August 17, 2026
Share on
In Citadel’s Equity Quantitative Research strategy, alpha researchers work on a fundamental challenge in systematic investing: how to predict returns in a way that is rigorous, repeatable and profitable in live markets.
Their work sits at the front end of EQR’s research engine. Data comes in, predictions are produced, portfolios are constructed and trades are executed in live markets. Alpha research is where many of those predictions begin. Researchers identify opportunities, build models to forecast returns and test whether those forecasts can improve trading decisions after accounting for risk, transaction costs and real-world market frictions.
Max Bane, a Quantitative Research Lead in EQR, views the role as a blend of creativity and discipline. A typical day can involve generating hypotheses about how markets behave, expressing those ideas as models and determining if they generalize to new data and future time periods.
“Simply put, the main goal of an alpha researcher is to predict returns,” Max said. “But crucially, it is not just to predict returns on some static historical dataset. It is to do so in the true out-of-sample of live trading, where your model is consuming input data it has never seen before.”
EQR approaches portfolio construction as an optimization problem. Alpha research provides the return forecasts that help the system understand where opportunity may lie.
“The role of alpha research is to build models that identify the opportunity set,” Max said. “We identify the trades that could be profitable, against which costs are compared to decide whether we want to make them.”
The hypotheses driving these models can be derived from a wide array of sources, e.g., a new dataset, an academic journal, a market observation, a news article, a new interpretation of existing information or a new modeling technique.
The difficulty usually lies in deciding which ideas merit further research. Generating ideas is the easy part.
“Once you have a room full of creative people, the ideas become cheap,” Max said. “What really matters is executing and prioritizing among them.”
That prioritization requires the researcher to exercise good judgment. They have to consider an idea’s plausibility, scale, frequency and breadth, then use first-principles thinking to theorize about why an effect should exist before deciding how to test it.
EQR manages systematic equity portfolios across multiple time horizons, creating a broad research environment where many different styles of inquiry can coexist.
At shorter horizons, researchers often have more observations, which can support highly empirical approaches and more expressive models. At longer horizons, there are fewer independent observations, making causal reasoning and simpler, more interpretable hypotheses particularly important.
Across both horizons, researchers ask why a pattern should exist, whether it can be acted upon and if it’s likely to persist amid changing markets.
“We live and die by the out-of-sample R-squared of our predictions,” Max said. “We cannot afford to fool ourselves with stories about our models. We cannot afford to fall in love with our hypotheses.”
That discipline is crucial because returns are extremely noisy. Even strong models may explain only a minuscule fraction of the variation in returns. Alpha research often entails extracting a small amount of signal from an abundance of noise, then applying that signal with enough breadth for the edge to accumulate.
“We are trying to squeeze out just a little bit of predictive skill above chance,” Max said. “If we can make many thousands of uncorrelated predictions with a little bit of skill in each one, that starts to add up into something commercially meaningful.”
An alpha research project begins with an idea about how returns behave, how markets respond to information, how a dataset might be used or how a model could be trained more effectively.
The researcher then has to express that idea concretely, which could involve constructing a feature, designing a loss function, building a forecasting model or extracting structured information from unstructured data.
Momentum provides one example of how this thinking works. A researcher might begin with the observation that stocks moving in one direction can sometimes continue moving in that direction. But the researcher then has to figure out why the pattern should exist.
Perhaps information spreads gradually through the market, or maybe large investors divide trades into smaller pieces to manage transaction costs. Each explanation implies different data to examine and predictions to test.
“The reason to go through this exercise of thinking through the data-generating process is that it gives you something to latch onto in terms of how you would test your hypothesis,” Max said. “What is my theory, what predictions does it make and how can I try to disprove the theory?”
If the hypothesis survives that stage, the work becomes more quantitative. Researchers evaluate the strength and consistency of the relationship. Then they test the signal inside simulations that account for transaction costs and other market frictions.
“You want to know whether the signal is commercially actionable and profitable when you account for a more realistic view of costs and frictions,” Max said.
The final test is generalization. Researchers hold out data during development and use it only near the end of the process to guard against overfitting, look-ahead bias and spurious relationships. What matters is how well the model’s predictions perform on unseen data.
Alpha research at EQR is highly collaborative. Researchers work alongside colleagues in risk modeling, market impact, optimization, trading systems and quantitative development. Each discipline contributes a different part of the investment process, and the value of a signal depends on how well it fits into the broader system.
Alpha researchers need to understand how their forecasts will be used by optimization, the costs associated with trading on them and how to implement them efficiently and reliably. They also need to respond when live performance differs from expectations.
Research doesn’t end when a signal begins trading. Live performance is another source of evidence, helping researchers refine existing models and generate new hypotheses.
“There is a saying that you should never let a good drawdown go to waste,” Max said. “But maybe we should be more symmetrical and investigate the cases where something overperformed expectations, too.”
Unexpected results, positive or negative, can reveal that an assumption has changed, a model is behaving differently than expected or a new relationship is emerging. Those observations can improve an existing signal or spark the next research idea.
Strong technical foundations are essential. Alpha researchers need statistical depth, programming ability, modeling fluency and an affinity for working with large datasets.
But the technical side is table stakes. Strong researchers exhibit judgment, creativity and the discipline to abandon an idea once the evidence turns against it. They can move between abstract hypotheses and concrete implementation. They can ask why a market behavior should exist, identify data that might reveal it, build a model to capture it and determine if it has commercial value.
Equally important, they also know when to stop. If an idea doesn’t work, they have to be willing to move on.
“We cannot really afford to fall in love with one problem,” Max said. “The world changes. The markets change. The problems change.”
EQR gives researchers meaningful responsibility early. Max describes a collaborative culture in which decisions are pushed to the most junior person possible, fostering a more nimble organization that actively develops future research leaders.
New researchers have colleagues with deep expertise across disciplines at their disposal and work on problems that directly affect investment outcomes. They’re expected to take ownership of their work by staking out positions and then building and testing their ideas.
As EQR has grown, the research agenda has expanded with it. Increased scale creates new questions about how strategies interact with markets, while generative AI and agent systems enable researchers to rethink the research process itself.
For Max, that points to a broader frontier where we build more innovative ways to conduct research while we’re building better models.
As Max put it, the work’s arc runs from idea to live trade.
“We generate ideas for how to predict returns,” he said. “We implement them as models. We simulate them, and if they look good, we trade on them.”
