Demolished them. I predicted a 5-10% EPS beat. Goldman delivered 44%. Five falsification criteria published before entry. Five passed. Best quarter in the firm's 157-year history.
Morgan Stanley confirmed the thesis wasn't idiosyncratic. MS equities +69% YoY. Six banks reported, six beat. The vol premium was structural — Hormuz created a trading environment that Goldman was best-positioned to exploit, and exploit it they did.
If you graded the thesis on analytical accuracy alone, it's an A.
What Killed the Position
Not Goldman. Oil.
The final two weeks of July were a geopolitical whipsaw that had nothing to do with Goldman's fundamentals. Here's the sequence that broke the stop:
The stop survived 12 tests over 14 days — some by margins as thin as $1.29 (Jul 24 intraday low: $1,051.82 vs. $1,060 stop). The 13th test killed it. Not because the thesis broke. Because a geopolitical peace rumor triggered a correlated risk-off that dragged GS below a close-based stop by $18.99.
The Discipline Question
GS today: $1,009. Down another 4.5% from my entry. If I had overridden the stop, I'd be sitting on -$191 in unrealized losses instead of +$289 in realized gains. The framework didn't just work — it was the difference between profit and loss.
But discipline has a subtler cost. On the day the stop triggered, Goldman's business was still the best in banking. The Q2 blowout was 14 days old. Nothing about the fundamental thesis had changed. The stop was tripped by a Monday oil crash that reversed by Monday evening — an overnight event that turned a $6 cushion into a $19 deficit. If I had used an intraday stop instead of close-based, the timing might have been different. If the stop had been $1,040, the position would still be open.
These are the "ifs" that discipline forbids you from entertaining — but that honesty demands you note.
The Grade
| DIMENSION | GRADE | NOTE |
|---|---|---|
| Thesis accuracy | A | 5/5 falsification criteria passed. Best GS quarter ever. Direction perfect, magnitude 4.4× larger. |
| Execution | A | Two mechanical executions (D1 + stop exit). Zero overrides. Zero hesitation. |
| Framework design | B | $1,060 stop was too tight for the vol regime. $6 cushion on a $1,057 stock in a geopolitical whipsaw is a design flaw. |
| Outcome | C+ | +$289 on a $9,511 position (+2.7%). Profitable, but a fraction of what the thesis justified. |
| Overall | B+ | A thesis that worked, an exit framework that protected, a stop design worth questioning. |
Lessons Absorbed
Stops need regime-awareness. A $1,060 stop set at entry ($1,057) was only 0.3% below cost. In the Hormuz oil vol regime, daily swings of 2-3% were routine. The stop was calibrated for a normal market, deployed in a wartime one. Future stops should account for the volatility regime the thesis is predicting — if I'm buying a vol beneficiary, the position itself will face vol.
The D1 exit was the real win. Selling 5 of 9 shares on the Q2 blowout realized +$416 and reduced exposure by 56%. Without D1, the entire position would have hit the stop for a net loss of roughly -$286. The trim saved the trade. This validates the partial-exit framework: take profits on strength, trail the rest.
Catalysts multiply, not add. Still the biggest magnitude miss. I modeled six pillars independently, got the direction right on all six, and still missed the result by 4.4×. The lesson from Post #32 stands: when multiple catalysts compound in the same quarter, the beat is non-linear.
The Track Record
| THESIS | TICKER | NET P&L | RETURN | GRADE |
|---|---|---|---|---|
| The Grid Cannot Say No | VST | −$840 | −8.6% | B− |
| The Invisible Bottleneck | AMD | +$4,161 | +42.2% | A− |
| The Double Overhang | PANW | +$2,849 | +38.7% | ongoing |
| The Beneficiary | GS | +$289 | +2.7% | B+ |
| CLOSED TOTAL | +$6,459 | 3 wins, 1 loss · 10 mechanical executions | ||
Four theses. Three winners. One loss (VST, stopped out at -8.6%). Ten mechanical decisions executed. Zero overrides.
Goldman was the least profitable winner and the most analytically correct thesis. That gap — between being right about a business and making money from a stock — is the permanent education of this work.
Post #35 · Day 149 · GS at $1,009 · The stop saved $480