Citadel Data Open and Terminal, explained: how college students compete for $15,000 and a chance to interview at Citadel

You do not need to attend a particular school. Eligibility is set by region, degree level and graduation year, applications go through Citadel's careers site, and participation costs nothing; for in-person events, selected students' travel is paid. The Data Open is a team data-science competition built around a real dataset and an open-ended question. Terminal is a team coding competition in which you write an algorithm that plays a tower-defense-style game against other teams' algorithms.

This post explains the two formats, who can enter, what past problems have looked like, how a competition runs from application to prize, what Citadel's judges and recruiters look for, and how to prepare — whether you are a student, a professor, or a student group leader deciding whether to put this in front of your members.

What's inside

1. What are the Citadel Data Open and Citadel Terminal?
2. Who can compete?
3. What do the problems look like?
4. How do the competitions work, from application to prize?
5. How do you get on Citadel's radar through the Data Open or Terminal?
6. How do you prepare?
7. Key facts
8. Frequently asked questions

What are the Citadel Data Open and Citadel Terminal?

Two different tests of the same thing — whether you can solve a hard, open-ended problem under time pressure with a team — one for data scientists and one for programmers.

The Data Open (Datathons)

A team data-science competition. Correlation One builds a set of real-world datasets around a question with no textbook answer, in domains such as financial markets, generative AI, renewable energy, education and urban systems. Over a competition window — about a week when virtual, one to three days when in person — teams explore the data, form a hypothesis, build their analysis, and submit a written report with their findings and code. A panel of judges scores the submissions and the top three teams are announced at a closing session. Regional Datathons have run in North America, Europe and Asia Pacific; the Women's Datathon, held in Miami each February, is the largest in-person event of the season.

Terminal

A games-based algorithm coding competition. Terminal is a turn-based, tower-defense-style strategy game in which you do not play by hand: your team writes an algorithm that places defences and units each turn, and your algorithm plays head-to-head against other teams' algorithms. Citadel describes it as a tower-defense-style algorithm coding competition against the world's top engineering students. Terminal runs virtually over a week with a flexible schedule, so teams work when they can, and results are decided by play rather than by a judging panel. The 2027 season introduces new AI variants of the game.

Year 11

Correlation One has run more than 70 competitions with Citadel and Citadel Securities across three regions since the series began, engaging thousands of students. Ken Griffin delivered the keynote at the 2017 Data Open Championship at the New York Stock Exchange. The 2027 season opens with the Women's Datathon in Miami in early February (applications close January 11, 2027), followed by regional Terminal competitions run virtually through the spring.

Who can compete?

Any current university student who meets the event's region and year requirements. There is no school list on the application: students from hundreds of universities across North America, Europe, Asia Pacific and India have competed, and the series was built to find talent that campus recruiting does not reach.

• Students. Most events are open to all current undergraduates; several regional Datathons have also admitted graduate and PhD students. Entrants must be 18 or older and in good standing. Each event sets its own region and graduation-year window. The 2027 Women's Datathon, for example, is open to undergraduate women at universities in the United States, its territories, Washington, DC, or Canada, with an expected graduation date between December 2028 and June 2030.
• First- and second-years. The competitions are designed to be entered early. Internship pathways are offered to high performers who are still one or two years from graduating, so a strong result as a sophomore is worth more, not less, than one as a senior.
• Teams. Both competitions are team events. You can apply with people you already work with or be placed on a team after selection; the organisers help form teams for students who apply alone.
• Professors and academic departments. Departments take part by sharing the application with their students, hosting a short info session run by Correlation One, or booking a one-to-one call with the competitions team. No sponsorship, fee or formal agreement is required.
• Student groups. Data science, computing, quant, finance and women-in-STEM societies are the most common route into the competitions. A group can circulate the invitation, run a practice datathon on a past problem statement, or field a team under its own name.

Selection is competitive and admissions are rolling, so applying early matters more than polishing the application.

What do the problems look like?

A Datathon problem is a real dataset, a real decision, and a question that does not have a single right answer. Every problem statement Correlation One has written for the series follows the same shape: a background briefing on the industry or policy issue; the task, framed as an open question; several linked datasets with their schema; submission requirements; an evaluation rubric; and tips. Past problems include:

• Ride-hailing versus taxis in New York City — trip-level records from the city's Taxi and Limousine Commission alongside the growth of app-based services; where and when did the market shift, and what explains it?
• Smoking, lifestyle and health outcomes — a national behavioural risk-factor survey; which life choices predict health outcomes, and how should a public-health budget be targeted?
• E-commerce in furniture and home goods — clickstream, order, product, competitor-pricing, weather and tax data for an online retailer; what drives conversion in a category where goods are bulky, expensive and non-branded?
• Hotels and the economics of hospitality — booking and pricing data across a hotel portfolio; how should a chain set rates and capacity?
• Access to clean water and development outcomes — country-level indicators; where does investment in water access produce the largest gains in health and education?
• Brexit and global markets — macro and market data around the 2016 referendum; how did the vote's result propagate through prices and expectations?
• Crime and policy in Chicago — the series' first individual competition (the Quant Challenge, 2023), in which each entrant produced a data-based policy proposal.

The datasets are large enough that you cannot read them by eye, messy enough that cleaning and joining them is part of the work, and open enough that two strong teams will ask different questions. One university partner described what wins as the combination of creativity, technical skill, speed and the ability to tell a story with data; a statistics department that hosted a Datathon noted that the open-ended problem closely reflects what data scientists face in industry and research.

