Top 31 Optiver Data Scientist Interview Questions + Guide in 2024

Optiver is a renowned trading firm with a reputation for leveraging technology and mathematical prowess to excel in the financial markets. Established in 1986, Optiver stands out for its collaborative culture and commitment to continuous learning.

As a Data Scientist at Optiver, you will be tasked with developing models and tools that drive trading strategies and optimize operations. This role demands a high level of expertise in areas like probability, statistics, machine learning, and coding. Candidates should be prepared for a rigorous and multi-faceted interview process that includes a series of online assessments, technical interviews focusing on mathematical and statistical challenges, and behavioral interviews to evaluate cultural fit and motivation.

Whether you’re intrigued by the intersection of data science and financial markets or up for the challenge of solving complex problems in a high-stakes environment, this guide will help you navigate the process and commonly asked Optiver data scientist interview questions. Let’s get started.

The interview process usually depends on the role and seniority; however, you can expect the following on an Optiver data scientist interview:

Recruiter/Hiring Manager Call Screening

If your CV passes the initial screening, a recruiter from the Optiver Talent Acquisition Team will contact you to verify key details like your experiences and skill level. This call typically covers why you want to work at Optiver, what you know about the company, and a summary of your career history. Behavioral questions may also be part of the screening process.

The recruiter call should take around 30 minutes.

Online Assessment

Optiver often starts the interview process with an online assessment, which can take 2 to 4 hours. The assessment includes a range of difficult mathematical and programming questions, games, and probability puzzles. Questions may cover numerical calculations, coding problems, logical reasoning, and mental math tests.

Some candidates find this part of the process very rigorous and time-consuming, so preparing adequately is essential.

Technical Virtual Interview

After passing the online assessment, candidates are usually invited to a technical interview. This round can include questions on probability, brain teasers, and programming tasks. For example, you may be asked to solve Fermi problems, calculate expected values, or work through regression tasks.

It’s common for this stage to include multiple rounds with different interviewers focusing on different technical aspects of the job. This interview typically lasts 1 hour.

Take-Home Assignment

Some roles require a take-home assignment where you need to solve problems related to data analysis or quantitative research. The assignment might involve complex topics such as regression analysis, volatility calculations, or similar quantitative problems.

In the final round, you may need to present your findings from the take-home assignment and explain your thought process.

Onsite Interview Rounds

If you move past the technical virtual interview and take-home assignment, you’ll be invited to an onsite interview day or a series of virtual interviews. The onsite process usually includes several rounds, such as:

• A reasoning assessment
• HR interview for behavioral questions
• Multiple technical interviews focusing on various skills such as probability, coding, quantitative analytics, and data analysis
• A final round which can include a simulation exercise like trading

The set of interview questions we’ve compiled for Optiver’s data scientist role is designed to comprehensively evaluate a candidate’s suitability for a data-driven and dynamic trading environment. These questions range from behavioral inquiries, like understanding the rationale behind career transitions, to technical assessments, such as proficiency in SQL, statistics, and machine learning concepts.

Read the details below for a comprehensive understanding of these interview questions.

1. Why did you move from jobs A to B in the past years?

Recruiting for a data scientist role costs time and money. Optiver doesn’t want to bring in people who are likely to be disloyal, nor do they want to bring in under-performers. This question seeks to establish if you have the potential to be a long-term hire.

How to Answer

The goal is to put the events in a positive light. Be honest, factual, and confident, and avoid talking negatively about your previous place of employment. Stick to the facts, leave out personal opinions, and divulge only as much information as is required.

Example

“My skillset did not suit what my employer needed then. It had been assumed that someone with my skills would be a good fit when I was hired. It later became apparent that they needed someone with domain knowledge, and we decided it was best to part ways after one year.”

2. What other companies are you applying to?

Interviewers ask this question to gauge whether you’re serious about the role you’re applying for and the industry you’re looking to work in. Passion for the industry is a common theme in Optiver interviews. Interviewers also want to know whether you are likely to accept their offer.

How to Answer

Be brief and honest when answering this question while showing that you’re excited about the prospect of landing the role. Avoid saying you have no interest in any other company or showing that this company ranks low on your list.

Example

“I have applied for a similar position at Akuna Capital, and I also have an interview coming up at Tower Research. Even so, I believe Optiver, with its values and work culture, offers the right mix of challenge and reward that I need right now to become an even better data scientist.”

