The Woes Of A (Portfolio) Manager

Introduction

Nobody tells you what it actually feels like. It seems like stardom, large checks and big bets, but there’s a subtle sadness and fear behind every PM.

I am writing this for two audiences. First, for PMs, especially emerging, who need to know they are not alone. Second, for researchers and developers who want to understand what keeps their boss up at night.

Most of these problems have no clean solutions. They are tensions to be managed, not puzzles to be solved.

The Talent Paradox

You need to pay your people enough to keep them, but not so much that they stop being hungry.

The best researchers know what they are worth. Underpay them, they leave. But overpay them relative to their current production and something strange happens. They might get too comfortable. The urgency that made them exceptional can start to fade. It doesn’t happen to everybody, but for some, comfort is a motivation killer. This is anecdotal, but I see this mostly in young people who became a “quant” for the money and not for the love of the game. After having some semblance of “fuck you” money, they become slower and significantly less hungry. It is NOT a profit-share / fairness problem, purely a diminishing marginal utility one.

I know of at least two firms that have recently implemented time-based decays on profit sharing. For example, in the first couple of years after a signal or strategy has been created, the researcher/PM gets full profit sharing on the signal or strategy. After some time threshold, the profit sharing % starts to decay. Presumably, this is to try to ensure that their researchers and PMs do not get too comfortable from past successes. In this industry, you are only as good as your next innovation.

Then there is also the cold start problem. When launching a new pod, you need exceptional talent to generate the PnL that would justify paying for exceptional talent. But exceptional talent already has offers from pods with track records. So, what do you do? You bet on yourself and pull forward compensation you cannot yet afford. It’s nothing new, it’s nothing fancy, but we all do it and we all think about it anyway. There is no hack here. You won’t know if you were right for at least a year, minimum.

The Investment Process Maze

Every PM has the same question running in the background: am I doing this right?

Data to features to forecasts to portfolio to execution. At every junction, you made a choice. Each choice seemed reasonable at the time. But you cannot run the counterfactual. Maybe the path you didn’t take was the path to 200 bps more return, maybe it would have reduced your drawdown just enough to prevent being fired. You will never know.

The research literature does not help as much as you would hope. Academic papers and anecdotes tell you what worked on average across some sample. They do not tell you whether it will work for your specific style, your specific universe, your specific turnover constraints, your specific execution costs, your specific platform/firm limitations.

Drawdown Psychology

PnL pressure is abstract until you are living through a drawdown. In a backtest, your eyes move from left to right diagonally, and you mentally note the little bumps along the way.

In reality, you feel and live through the daily bleed. Watching your Sharpe decay in real-time. The firm asking for “updates” that are really interrogations. You sit in meetings explaining variance and “rotations” and “headwinds” to people who are quietly wondering if you have lost it.

The temptation to intervene manually is overwhelming. Turn off signal X. Reduce position size. Override the optimizer. Almost always wrong. Discretionary interventions during drawdowns are usually just some form of panic with a narrative attached.

The worst part is the uncertainty. Is this variance or is the strategy broken? You cannot know in real-time. The sample size is too small. The signal-to-noise ratio is too low. You might be pattern-matching on noise and deluding yourself into think its analysis.

This is where careers are made or destroyed. On whether you can hold your nerve when the equity curve goes against you for months. The PMs who survive are the ones who built conviction and social capital with their team, management and investors during the good times and spent it during the bad.

Alpha Decay and the Treadmill

Signals have half-lives, and those half-lives are shrinking.

A signal that was a champion and worked at institutional scale five years ago is now unironically found as an ETF. What worked eighteen months ago is crowded. The alpha you discovered last year is being arbitraged away by funds that reverse-engineered something similar. You are on a treadmill that speeds up.

This means your research pipeline cannot pause. The moment you stop producing new signals is the moment your existing book starts dying. But your researchers are finite, and some of their time is consumed by maintenance, debugging, and the boring infrastructure work you already convinced them to do.

So you are always behind. Always choosing between maintaining the current book and building the next one. There is no equilibrium where you have enough signals and can relax.

PnL Pressure and the Balancing Act

The firm wants consistent, scalable returns and boring infrastructure work that enables scale. Your researchers want to work on intellectually interesting problems. They did not get PhDs to spend their days debugging data pipelines.

