“When you’re in a PhD program, you work on one project for five years,” says Ben Wellington, Head of Complex Feature Engines at Two Sigma. “Your job is to become the world’s biggest expert on this very small slice.”
The instinct to go deep and exhaust every single angle of a challenge is exactly what makes someone a great researcher – and this someone we would love to hire. It’s also the very thing that can slow them down the most. At Two Sigma, we’ve spent more than two decades applying the scientific method to financial markets. Along the way, we’ve learned that what may be a five-year idea in an academic setting might become a three-month idea here. And that means getting comfortable with a skill that almost nobody is formally taught and most high achievers actively resist: knowing when to stop.
The persistence trap
Today’s culture treats persistence as a virtue and quitting as a character flaw, and that framing isn’t always wrong. Sometimes the answer really is around the next corner. But persistence has a downside that nobody warns you about.
When you hire curious, driven people and hand them fascinatingly complex problems, of course they want to solve them completely. You can always run one more test, explore one more angle, refine one more parameter. The work feels productive because it is productive — just not always productive enough to justify what it costs in time and attention.
“Any one of our ideas could be its own PhD for five years,” Wellington says. “And it is really hard to know when to stop a project.”
The same drive that produces breakthroughs can also trap you. Not because you’re doing bad work, but because you’re doing good work on something that stopped being the best use of your time two weeks ago. That may be because what you are exploring isn’t working. Or, maybe because you have discovered something real that works, but you want it to make it work better, as any good researcher would. But oftentimes, getting it out the door into production is much more value accretive than keeping it in a lab notebook so that one day you might make a better version of it.
The compounding advantage of small wins
The alternative isn’t giving up easily. It’s deliberately choosing incremental progress over moonshots, often referred to as the Kaizen method: making frequent, marginal gains.
As Stan Sporgis, an Algorithmic Trading Analyst explains, “Two Sigma is not really a place where we try to hit huge home runs all at once. It’s really about slow, iterative improvements done on a consistent basis, because even a small difference is a step in the right direction. And when you add two, three, four or five of those differences over the course of the year or several years, you’ve now made considerable progress. And if enough folks are doing that across the company, then we’re really getting somewhere.”
A modest approach, but consider the math: a 1% improvement, compounded weekly, produces a 68% gain over a year. String together enough small wins, and you’ve built something competitors can’t easily replicate, because there’s no single breakthrough to copy. Just thousands of micro-improvements, each one barely visible, collectively forming an insurmountable lead.
Making stopping acceptable
None of this works if stopping feels like failure. A researcher who kills a project and worries it’ll show up in their performance review is likely going to keep grinding on it quietly, hoping for a breakthrough that justifies the time already spent. So the culture has to make quitting safe before it can make quitting strategic.
At Two Sigma, that starts with data. “Data adds a layer of healthy rigor to the process, so that we aren’t doing things off of whims,” says Joel O’Neil, a Product Manager in modeling and engineering. “We’re not going to continue to pursue a direction unless we have the evidence to back it up.” The decision to stop becomes less personal. It’s no longer a failure when it’s clear that the numbers didn’t get there.
It also means defining what “get there” looks like before you start, not after you’re emotionally invested. When a team sets a clear bar upfront — X needs to happen by Y, or we move on — the stop decision is half-made before anyone begins working.
“It’s okay to try it and then it doesn’t work out,” Sporgis says. “That experiment teaches you a lot. You will learn something new and then be able to improve.” At Two Sigma, the pivot gets respected the same way the breakthrough does. Making the hard call to walk away from a stalled project and redeploy your energy somewhere more promising is showing good judgment. Every hour on one project is an hour not spent on another.
The courage to move on
Obviously there’s a reason persistence gets all the glory. Stopping can feel like failure. When you’ve invested so much, you don’t want to “waste” it. But optimizing for sunk costs rather than future returns is a tricky place to be.
“You can always make anything we’re doing better, but at what point is it better to go to a new project and push that forward versus push this one a little higher?” said Wellington. “That is one of the hardest skills for modelers to learn, but it’s an important one. We have to produce models. We have to produce predictors. So we have to find that balance where you say, ‘Alright, I’ve tried enough, let me try the next project.’ And that’s a really hard part of the job.”
The best modelers, researchers, traders, and builders at Two Sigma fall in love with problems, not solutions. And when the evidence says “move on,” they have the discipline to listen.
The scientific method has guided our work from the very beginning, and we see again and again that the most productive teams aren’t the ones grinding longest on any single project. They’re the ones cycling through ideas, forming hypotheses, testing, learning, and reallocating resources to whatever shows the most promise.
The best researchers and modelers we have don’t cling to their ideas. When the evidence says move on, they move on. That sounds easy, and it never is.