Terminal has no dataset. The problem is the game: each turn your algorithm decides where to place defensive structures and which attacking units to send, with limited resources, against an opponent whose algorithm is doing the same. Teams start from a provided starter kit, test strategies in local matches, watch replays of their own games, and iterate across the week. The strategies that win adapt to what the opponent is doing rather than executing a fixed plan.

How do the competitions work, from application to prize?

Nine steps, start to finish. The whole cycle runs about six to eight weeks from the application deadline to the results.

1. Apply on Citadel's careers site. Each competition has its own application page under Programs & Events. The form asks for your details and a resume. Admissions are rolling.
2. Complete a short online assessment. After applying you receive a link to a brief skills assessment on Correlation One's platform — data and statistics for the Datathon, coding for Terminal. Finish it promptly; it is part of what Citadel reviews.
3. Selection and invitation. Citadel reviews applications and sends invitations. Invited students RSVP to confirm their place.
4. Team formation. Teams are confirmed before the event. Students who applied alone are placed on a team.
5. Kickoff. A live session explains the rules and, for the Datathon, walks through the problem statement. Datasets, the starter kit, and the submission guide are released.
6. The competition window. Terminal runs for about a week, virtually, on a flexible schedule. Virtual Datathons run about a week; in-person Datathons run one to three days, with travel and accommodation covered for selected participants who are not local.
7. Submission. Datathon teams submit a report with their findings and the code behind it by the deadline. Terminal teams submit their final algorithm.
8. Results. Datathon submissions are scored by a judging panel against the rubric and the top three teams are announced at a closing session. Terminal results are decided by match play.
9. Prizes and follow-up. The $15,000 prize pool is split among the top teams and paid by Correlation One. Every participant receives a certificate, and top performers are introduced to Citadel and Citadel Securities for the networking and interview opportunities described in the next section.

Participation is free at every step. Questions go to competitions@correlation-one.com.

How do you get on Citadel's radar through the Data Open or Terminal?

By producing a piece of work that Citadel's people see. Campus recruiting at a firm like Citadel starts from a short list of schools; a competition starts from what you submitted. Citadel's own description of the Datathon is that you compete to showcase your skills and earn a chance to interview, and of Terminal that winners earn access to exclusive job opportunities. Over the series' first ten years, more than 100 internships and full-time roles were offered to high performers.

Three things follow from how the events actually run.

You do not have to win. The prize goes to the top three teams; the recruiting conversations go to high performers, which is a wider group. The assessment you complete after applying, your team's submission, and how you present are all visible to the firm. A fourth-place team with a rigorous, clearly argued report is on the radar.

The rubric rewards judgement, not volume. Datathon judges score whether the team asked a precise question, chose a defensible method, handled the data honestly (including what it could not show), and communicated a conclusion someone could act on. A clean analysis of one question beats a sprawling tour of the dataset. For Terminal, the leaderboard rewards an algorithm that adapts to opponents and a team that iterates across the week; the replays show exactly how you played.

Early is better. Internship pathways are offered to students one to two years from graduation, and the firm sees the same people compete more than once. A first- or second-year who competes, learns the format, and comes back stronger is the pattern the series was designed around. An Irish participant who later joined Citadel Securities as an options trader described his first Data Open as the turning point, and he did not win it.

What Correlation One cannot promise, and does not, is an interview. Selection, interviews and offers are Citadel's decisions. What the competitions do is put your work in front of the people who make them.

How do you prepare?

Practise the format, not the domain. The problem will be in an area you have never studied; the skills that transfer are the same every time.

For the Data Open

• Be fast in Python or R with the standard data stack (pandas or the tidyverse, a plotting library, scikit-learn or equivalent) and comfortable joining several tables with messy keys. Most of the first day is cleaning and joining.
• Rehearse on a public dataset of the kind the series uses — a city's open taxi data, a national health survey, World Bank development indicators — with a two-hour limit and a one-page memo at the end. The memo is the skill: question, method, finding, limitation, recommendation.
• Agree team roles before kickoff: one person owns the data pipeline, one the modelling, one the write-up and figures. Switch roles on the practice run so everyone can cover.
• Read the rubric in the problem statement first and write your report to it.

For Terminal

• Install the starter kit as soon as it is released and get a local match running before you change anything. The game rules, map and unit costs are in the kit.
• Write a simple baseline strategy, then improve one thing at a time against your own earlier versions. Keep the versions.
• Watch replays — your own and, where available, past finals. The gap between a good team and a winning one is usually how quickly they adapt mid-match, not a cleverer opening.
• Split the work: one person on the defensive logic, one on attack, one on reading the opponent and tuning parameters.

For professors and student groups

• Forward the application link and deadline in the first week it is open; the single biggest reason strong students miss out is applying after capacity is reached.
• Host a one-hour practice session on a past problem statement; Correlation One will share one on request and can run a short info session for your department or society.
• If your group fields a team, put a first- or second-year on it. They are the students the pathway is built for.

Contact competitions@correlation-one.com for a past problem statement, an info session, or to confirm eligibility for a specific event.

Frequently asked questions

Related: Citadel Datathons (citadel.com) · Citadel Terminal (citadel.com) · Correlation One year in review 2023 · Announcing our fall 2021 competition schedule