Citadel Datathon & Data Open Guide

What is the Citadel Datathon and Data Open?

The event brings students together to investigate a challenging dataset, extract useful insights and defend an analytical recommendation. Teams combine programming, statistics, visualisation and clear communication rather than competing on a trading profit leaderboard.

Citadel and Citadel Securities use Datathons to meet strong technical students and assess how they approach ambiguous problems with other people. Correlation One helps administer the application and assessment process.

How the Data Open name fits

The Data Open was the earlier name for Citadel's datathon competitions, including regional events and a championship. The current organiser pages use Datathons, so use the Datathons event page to find active applications rather than searching for an old championship entry form.

A Datathon is also different from Citadel Securities Terminal, which is a coding tournament using automated game strategies. Neither event should be confused with the ordinary recruitment interview process.

Eligibility, upcoming events and application dates

Eligibility depends on the individual Datathon. Start with the university region, degree status and expected graduation date for the exact event you want to enter.

Next listed event

• Event: Women's Datathon, in person in Miami, early February 2027.
• Application deadline: 11 January 2027.
• Selection notification: 19 January 2027.
• Application: official Women's Datathon application.

These are the dates available as of 4 November 2026. For later Europe, Asia, India, Summer Invitational or PhD events, select the new event from the official calendar before using a prior edition's schedule.

How to apply and get selected

Start by completing the event application with accurate education and contact details, a current resume and the requested background information. The Miami application also asks about your role interests and offers a place to add links to technical work such as GitHub or a research profile.

1. Choose your event. Match the university location, degree and graduation window, then open its application form.
2. Submit your application. Include the requested resume and experience details, then retain the confirmation.
3. Complete the assessment. Look for the Correlation One email, open the quiz invitation and finish the required 60 minute test by the stated deadline.
4. Confirm participation. If selected, complete the required paperwork, accept the event instructions and arrange the team and travel details.

Citadel's general Datathon selection process considers academic performance, assessment performance, qualifications and interest in investing. Completing the quiz is part of the application, rather than an optional practice exercise.

Citadel Datathon assessment: topics and preparation

The entry quiz lasts 60 minutes. Prepare for short technical reasoning problems involving probability, statistics, linear models, data manipulation and coding logic, with some questions combining several ideas.

Prioritise fundamentals and careful interpretation before advanced machine learning. For example, be able to explain why a random train and test split can leak future information when observations form a time series.

Original probability practice question

EverythingQuant practice: a detection system flags 80% of genuinely faulty records and incorrectly flags 10% of clean records. If 5% of records are faulty, what percentage of flagged records are actually faulty?

A realistic timed practice benchmark

As an EverythingQuant preparation exercise, complete 20 original short problems in 60 minutes. Divide your mock between probability and statistics, regression and model reasoning, and data manipulation, then use the final five minutes to revisit uncertain answers.

Track errors by topic and whether they came from incorrect mathematics, code reading, missed assumptions or rushed interpretation. Repeat the weakest topic the next day instead of only completing more generic puzzles.

How the team competition works

After selection, participants are helped into teams. Citadel's Datathon programme terms describe teams of three or four people working with a supplied dataset on a defined topic, then presenting an original analysis to a panel of judges.

Each event provides its own challenge brief, time limit, submission instructions and judging criteria. The current Miami event is in person, while earlier Data Open competitions have also used virtual formats, so prepare around the format of the event you enter.

1. Understand the challenge. Restate the question, target outcome, evaluation criteria and constraints before touching the data.
2. Audit the dataset. Check units, duplicates, missing values, outliers, coverage and whether variables could reveal the answer improperly.
3. Produce defensible insights. Start with descriptive comparisons, then use an appropriate statistical model or experiment to investigate the strongest hypothesis.
4. Stress test the conclusion. Compare against a baseline, change key assumptions and identify the largest source of uncertainty.
5. Present the result. Make the finding, method, impact and limitations understandable to judges and prepare for questions.

What to bring and what data you can use

Bring your own laptop and charger and be ready to work with data supplied by the organisers. The general programme rules permit organiser data and appropriate public data, but prohibit confidential third party information and misuse of proprietary material.

Python with pandas, R or SQL can help with cleaning and exploration, while simple charts work well for the presentation. Treat these as practical preparation tools rather than a claim that a particular programming language or library is mandatory.

How to build a strong Datathon submission

Build the analysis around one defensible question instead of collecting unrelated charts. The strongest presentation connects a real decision to a method, a measurable finding and a careful statement of what the data supports.

Divide work by deliverable

A practical four person team can assign initial ownership to data quality, modelling, visualisation and presentation. Everyone should still understand the central claim, the data definitions and the limitations well enough to answer a judge.