3. Why are you interested in trading?

Trading is the main activity at Optiver, and their data scientists will be expected to develop insights that determine what trades they make and how they make them. Interviewers will want to establish if your motivations and objectives in this regard are aligned with the Optiver’s.

How to Answer

Demonstrate that you know what the trading industry is about and what comes in the package when you sign up for a job in that industry. Secondly, show how your skills and experience align well with the role and industry. You want to communicate a strong but sincere passion for trading.

Example

“I like the fast pace of the trading industry and the idea that the value of my work will be reflected in many of the daily decisions. This industry is dynamic with constantly changing inputs, and there’s always an opportunity to get regular feedback. It would be hard to find an environment that encourages a similar level of learning and growth in other industries.

4. How would you handle missing data in a dataset?

Dealing with missing data is part of the fundamentals of data science. You could be asked this type of question at your Optiver data scientist interview to test your grasp of such fundamentals.

How to Answer

There are established methods of dealing with missing data depending on the specific situation. Give the interviewer some common methods, plus where and why you’d use them.

Example

“I’d assess how much missing data there was in proportion to the whole dataset, what specific data was missing, and what assumptions could be made about the data. If the data is randomly missing and forms a small part of the dataset, the affected observations could be removed. I may also replace the missing data with the mean or median values if the missing data is a small part of the dataset.”

5. What is the definition of variance?

A data scientist at Optiver is expected to grasp basic statistics well. Variance is one of many metrics you may be asked to define to test your statistical knowledge.

How to Answer

Give a direct answer to the question, explain variance, and give its general equation.

Example

“Variance is a measure of how spread out the points in a dataset are from their average value. It is calculated by summing the squares of the differences of all values and the mean and then dividing the result by the sample size minus 1.”

6. What is the probability of having a double-headed coin given ten consecutive heads?

This question evaluates your understanding of probability theory in a scenario involving a combination of fair and biased coins. It is applicable for a data scientist at Optiver because it mirrors the intricacies faced in financial data analysis, where recognizing patterns and probabilities from diverse data sources is customary.

How to Answer

Approach this with Bayes’ theorem. Define the prior probabilities of picking each type of coin and calculate the likelihood of getting ten heads with both a fair and double-headed coin. Use these to determine the posterior probability of the coin being double-headed.

Example

“In a jar of 1000 coins with one double-headed coin, the chance of picking the double-headed coin is 0.001. With ten heads observed, applying Bayes’ theorem, we find there’s about a 50.6% chance that we’ve picked the double-headed coin. For the next toss, the probability of getting heads, considering both coin types, is approximately 75.3%.”

7. What is the probability of getting two 4s in 8 dice throws?

Data scientists at Optiver have to make decisions under uncertain conditions. A good understanding of probability distributions helps in such situations, and these types of questions test if you possess this knowledge.

How to Answer

Identify the distribution type represented by the problem and simply plug the relevant variables into the formulas.

Example

“The binomial distribution applies here because we are trying to calculate the probability of an event in a certain number of trials. The values needed for the equation are number of tests, n=8, the number of successes, k = 1, and the probability of getting a 4 in each roll = ¹⁄₆. If we plug the values into the equation, we get that the probability of getting a four twice in 8 throws is 0.2605.”

8. How would you analyze focus group data for new TV series pilots to determine which to feature?

This tests your ability as a software engineer to analyze qualitative data, a skill Optiver requires for interpreting complex market data and making informed decisions. It reflects on processing and deriving insights from diverse data sources.

How to Answer

Detail how to analyze the rating data, considering factors like average ratings, variability, and audience preferences. Discuss using statistical methods to compare and rank the pilots.

Example

“I’d calculate the average rating for each pilot and assess the variability of ratings. High average ratings with low variability might indicate a universally appealing pilot. I’d also consider audience segments to see if certain pilots appeal strongly to specific groups, which might be crucial for niche markets.”

9. Which is higher in Japan, the mean or median age?

As an Optiver data scientist, you must know about probability distributions and other fundamentals. This type of question tests this, focusing on the common measures of center.

How to Answer

Information on the age distribution in different countries can be found on sites like Statista. This will give you an idea of what shape the data will form when plotted. You can also infer the information from other information you already know.

