What is quantitative finance?

When people hear “quant” they usually think of a trading algorithm. That is one piece of it. The simpler way to put it: quantitative finance is finance done by measurement. You buy, sell, hedge or hold because a number tells you to, and that number comes from a rule that somebody wrote down, tested on past data, and can run again tomorrow. For most of the industry’s history the decision was made the other way, by judgement. An experienced person read the news, talked to people, and decided. A lot of finance still works like that.

Quants are not cleverer than those people. The difference is that a written rule can be checked against the past, you can work out what it costs, and when it fails you can see why. A feeling gives you none of that. This post covers what the field is, how it came to run so much of the market, who works in it, and what they need to know.

Where it came from

Three papers started it. All three are short, and worth reading in the original.

• 1952, Harry Markowitz, “Portfolio Selection” (Journal of Finance). Risk is the standard deviation of returns; a portfolio’s risk depends on how its holdings move together, not just on each one; and for any level of risk there is a best mix. Everything called “diversification” with a number attached descends from this.
• 1964, William Sharpe, “Capital Asset Prices” (Journal of Finance). A stock’s return splits into the part that follows the market (beta) and the part that does not (alpha). Fund managers have been paid on alpha ever since, and the Sharpe ratio, return per unit of risk, is the industry’s yardstick.
• 1973, Fischer Black and Myron Scholes, “The Pricing of Options and Corporate Liabilities” (Journal of Political Economy). A formula for what an option is worth, and with it the idea that a derivative can be hedged by trading the thing it is written on. Derivatives desks, and the risk departments that watch them, are built on this.

For about thirty years those ideas stayed in universities and a few banks. Using them for real needed data and computing power that did not exist yet. Romero and Balch, in What Hedge Funds Really Do, date the change to the 1990s: quantitative finance had matured and computers were cheap enough to examine thousands of price relationships at once, and the economists, physicists and mathematicians hired to do it were nicknamed “rocket scientists”. The first wave had a famous failure, Long-Term Capital Management, whose partners included Myron Scholes and Robert Merton, the two who shared the 1997 Nobel prize for the option-pricing work, and which lost its capital in 1998 when a bet that price gaps would close met a market in which they widened instead. The industry did not conclude that measuring was a mistake. It concluded that the risk had to be measured as carefully as the return, and that is where the modern risk department comes from.

Who uses it, and for what

The hedge fund is the one most people picture, so a few facts about it. Romero and Balch describe a hedge fund as a lightly regulated pool of money from “accredited” investors, which means wealthy enough that the law assumes they can look after themselves. It is free to go short and to borrow, and it is paid “two and twenty”: 2% of assets a year plus 20% of the return. The arithmetic explains a lot about how the industry behaves, so here it is:

fee, what the manager keepsr, the fund's return for the year

A fund that returns $\good{4\%}$ pays its manager $\key{2.4\%}$, which is 60% of the return; one that returns $\good{10\%}$ pays $\key{3.6\%}$, 36% of it; one that loses money pays no 20% but still pays the 2%. So the manager earns most in a good year and is never paid nothing. That is why a fund gets judged on three things together: the absolute return, the return against a benchmark such as the S&P 500, and the risk taken to get it, usually combined in the Sharpe ratio. Often what the investor is paying for is not a higher return than the index. It is a similar return with a smoother ride.

The industry is no longer a niche. HFR’s industry report for the end of 2025 put global hedge fund capital at $5.15 trillion, the first time above $5 trillion, after the strongest year of inflows since 2007 (HFR, January 2026). The approach goes well beyond funds, too. The SEC’s 2020 staff report on algorithmic trading describes the US stock market as “primarily automated”, where “most of the lifecycle of trading is automated” and algorithms “facilitate the provision of and search for liquidity by a broad range of participants”; its reading of the research is that algorithmic trading has improved liquidity and market quality in normal conditions, while some kinds of it can worsen periods of stress (SEC staff report). Whoever you are, when you send an order today it is matched by a program, priced by a program, and most likely answered by one.

What the work is

Most of a quant’s week is research, not trading. The loop looks like this.

