Quick note before I start: this is a personal account of how I ran my own book during 2025. It is not advice, not a recommendation, and not a track record. No numbers, no claims of performance. Just the decisions and the reasoning behind them, written down because writing it down is the only way I get any real honesty out of a year.
How I came into 2025
I rolled into the year with the same momentum system I had been running at the end of 2024. It was working in the sense that I understood it. I knew what it bought, why it bought it, when it rotated, when it sat on its hands. That kind of comfort is useful right up until it becomes the reason you stop looking for better tools.
Around the turn of the year I made the call to swap the engine. I moved to a multifactorial machine-learning model with a clear tilt toward low-volatility names. The reasoning was simple. Momentum had earned its keep, but the breadth of factors the new model was reading felt like a more grown-up way to express what I actually wanted out of the book: a long roster of decent businesses that did not breathe too hard when the tape got noisy.
The hedge
Because the new model leaned hard toward low-vol, the resulting book was structurally long, and not just a little. I wanted something closer to a pseudo long/short profile without actually shorting. So I added a layer of out-of-the-money puts on the benchmark.
The trade had two purposes. First, it capped tail risk on the long book in a way I could quantify ahead of time. Second, it gave me a position that responded to a jump in implied volatility and to sharp directional moves, the long vega and long gamma combination, without committing me to any kind of directional call on the index. I was not betting on a selloff. I was paying a clearly priced insurance policy and accepting the carry as the cost of sleeping at night.
The spring tariff-driven selloff
When the spring tariff-driven selloff arrived, the hedge did what hedges are supposed to do. Implied vol spiked, the protection repriced, and the convexity in the position carried weight while the underlying book behaved the way a low-vol book tends to behave in that kind of tape, which is to say, with less drama than the index.
Here is the part I want to be honest about. A defensive trade that pays off cleanly is a dangerous moment, not a victory lap. There is a particular flavor of overconfidence that shows up right after. You start to feel like you saw something. You start to look for the next thing to see. The hedge was a structural decision. It was not a forecast. I had to keep reminding myself of that for weeks.
Lifting the hedge
A few weeks after the dust settled, I closed out the put protection and moved the book back to a cleanly long-only stance, still on the same multifactorial ML model. That part I want to defend carefully, because on the surface it can look exactly like the kind of discretionary call I try to avoid.
The decision was not based on a view that the worst was behind us. It was based on a risk question, framed in the way I have come to trust most. Given the carry of running protection, the remaining time value in the puts, and the structural defensiveness already living inside the low-vol book, how much more worst-case loss was I actually neutralizing per dollar of premium. The answer, after I forced myself to write it out, was not enough to justify the position. The hedge had done its work during the event. Holding it from there forward was no longer reducing the plausible bad outcome in a meaningful way. So I lifted it.
The model itself stayed where it was for the rest of the year.
Where I almost slipped
The real risk during all of this was never the model. It was me. Specifically, two moments where I caught myself reaching for a trade that had no business being in the book.
The first was right after the hedge worked. Adding more puts, going further out the curve, stretching duration on the protection. None of that was in the playbook. It was just the feeling of having been right, looking for somewhere to spend itself.
The second came later in the year, when the book was running quietly. The urge there is different. It is boredom dressed up as opportunity. Tactical overlays, sector tilts, a quick rotation into something thematic. I kept a written rule for the whole year that any trade not produced by the model needed a backtest before it touched real capital. The rule did most of the work. Writing it down did the rest.
The lesson behind it all
The single sentence I want to carry into next year is this. When you are deciding whether to change your approach, ask how much you can lose if you change, not how much you can gain. If the honest answer is that you could lose less by changing, the decision is probably correct. If the honest answer is that you could gain more, the decision is probably an itch.
Flexibility is one of the better tools a systematic trader has. The mistake is letting flexibility be driven by a story rather than by a risk number. A switch made because something might pay off is a discretionary trade with a quant costume on. A switch made because the worst plausible outcome gets smaller is a portfolio decision, full stop.
I came into 2025 wanting to make more. I am leaving it convinced that the cleaner question, almost every time, is how to lose less. A methodology survives that question. It rarely survives the other one.