Mandelbrot
Howard Marks has a line I cannot stop circling back to:
But if the environment is going to be different, maybe different things will work better. I’ve been thinking lately about what is the decision, what is the error that investors commit the most often and in the biggest way with the most detrimental effect? And I think the answer is they believe that the things that are going on will always go on. And they extrapolate to infinity as opposed to expecting regression toward the mean.
The error he describes is the one that feels most natural. Whatever has just happened becomes the template for what will happen next. A stock that has run for three months will run for three more. A sector leading the market today will lead the market tomorrow. The mind treats the recent as the eternal because the recent is what it can see. Marks has spent thirty-five years writing memos, in part, against this single tendency.
What he does not say, but what is obvious to anyone who has tried to live it, is that the discipline of not extrapolating is most painful precisely when extrapolation looks like genius. Everyone in the room is making money. Everyone is showing you their screenshots. The narrative arrives fully formed: the AI buildout is real, the hardware capex is real, the demand is real, and the stocks reflect the demand. You cannot argue with any of it. You can only ask whether the price already contains the story, and whether the story will keep being told at this volume.
I spent most of April asking exactly that question and finding that the question itself was the discipline. The answer was not. The answer is unknowable. The discipline was simply refusing to abandon the framework on the basis of one month of relative pain.
The shape of the month
The recovery from March’s drawdown arrived faster than I had any reason to expect. The war did not end. The uncertainty did not lift. Yet markets, doing what markets do, climbed straight through it. The portfolio finished April up 8.56% on a time-weighted basis, and the year-to-date return ticked back into positive territory at 1.28%. The drawdown from peak still showed 8.88% within the month, so this was a substantial recovery, not a complete one.
The attribution mattered more than the headline number. Of the 8.56%, stock contributions accounted for 7.76 percentage points. Options added 0.06. Income costs were positive at 0.74, mostly because puts I had sold in March came back into profit as volatility collapsed. GOOGL alone contributed 5.09 percentage points. More than half the month’s return came from a single position I had refused to trim in March under a formal exception.
The sector breakdown made the discomfort literal. Technology contributed 5.89 points. Within that, the engine was a single name. Hardware, the part of technology that everyone on my WhatsApp groups was riding, barely existed in my portfolio. I held Micron, briefly, before I trimmed and exited part of the position. Beyond that, my exposure to the AI buildout sat at zero (well there is Alphabet but that’s not hardware pureplay). The recovery rewarded a posture I had not taken.
Cash sat at negative 19.7% by month-end. I came into April on margin and finished April more on margin. One bear signal was active going in: the S&P 500 near key support. Three were active by the close, with the McClellan Oscillator and the VIX Golden Cross joining the list. Going more aggressive in a month when the system was beginning to flag more caution is a sentence I want to sit with for a while.
What I bought, what I sold
Eighteen transactions across the month. Three exits and a string of new or expanded positions that point in a particular direction.
The exits first. I closed SLM and SSRM in mid-April as they did not meet the quality threshold. And I trimmed Micron in two tranches. The MU trim required the most thinking and produced a formal exception on file. I sold MU mainly because the system had marked it a D quality business. My rule is to hold A and B businesses; I can live with a C if the thesis is intact; but a D quality position sitting on a 75% gain looks more like a gift than a holding. The exception was bounded with an expiration through the end of the year, and the rationale was honest: take the money, walk away from the lower-quality book.
I should note what this trim did not do. It did not give me a meaningful position in hardware. It removed the only hardware exposure I had. The trade was correct on its own terms. It also coincided with the moment when staying in hardware would have been most rewarding.
The buys were where the weight of the month sat. I added 250 shares of Whitbread (WTB), the UK hospitality holding, accumulated across two days. I bought 80 shares of Salesforce (CRM) at three different price points. I added 120 shares of Netflix (NFLX) over three trading days. I accumulated 750 shares of Rolls-Royce ADRs (RYCEY) across two days. And I took 500 shares of Blue Owl (OWL) on the 14th.
These are not hardware names. CRM is software, sitting inside the SaaSpocalypse the system has been navigating around. NFLX falls under the Product Love commandment, which I have a formal bounded exception on with the rationale that I bought because the system told me to, not because I love the product. WTB and RYCEY are British names that diversified me geographically away from the US technology weight. OWL is alternative asset management. The pattern is value-leaning, software-heavy, deliberately not chasing the hardware tape.
The April 12 reflection captured the moment I had to make these calls:
The SaaS stocks continued their downward decline. Prior to building the system I would have probably bought more into CSU and ADBE as the conviction still remains. But Mandelbrot told me to hold my hand steady and I continued.
The Mandelbrot reference is to the rules I have set that bound concentration even when conviction screams to add. I held the line on CSU and ADBE. I extended elsewhere. The question of whether software is the next rotation or a value trap will be answered in quarters, not weeks.
The options book
The options activity was larger in turnover and quieter in P&L. Seven new trades opened. Fourteen puts sold. Six calls sold. Eighty-two contracts closed across the month, mostly through expiry, assignment, or buyback at profit thresholds. Total premium collected came to 8.53% of portfolio. The contribution to the month’s TWR was 0.06%. The premium offset income costs almost exactly.