• Data comes first and takes most of the time. Prices must be adjusted for splits and dividends; the set of companies must be the one that existed on each date, not today’s survivors; a quarterly result must be used only from the day it was published. Romero and Balch devote three chapters to data quirks, data sources and backtesting, and their list of what a data feed must have is a good summary of the job: history, freedom from survivor bias, ease of integration, low latency.
• The hypothesis is a rule, written so a computer can run it: “buy the 20 stocks with the highest 12-month return, hold a month, repeat”. The “why” matters as much as the rule. A pattern with no reason behind it is usually the first one to stop working.
• The backtest is the experiment. It is also where the field’s three famous mistakes live: using information that was not yet available (look-ahead), trying so many rules that one fits by luck (over-fitting, or data mining), and assuming the market that produced the past will produce the future.
• Risk gets measured rather than felt: volatility, drawdown, how much the strategy depends on the market as a whole, the worst day in a thousand. A strategy that earns 10% a year and loses 40% in one bad month is not really a 10% strategy.
• Execution is its own specialism. A large order moves the price against itself, so it is cut into pieces and fed to the market by algorithms, and the cost of doing that badly can exceed the edge of the idea.
• Monitoring closes the loop. Markets adapt. The CFA Institute’s handbook on AI in investments points out that once a pattern is published, it fades. So a strategy has a shelf life, and the research never really stops.

The skill set

The job needs three kinds of knowledge. Most people who do it well are strong in one or two and competent in the rest.

Two things about that table are easy to miss from outside.

First, the statistics are easy to describe and hard to do honestly. Financial data is small and noisy. Thirty years of daily closes is about 7,500 numbers. A monthly signal gives you a few hundred at most. The CFA handbook’s chapter on machine learning in investments lists the consequences: low signal-to-noise, tiny datasets by the standards of other fields, markets that change and adapt, and a replicability problem in published results. Its advice is what experienced quants say anyway: keep the models simple, use what you know about markets to sanity-check the output, test strictly in time order, and combine models rather than trusting one. The hard part is the discipline, not the model.

Second, the programming is not a support role. Hilpisch’s Python for Finance describes the old division of labour, quants prototyping in one language and a technology team re-implementing in another, and the double work and translation errors it produced; the industry’s move to Python for both research and production ended it. A quant who cannot ship their own code is now unusual. Look at the headcount and the budget of a bank or a fund and you are looking at a technology company that trades.

The CFA handbook adds a third observation about teams rather than individuals: the firms that succeed with new methods build “T-shaped” teams, where investment people, data scientists and a bridging role that can translate between them work together, and they hire finance people who learned the technology more often than technologists who learn the finance. Both routes in exist. This blog is written from the second.

Why it matters even if you never trade

The methods carry over. A backtest is a replay test. Look-ahead bias is the same bug as reading a field that was only back-filled later. Survivorship bias is what you get when you report only the requests that completed. A Sharpe ratio is a signal-to-noise ratio with money in the numerator. If you have built or run a system that is not allowed to be wrong, in payments, settlement or fraud detection, you already think the way a risk quant thinks, and the habits will look familiar. The field learned its rules about data, testing and noise the expensive way, and machine learning in general is still learning the same ones.

Next

The finance stats series on this site works through the numbers behind this post, from probability and expected value to the Sharpe ratio and portfolio optimisation, on real daily prices. The ML basics series does the same for the models.

References

• Philip J. Romero and Tucker Balch, What Hedge Funds Really Do: An Introduction to Portfolio Management (Business Expert Press, 2014): ch. 1 and 2 on what a hedge fund is and how it is paid, ch. 12 to 14 on data and backtesting, ch. 15 on Long-Term Capital Management, ch. 16 on the industry since.
• Handbook of Artificial Intelligence and Big Data Applications in Investments, ed. Larry Cao (CFA Institute Research Foundation, 2023): the introduction on T-shaped teams, and ch. 1 (Mike Chen and Weili Zhou) on why financial data is hard and how practitioners cope.
• Yves Hilpisch, Python for Finance: Mastering Data-Driven Finance (2nd ed., O’Reilly, 2018), ch. 1 on technology in finance and the move to one language.
• Harry Markowitz, “Portfolio Selection”, Journal of Finance 7(1), 1952, doi:10.1111/j.1540-6261.1952.tb01525.x.
• William F. Sharpe, “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk”, Journal of Finance 19(3), 1964, doi:10.1111/j.1540-6261.1964.tb02865.x.
• Fischer Black and Myron Scholes, “The Pricing of Options and Corporate Liabilities”, Journal of Political Economy 81(3), 1973, doi:10.1086/260062.
• US Securities and Exchange Commission, Staff Report on Algorithmic Trading in U.S. Capital Markets (August 2020).
• HFR, Global Hedge Fund Industry Capital Surges Past Historic $5 Trillion Milestone (January 2026).