A model is a lever on real money

Graduate · enrolled learners

This lesson opens with Two Sigma, disclosed 2023.

What happened

Two Sigma is one of the most sophisticated quantitative hedge funds in the world, running strategies built and governed by teams of scientists, and it disclosed that one of its own researchers had made unauthorized changes to the firm's trading models. The researcher adjusted the models in ways that boosted the performance of some funds and hurt others, producing around 170 million dollars of unintended profit-and-loss swings the firm had not sanctioned. The models were not wrong in a mathematical sense; nothing in the equations failed, and no market moved unexpectedly. The failure was one of governance: a single person was able to change the models that moved real money without the controls that should have caught the change catching it. Even at a firm defined by quantitative rigor, the danger was not the mathematics but the process around it, the question of who could alter a live model and whether anyone would notice, and the answer that day was that someone could, and no one did in time.

The decision point

A trading model is not a neutral piece of mathematics; it is a lever connected to real money, so the moment it goes live the central risk shifts from whether the math is correct to who is allowed to change it and whether a change will be noticed. Two Sigma is the case: at one of the most rigorous quant firms in existence, a researcher's unauthorized changes to live models produced around 170 million dollars of unintended profit-and-loss swings, not because the math failed but because the governance around the math did. This is the opening lesson of quantitative methods for a professional trader: sophistication in the model does not protect you if the process governing the model is weak, because a model that decides trades is a lever anyone with access can pull, and a lever anyone can pull unnoticed is a risk no equation captures. The error is to treat model risk as purely a question of whether the model is right, when the more dangerous question is who can alter it, under what authorization, and with what independent check, since a correct model changed by the wrong person at the wrong time moves real money in ways nobody intended. So the discipline is to treat control over a live model as a first-class risk equal to the model's correctness: to know exactly who can change a model that moves money, to require authorization and independent review for any change, and to detect changes rather than trust that they will not happen, because Two Sigma shows that even a firm whose entire edge is quantitative rigor is undone not by its mathematics but by a gap in who could pull the lever.

Recorded loss

$170,000,000

Orientation and Year One are open: anyone can read them without an account. From Year Two onward the lessons are for enrolled learners, because progress through the later years only means anything if it is tracked against a record.

It is free. We do not sell the list and there is nothing to buy at the end of it.