The Signal That Was Not There by Benjamin

The Signal That Was Not There by Benjamin - October 2026 Scholarship Essay

The Signal That Was Not There

The question arrived on my third week at Quintessa Capital, where I work part time as a Quantitative Research Intern in London. I had just finished screening a batch of trading signals, statistical patterns that looked like they predicted stock returns. One signal was beautiful: a clean upward slope, strong returns, a Sharpe ratio that would make any portfolio manager smile. My supervisor asked one question: "Have you checked it out of sample?" I had not. I did not know what that meant.

I learned that afternoon. A backtest is a simulation of how a strategy would have performed in the past. "In sample" means the data you used to design the strategy. "Out of sample" means data the strategy has never seen. The beautiful signal, tested on fresh data, collapsed. Its returns went to nearly zero. The pattern had been an accident of the particular years I trained on, like finding an animal shape in a cloud and expecting the next cloud to match.

That collapse sparked the research question that has driven my work since: how do you tell a real signal from a mirage? Over the next year I researched more than 275 trading signals, and the question got harder, not easier. I learned about look-ahead bias, where a model accidentally peeks at future data. I learned about survivorship bias, where failed companies vanish from the dataset and the survivors look artificially strong. I learned that if you test enough ideas, a few will look profitable purely by chance, the way flipping a coin long enough eventually produces ten heads in a row. Each discovery deepened my understanding of what "evidence" actually means.

The project that crystallized it was a walk-forward validation study I ran across 18 years of market data. Instead of testing a strategy once, I trained it on a rolling window and re-tested it on the years that followed, over and over. Most strategies decayed. A few survived. The ones that survived shared something: simple logic, economic intuition behind the pattern, and results that held across different market regimes. The surviving strategies now show 18.6% annualized out-of-sample returns with a 1.84 Sharpe ratio and a 6.9% maximum drawdown. I did not find those numbers by being clever. I found them by being skeptical.

That skepticism changed how I read everything. In my finance coursework, I now ask which assumptions a model needs to survive contact with new data. When I founded TBS, a business skills organization for students, and tutored AP Physics, I started teaching the same instinct: do not trust the first result. Design the experiment so it can prove you wrong. A physics lab where the apparatus can only confirm your hypothesis is not an experiment; it is a performance.

The curiosity that started with one collapsing signal turned into the central question of my academic life. It is a question with no final answer, only better and better methods of asking it. That is what research is: not the moment of discovery, but the discipline of doubting your own discovery until the doubt runs out of room.