Quantitative funds are described to students as places where mathematicians make money from patterns, which is true enough to be useless. The useful description is narrower: a systematic fund replaces the individual investment decision with a rule, applies the rule across a large number of positions, and spends almost all of its effort on whether the rule will keep working.
The defining difference
A discretionary investor studies a company and decides to own it. A systematic investor studies a hypothesis, tests it on history, implements it as code, and then owns whatever the code says to own today.
That shift changes what skill means. The discretionary edge is judgement about a specific situation. The systematic edge is a small statistical advantage, applied consistently, over enough independent positions that the advantage shows through the noise. A rule that is right 52 per cent of the time is worthless on one trade and highly valuable across ten thousand.
Why breadth beats accuracy
The central insight of systematic investing is that performance depends on the quality of a forecast and on how many independent times you can apply it. Doubling the number of genuinely independent bets improves the risk-adjusted return roughly as much as improving the forecast by forty per cent, and the former is usually easier.
Independent is the load-bearing word. Two thousand positions in the same sector responding to the same factor are close to one bet. Much of the real work at a quantitative fund is establishing how many independent bets a strategy actually contains, which is invariably fewer than the position count suggests.
The research process
| Stage | What happens | Where it usually dies |
|---|---|---|
| Hypothesis | An economic reason a pattern should exist | No reason beyond the data |
| Data | Sourcing, cleaning, point-in-time alignment | Survivorship and look-ahead bias |
| Signal construction | Turning the idea into a number per asset per day | Too many parameters |
| Backtest | Historical simulation with realistic assumptions | Costs and slippage ignored |
| Out-of-sample test | Data the researcher never touched | Performance disappears |
| Capacity analysis | How much capital it can carry | Works at $10m, not at $500m |
| Paper and small live | Real prices, real fills, small size | Live results diverge from simulation |
| Allocation | Risk budget alongside existing signals | Correlated with what the fund already runs |
Why most signals fail
Being wrong here is the normal state, and the discipline of a good research group is mostly a set of defences against fooling yourself.
- Overfitting: enough parameters will fit any history. The test is whether it survives data the researcher has never seen.
- Look-ahead bias: using information that was not available at the time the trade would have been placed. Restated financials and index membership are the usual culprits.
- Survivorship bias: testing on the companies that still exist. The ones that went to zero are exactly the ones the strategy needed to avoid.
- Transaction costs: a strategy turning over daily can look excellent gross and lose money net. Spread, market impact and financing are not details.
- Capacity: the alpha is finite. A signal that works on small caps may hold a tenth of the capital the fund needs to deploy.
- Crowding and decay: a public edge stops being an edge. Signals lose potency as others find them, which is why research is continuous rather than a one-time build.
The infrastructure is not a detail
At a systematic fund, technology is the product rather than a service to it. Data pipelines, a backtesting engine that cannot accidentally look ahead, an execution system that measures its own slippage, and risk systems that decompose exposure by factor in real time.
This is why funds hire so many software engineers and why engineering candidates are frequently better paid and more employable at these firms than they expect. A research idea that cannot be implemented, tested honestly and executed cheaply is not an idea.
Where these firms sit relative to each other
Three loose groups, with different hiring and different work. Systematic hedge funds run diversified statistical strategies across many markets at horizons of days to months, and hire researchers with statistics and machine learning backgrounds. Proprietary trading and market-making firms operate at much shorter horizons, are latency-sensitive, and hire for low-level engineering and fast quantitative reasoning. Bank quantitative desks price and hedge derivatives, which is a different discipline built on stochastic calculus and model validation rather than on signal research.
Candidates apply across all three with one preparation and are surprised when the interviews differ. They are different jobs that share a label.
Frequently asked questions
What is the difference between a quant fund and a regular hedge fund?
A discretionary fund makes decisions about individual investments; a systematic fund makes decisions about the rules, then executes whatever the rules say. The edge shifts from judgement on one situation to a small statistical advantage applied across a large number of independent positions.
Why do quant strategies stop working?
Crowding and regime change, mostly. A public edge attracts capital until the return is competed away, and a relationship that held under one rate or liquidity environment can break when that environment changes. This is why research is continuous rather than a one-time build.
What is overfitting in a backtest?
Fitting a strategy so closely to historical data that it captures noise rather than a real relationship. Enough parameters will fit any history. The only meaningful defence is testing on data the researcher has genuinely never seen, and being suspicious of anything with no economic reason behind it.
Do quant funds hire software engineers?
Heavily. Data pipelines, backtesting engines that cannot look ahead, execution systems and real-time risk are the product rather than support for it, so engineering seats at these firms are numerous, well paid and often more accessible than research roles.
Related guides
Want this applied to your recruiting?
Reading is the easy part. For practitioner feedback tailored to your situation, work 1:1 with Suro.