You are in the middle.

A huge part of being a PM is convincing smart people to work on boring problems. Data quality. Execution monitoring. Risk reconciliation. These will not get your researchers poached by Renaissance. But they are the difference between a backtest and a business.

Try selling that to someone who just turned down a superstar pod to join your “hot, new thing”.

Trading Is Harder Than You Think

Your backtest is lying to you.

I don’t mean it is fraudulent. I mean it is showing you a world that does not exist. The world where you can trade any size at the mid, your signal causes no impact, you have infinite shorting inventory and fills happen instantaneously.

You’d think these are the work of an amateur, but I assure it is not. Many firms and large pods in the realm of statistical arbitrage work with signals on a pre-cost, pre-friction basis. They expect researchers to only focus on the theoretical justification of a signal, not its realizability. The core assumption here made by these large firms and pods is that at the asymptote, given enough signals, whilst each signal is individually weak and cannot overcome transaction costs, their aggregate will reveal true predictive power and be able to overcome transaction costs.

Most signals are weak. When researchers show me something with a 1% information coefficient, they act like they have discovered gold. But 1% IC has to survive contact with the real world. You need to trade that signal on a large universe to get breadth. You need turnover to capture the alpha before it decays. And you need to do this while paying transaction costs on every trade.

The Research Hygiene Problem

I have a rule: researchers do not see out-of-sample PnL and statistics.

This makes them uncomfortable. It feels like I am hiding something.

The truth: when a researcher knows the out-of-sample results of their signal, every subsequent research decision becomes contaminated by that knowledge. They are no longer doing science. They are doing motivated reasoning with extra steps.

But from the researcher’s perspective, this feels like being stonewalled. They hand over the signal, and then... silence. They lose the feedback loop that makes learning possible.

I do not have a clean solution. Research cleanliness and researcher visibility are in opposition. You cannot have full measures of both.

Managing Brilliant Egos

Smart people have strong opinions. This is a feature. You want researchers who believe in their ideas, who fight for them.

The problem is that intellectual conflict between smart people can turn into personal conflict. And personal conflict destroys teams.

I have watched people with complementary skills refuse to collaborate because one made a dismissive comment about the other’s methodology years ago.

PMs sometimes have to play the adult in the room and mediate between large egos. The trick, if there is one, is to attack ideas harder than anyone else while visibly respecting the person who proposed them. Critique the signal, not the signal-maker.

Sometimes you have a researcher who cannot separate their ego from their work. These people are poison, no matter how talented. I have learned to fire them faster than feels comfortable.

The Key-Man Time Bomb

Your best researcher could leave tomorrow.

When they leave, they take everything in their head. The institutional knowledge not written down anywhere. The intuitions about which data sources are reliable. The tacit understanding of why certain feature constructions work.

You want to teach everyone everything so no single departure can cripple you. But the more you teach, the more you increase IP risk. Non-competes are increasingly unenforceable. A motivated researcher can leave, wait out any restricted period, and reproduce a significant chunk of what they learned.

Option one: hoard knowledge, create key-man risk. Option two: distribute knowledge, leak alpha to the competition. There is a middle path involving compartmentalization and strategic redundancy, but it is a compromise like everything else.

The Lonely Parts

Decision fatigue is real. Every day you make a sequence of judgment calls with incomplete information. Each decision feels small. The cumulative weight is crushing.

And you cannot voice your doubts to the team. If you say “honestly, I have no idea if our feature engineering approach is correct,” you undermine the confidence that holds the pod together. So you carry the uncertainty alone.

Then there is the responsibility. If the pod fails, it is not just your livelihood at risk. It is everyone who bet on you. The researcher who turned down some other (arguably) more successful pod. The developer who left a stable bank job. They trusted you.

I think about this more than I think about my own comp.

Conclusion

My first PM told me: “Every one wants to be a PM until they realize how toxic the job can be, enjoy your researcher years!” It’s been awhile since then, and I still think about it.

If you are a PM, I hope you feel less alone.

If you are a researcher or developer, the things that keep your PM up at night are the things that, if you solve them, will make you indispensable. The boring problems. The infrastructure problems. The execution problems. That is where careers are made.