One Number Was Hiding Everything That Mattered
For a long time, I judged things by a single number. Whichever version scored highest on my one chosen metric won, got built upon, and survived to the next round. It was clean, it was decisive, and it felt rigorous. It took me an embarrassingly long time to understand that the number I was so diligently optimizing was quietly hiding the one thing that would actually have mattered.
The number was not wrong, exactly. That is the insidious part. It was perfectly accurate and deeply misleading at the same time, which is a combination this domain produces with alarming regularity.
The comfort of a single score
A single metric is seductive for understandable reasons. It takes a messy, tangled, multidimensional reality and collapses it into one value you can rank, sort, compare, and optimize. It feels objective. It gives you a leaderboard, a clear sense of better and worse, a direction to climb. After enough time staring at complexity, the simplicity of a single score is enormously relieving.
But every act of summarizing throws information away. That is what summarizing is. And the only question that ever matters about a summary is the one I was not asking: what, exactly, did this number discard in order to be so conveniently small? Because whatever it discarded is now invisible to me, and invisibility is precisely where danger likes to live.
A high rate of being right that still lost
The cleanest example from my own mistakes was a strategy that was right the overwhelming majority of the time. By the metric I was watching, it looked wonderful — it won, and won, and won again. I was delighted by how often it was correct.
It also lost money. The resolution to that paradox was simple and humbling: it made a great many tiny gains and a handful of enormous losses. It was right constantly and ruined occasionally, and the occasional ruin dwarfed all the frequent little victories combined. The metric I had fixated on measured how often it was right. It said precisely nothing about how much I made when right versus how much I lost when wrong — which, it turns out, was the entire game. The number was true. It was also a near-perfect disguise for a losing strategy.
The average quietly hides the tail
The same lesson recurs in a hundred forms, and one of the most dangerous is the humble average. An average outcome can look perfectly healthy while concealing the actual shape of the distribution beneath it. Two things with an identical average can be utterly different animals: one steady and survivable, the other calm for a long, reassuring stretch and then catastrophic.
In this domain, the thing that ends you is almost always the tail — the rare, extreme event lurking out at the edge of the distribution. And an average is, by its very construction, a machine for washing the tail out. It takes the one region you most need to see clearly and smooths it into the comfortable middle. Judging by averages, I was systematically averaging away the exact events most capable of destroying everything.
Even better numbers get gamed
The natural response is to reach for a smarter metric — something risk-adjusted, something that claims to account for the bumps. And those are genuinely better. But I learned to distrust the next step, which is to optimize hard against that better number.
Because the moment you push relentlessly against any single metric, your process quietly begins discovering the ways to inflate that specific number without delivering the real thing it was only ever a proxy for. The metric and the reality it was meant to represent slowly peel apart, and you are left climbing a measurement while the thing you actually cared about stays flat or sinks. Any number, optimized against with enough force, eventually stops describing the world and starts describing only how to make that number go up.
What a single score cannot tell you
When I finally listed what a single metric structurally cannot tell me, the list was damning. It cannot tell me the shape of how I lose. It cannot tell me whether the whole result hangs on a handful of lucky moments that will not repeat. It cannot tell me how the thing behaves in the worst conditions rather than the average ones. It cannot tell me whether it would survive a genuinely bad stretch, or simply has not met one yet.
Every item on that list is a question about survival. And not one of them fits inside a single number. I had been optimizing a score while remaining almost completely blind to the questions that actually decide whether something lives or dies.
Looking at the whole shape instead
The fix was not a better number. It was giving up the fantasy that any one number was enough. I started forcing myself to look at the whole shape of a result instead of its summary: the full distribution of outcomes, not the average; the worst stretches specifically, not the typical ones; how much the result leaned on rare events; how it behaved precisely when conditions turned ugly.
This is slower and messier. It does not give you a clean leaderboard, and it resists the satisfying click of declaring one thing the winner. But it is the difference between actually understanding a result and merely knowing a single statistic about it. The summary is a headline. I had been reading only headlines and imagining I had read the story.
The deeper lesson: survival lives in what the summary discards
The principle underneath all of this is the one I most needed to learn. The things that kill you in this domain are, with eerie consistency, the very things a summary statistic is designed to smooth over — the tail, the worst case, the hidden dependence on a few rare moments. A good summary is built to suppress exactly that information in the name of being concise.
So the more cleanly I compressed reality into one comforting number, the more efficiently I was blinding myself to the only parts capable of ending the whole effort. Now I distrust any single number on principle, and I distrust a flattering one most of all. When something offers me one beautiful score and invites me to relax, my first question is no longer “how high is it?” It is “what is this number built to hide?”
— No signals, no returns, not investment advice.