Strategy names get used loosely, and a candidate who can place them on a horizon axis understands the field better than one who can define each in isolation. Holding period determines the data you need, the costs that bind, the infrastructure required and the kind of person a firm hires.
The horizon axis
| Strategy | Horizon | Exploits | Binding constraint |
|---|---|---|---|
| Market making | Seconds and below | The bid-ask spread, for providing liquidity | Latency and adverse selection |
| High-frequency arbitrage | Microseconds to seconds | Price differences across venues | Speed; the race is technological |
| Statistical arbitrage | Hours to days | Short-term mean reversion between related assets | Transaction costs and crowding |
| Event and news-driven | Minutes to days | Systematic reaction to scheduled information | Data quality and speed of parsing |
| Factor and equity market neutral | Weeks to months | Persistent cross-sectional return premia | Factor crowding and drawdowns |
| Trend following and managed futures | Weeks to months | Persistence in price direction across markets | Long flat periods and whipsaws |
| Volatility and options relative value | Days to months | Gaps between implied and realised volatility | Tail risk; short volatility is short insurance |
| Risk premia and carry | Months | Being paid to bear a risk others avoid | Correlated losses in a crisis |
Market making, which is a service before it is a bet
A market maker quotes a price to buy and a price to sell simultaneously and earns the difference. It is paid for providing immediacy to whoever wants to trade now, and in a competitive market that payment is small and repeated an enormous number of times.
The risk is not that prices move, because inventory is hedged quickly. The risk is adverse selection: the counterparty who lifts your offer may know something. Managing that means widening quotes when information is likely, skewing prices to unwind inventory, and being fast enough to cancel a stale quote before somebody else takes it. Interviews at these firms test exactly this, usually through a game where you must quote a two-way price on something uncertain and then live with the fills.
Statistical arbitrage
The original version is pairs trading: two historically related securities diverge, you short the expensive one and buy the cheap one, and you profit when the relationship reasserts. Modern implementations generalise this to hundreds of assets and dozens of relationships estimated simultaneously, but the intuition survives.
Two things break it. The relationship can widen further before it reverts, which is a leverage and risk management problem rather than a research one. And the relationship may not be a relationship at all: two series can be correlated by coincidence, especially when you have searched a thousand pairs to find them. This is why an economic reason for the linkage matters more than the strength of the historical fit.
Factors, and what they actually claim
Factor investing says a handful of characteristics explain much of the cross-section of returns: value, momentum, quality, size, low volatility and carry among them. A market-neutral factor portfolio goes long the assets scoring well and short those scoring badly, so that the market direction cancels.
The honest debate is why any premium persists. Either it compensates for a real risk, in which case it should continue and will hurt when that risk materialises, or it reflects a behavioural bias, in which case it may be arbitraged away as it becomes known. Momentum and value both had long and painful drawdowns in the last two decades, which is the practical answer to anyone presenting factor returns as a free lunch.
Trend following, and why it survives criticism
Trend following buys what has been going up and sells what has been going down, across futures markets in equities, bonds, currencies and commodities. It is criticised for having no fundamental basis, wins in extended directional moves, and loses steadily in choppy, range-bound markets.
Its durability comes from two properties rather than from forecasting skill: it is diversified across dozens of uncorrelated markets, and it tends to perform in sustained crises, when other strategies correlate to one. That makes it valuable in a portfolio even during years when it makes nothing on its own, which is a distinction worth being able to draw in an interview.
Execution, the strategy nobody applies for
A large fund that has decided to buy is left with a hard problem: buying without pushing the price up. Execution research is the study of how to split an order across time and venues to minimise the cost of trading, and it is a genuine quantitative discipline with a direct and measurable payoff.
It is also the most underrated entry point in the field. Fewer candidates apply, the work is tractable, and the skills transfer directly into signal research later. A candidate who can discuss implementation shortfall, market impact and the trade-off between speed and price is signalling an understanding of how money is really made and lost.
What this means for how you prepare
- For market making and high-frequency firms: mental arithmetic under time, probability, expected value, and the discipline of quoting a price and living with it. Low-level engineering for the technology roles.
- For statistical arbitrage and factor funds: statistics, regression, time series, and the ability to discuss overfitting without prompting.
- For trend following and macro systematic: portfolio construction, risk parity, correlation and drawdown behaviour.
- For volatility strategies: options pricing, the Greeks, and what implied volatility represents.
- Across all of them: coding. Python for research is the baseline, and C++ where latency matters.
Frequently asked questions
What is statistical arbitrage?
Trading short-term deviations from a historical relationship between related securities, most simply by shorting the expensive one and buying the cheap one and profiting as the relationship reasserts. Modern versions estimate many relationships at once, and the main danger is finding relationships that were coincidence.
How does a market maker make money?
By quoting both a buy and a sell price and earning the spread, repeated an enormous number of times. The main risk is adverse selection rather than price movement: the counterparty who trades against you may know something, so quotes widen when information is likely and skew to unwind inventory.
Does trend following still work?
It has long flat periods and loses in range-bound markets, which makes it hard to hold. Its value in a portfolio comes less from standalone returns than from being diversified across dozens of markets and tending to perform during sustained crises, when most other strategies correlate to one.
What is execution research?
The study of how to complete a large order without moving the price against yourself: splitting it across time and venues, and measuring implementation shortfall and market impact. It is a genuine quantitative discipline, less contested by applicants, and it transfers directly into signal research later.
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.