The underlyings span a range I should be honest about: QQQ, MU, XLE, SLV, ORCL, TGEN, OWL, AMSC, SPY. Some are core positions. Some are not. TGEN and AMSC have no place in my equity book; I traded them because the option chain offered something the model liked. The XLE short from the previous month came back into profit after the energy complex sold off. The ORCL position I closed left some premium on the table because I exited at my profit threshold and the contract continued to decay, an outcome that looks like a mistake in hindsight and a discipline at the time.
The April 12 reflection flagged a regret check that I want to record honestly:
Not check markets continuously when monitoring open option positions.
That is the right rule. The compliance score for Probabilities Not Possibilities dropped to 0.6 the same week, and I think those two sit close together. Watching options P&L tick by the second is a way of converting a probabilistic structure into an emotional one. The trade is the trade. The premium is the premium. The expiry is the expiry. Watching does not change any of it.
The mirror
Six weeks of compliance scores. Overall average of 0.898. The research commandments held at 1.0 across the board: Frameworks First, Valuation Discipline, Accounting Skepticism, Capital Cycle Awareness. The work upstream of decisions was clean. The work at the moment of decision was messier.
A few patterns inside the numbers. Study in Bull, Deploy in Bear sat at 0.5 for the last three weeks of the period; my margin balance did not allow me to deploy more, and arguably I should have deployed differently with the cash I had. Concentration Risk dropped to 0.7 from week two onward and stayed there, which is the GOOGL exception making itself felt in the score. Geographic/Sector Diversity oscillated between 0.7 and 1.0, structurally limited by my technology weight. Product Love Is Not a Thesis dropped to 0.7 for the last two weeks, which I read as the system flagging the NFLX entry even though I have the bounded exception on file.
The exception list itself is worth examining. Four formal exceptions are now active. GOOGL bounded through September. NFLX and MU bounded through year-end. RMS and RACE permanent. Permanent exceptions are still the part of the architecture that I am least comfortable with. I wrote in March that a permanent exception to a discipline system is a contradiction in terms, and I still believe it. I have not yet written the conditions under which I would close the RMS/RACE exception. Until I do, the words “until I liquidate my portfolio” remain the rationale, and that is not analysis.
The April 26 reflection captured the harder version of the question:
It has been rough in the market. I have not followed the hardware momentum but instead tried heavily invested in the value opportunities provided by software but the software has continued to bleed. Do I trust myself to stay put even as my portfolio not only underperforms the market but underperforms everyone and anyone I come across on WhatsApp.
The question is rhetorical and not. The system tells me one thing. The room tells me another. The compliance score is the receipt for which voice I listened to.
Posture
Marks was talking about something specific when he wrote the line about extrapolation. He was talking about the moment when an investor needs to change posture for a different environment, and the ways the mind resists doing so. The hardest part is that you cannot tell, in real time, whether your posture is wrong because the environment has changed, or whether the environment is in a moment of noise and your posture is exactly right. You only know in retrospect.
I have been reading Edward Chancellor in parallel with this month’s events. One line has stayed with me:
High returns attract capital just as low returns repel it. The competitive process that ensues acts to drive returns back to the cost of capital.
The line describes the long arc of any sector, the AI hardware buildout included. Capital is flowing in. Capacity is being built. The returns the recent winners are showing on a one-month chart are a signal both of present demand and of future supply. Regression to the mean is not a forecast about next month. It is a description of the gravitational pull on every period of unusual return. Eventually the pull asserts itself.
Eventually is not a tradeable timestamp.
The gap between knowing and acting opens here. I know that hardware capex cycles end. I know that software, beaten down for two years now, may be where the next cycle of returns begins, or may be a value trap that absorbs capital for another year before recovering, or may be a structurally impaired sector whose moats are being eroded by the very AI infrastructure layer the hardware names are building. I do not know which of these is true. The system does not know either. What the system does is bound my behavior so that being wrong is survivable.
The May 3 reflection ends with a sentence that has been sitting on me:
I am being disciplined with holding onto positions, not making impulsive buys but the results are not there. I am seeing people generating extraordinary returns through hardware stocks but I cannot get myself to enter these. The system is not allowing me and I am being disciplined about it. But it is difficult to see others making returns while my portfolio languishes and it has started to feel like that I might be sitting on a bunch of value traps.
The honest reading is that discipline is currently producing pain without obvious reward. The dishonest reading would be that discipline is wrong. The Marks framing helps me hold the honest reading. If the environment shifts, different things will work better. The error is to assume it will not shift.
What April was
April was a month where the market gave me 8.56% and the framework gave me a pass. The numbers look like recovery. They are also, in another light, a record of opportunity I was not positioned to take. I do not yet know whether the WhatsApp screenshots will look prescient or embarrassing in six months. I know that my job is not to know. My job is to maintain a posture that survives both outcomes.
The harder question is one I have been carrying since the system went live. If the environment continues to reward the posture I have refused to take, at what point does discipline become stubbornness? At what point does the framework itself need to be revised? Marks would say: when the environment is genuinely different. He would also say that most environments people declare to be different turn out, in the end, to be the same one wearing new clothes.
I am still trying to figure out which kind of moment this is.