Agree on column definitions, notebook or file ownership and a shared results table at the start. Save final chart values and calculations in one reproducible notebook so the slides cannot drift away from the analysis.

A six hour EverythingQuant team drill

Use this as a preparation exercise, not as the official length of a Citadel event.

• 30 minutes: frame the research question and identify the outcome measure.
• 60 minutes: clean the data, inspect distributions and create a data dictionary.
• 75 minutes: build descriptive comparisons and a baseline.
• 60 minutes: test one meaningful model or hypothesis.
• 60 minutes: run robustness checks and document uncertainty.
• 45 minutes: assemble the presentation around three decisive figures.
• 30 minutes: rehearse the talk and field skeptical questions.

Worked example: did a bus lane reduce commute time?

EverythingQuant illustration: a city introduces a priority bus lane in one district and compares it with a similar district without the change. The fictional dataset contains average commute times before and after the intervention.

Both districts improved, but the pilot district improved by an additional six minutes. This is the difference in differences estimate, which removes the shared three minute improvement seen in the comparison district.

import pandas as pd

data = pd.DataFrame({
    "district": ["pilot", "pilot", "comparison", "comparison"],
    "period": ["before", "after", "before", "after"],
    "mean_minutes": [42, 33, 40, 37],
})

values = data.pivot(index="district", columns="period", values="mean_minutes")
changes = values["after"] - values["before"]
extra_reduction = changes["pilot"] - changes["comparison"]

print(extra_reduction)  # -6 minutes
assert extra_reduction == -6

What the result means

The initial finding is a six minute additional reduction in average commute time for the pilot district. A strong submission would now check sample sizes, differences in commuter composition, route changes and whether both districts were following similar trends beforehand.

The estimate is descriptive until those assumptions are tested. A concise presentation would show one comparison chart, a sensitivity test using peak hours only, and a recommendation to gather more observations before claiming a causal effect.

Turn the analysis into a five slide presentation

1. The question: define what a successful transport intervention would achieve.
2. The dataset: give coverage, time period, data quality issues and the comparison group.
3. The result: show the raw changes and the additional six minute reduction.
4. The checks: explain alternative causes, sensitivity analyses and remaining uncertainty.
5. The decision: recommend a next test or policy action justified by the observed evidence.

Judging, prizes and career opportunities

Datathon teams present their findings to judges, who evaluate the work under criteria supplied for the event. Aim for a correct, original and useful analysis that can survive questions about assumptions, data quality and alternative explanations.

Citadel describes cash prizes and eligibility to interview as potential rewards. Under the general Datathon programme terms issued in 2023, standard team awards were US$10,000 for first place and US$2,500 each for second and third place, with a US$20,000 first prize for the PhD competition.

Those figures describe the older programme terms, not a guaranteed award schedule for the upcoming Miami event. Use the current event instructions for its actual prizes, team allocation and submission criteria.

Does winning lead to a Citadel job?

Strong performance can create access to interviews and conversations with recruiters, but a prize does not automatically become a job offer. If quantitative research is your goal, continue with the Citadel Securities quantitative research interview guide to prepare for the separate hiring process.

Travel and practical costs

For eligible selected attendees at the Miami Women's Datathon, the linked programme terms describe economy travel, airport transfers, hotel accommodation and certain meals, subject to their conditions. Participants remain responsible for personal expenses, applicable taxes and any required travel documentation.

A focused preparation plan

Prepare for two distinct tasks: the individual entry assessment and the later team challenge. The first rewards fast, accurate technical reasoning; the second rewards reproducible analysis, teamwork and a convincing presentation.

1. Day 1: review conditional probability, Bayes' rule, expected value and common distributions; solve 15 short questions without a calculator.
2. Day 2: practise interpreting linear regression coefficients, model assumptions, hypothesis tests and the impact of confounding variables.
3. Day 3: load a public dataset in Python or R, filter rows, join tables, handle missing values and explain the result of each operation.
4. Day 4: complete the 60 minute EverythingQuant mock benchmark, review every mistake and revisit the two weakest topics.
5. Day 5: work with a team on the six hour analysis drill, reproduce the figures from code and deliver a five minute presentation followed by questions.

Before the actual competition, rehearse answering why the outcome matters, how the data was collected, what assumption could break the result and what you would investigate next. Judges should be able to follow the reasoning without reading your code.

Citadel Datathon FAQs

Official applications and useful resources

For a wider overview of interview preparation, explore EverythingQuant guides. For Datathon entry, always start with the specific event application and its rules.