Example

“Present-day Japan has one of the lowest birth rates in the world. This has likely resulted in the older generation forming a larger subset of the population than the younger one. In that case, the population distribution is likely a skewed left-normal distribution with the median age higher than the mean age.”

10. How do you evaluate the success of an A/B test regarding free shipping mentions on a product page?

Understanding and conducting A/B testing is critical in a data-driven environment like Optiver’s. It’s crucial for optimizing strategies and making data-backed decisions, similar to testing trading algorithms or market strategies.

How to Answer

Discuss how to analyze the results using statistical methods to determine if the observed difference in conversion rates is significant. Mention considering factors like sample size, conversion rates, and statistical significance testing.

Example

“I’d compare the conversion rates of both groups using a hypothesis test, like a chi-square test, to see if the difference is statistically significant. If the experiment group with the free shipping mention shows a significantly higher conversion rate, then the test is successful.”

11. How would you explain the idea of a p-value to a non-statistician?

Many people who make data-driven decisions at Optiver don’t have a statistical or data science background. Interviewers want to know if you can effectively communicate these ideas to them.

How to Answer

Using an example, you can start by explaining the nature of statistical testing and the concepts of the null and test hypotheses. Instead of simply defining the p-value, this approach allows you to put the concept in a context that is easy for a non-statistician to relate to or understand.

Example

“If we had a new algorithm and needed to test if it was, in fact, better, we would work with two hypotheses, a null and a test hypothesis. The null hypothesis would state that the new algorithm does not perform better, while the test hypothesis would say it does perform better. In this case, the p-value would measure the likelihood that we’d get the same results if the new algorithm wasn’t better. This null hypothesis is usually rejected if the p-value is less than 0.05.”

12. How can you explain neural networks to children in a simple and understandable way?

During an Optiver data scientist interview, this question assesses your capability to distill complex ideas into understandable terms. This skill is vital in a trading firm setting, where explaining sophisticated algorithms and trading models clearly and concisely is key to collaborative success. It ensures effective communication within teams and aids in making complex strategies accessible to all stakeholders.

How to Answer

Use simple, relatable analogies to explain neural networks. Focus on conveying the idea of learning from examples, like how the brain works, without delving into technical details.

Example

“I’d compare a neural network to the process of learning to recognize animals. Just as children learn to identify animals by seeing many examples, a neural network learns patterns from data. Each ‘neuron’ in the network helps to make a decision, similar to how each child in a group might have a small piece of information about an animal.”

13. How would you test if a new product is likely to significantly increase the value of a company?

As an Optiver data scientist, you’ll test different hypotheses to develop new models. This question tests if you know about hypothesis testing and the steps involved in the process.

How to Answer

You’ll need to develop a hypothesis and then explain the process you would use to test that hypothesis. The goal is to demonstrate that you understand the steps and objectives of hypothesis testing.

Example

“In this situation, my null hypothesis would be that the product is not likely to significantly increase the company’s value, and my alternative hypothesis would be that it would lead to a significant increase in the company’s value. A significance value of 0.05 would be sufficient, but we can also use 0.01 depending on the sample size and variance. After that, I’d do the computations for the sample mean, sample standard error, and, lastly, the t-statistics. We can finally compare the p-value from the computations with the significance value, rejecting the null hypothesis if it’s lower.”

14. How would you model electricity supply to avoid over or under-supply in a town?

This question evaluates your competence in time-series forecasting for resource management, an important skill for a Data Scientist at Optiver. It assesses your capacity to analyze and forecast crucial resources, similar to financial forecasting in trading environments, where accurate predictions are necessary for strategic decision-making.

How to Answer

Describe using time-series models like ARIMA, which can handle seasonality and other factors affecting electricity consumption. Discuss how these models predict future demands based on historical data and patterns.

Example

“I would use an ARIMA model to forecast electricity supply. This model is ideal as it accounts for past consumption trends and seasonal variations, like increased usage in winter due to heating. I’d analyze historical electricity usage data, adjusting for seasonal factors to create a forecast that balances the supply with expected demand, minimizing the risk of outages or wastage.”

15. Should you roll a die the second time, assuming that you currently earn $4 due to rolling a 4 the first time and will forfeit the $4 and earn whatever you get from the second roll?

For market makers like Optiver, the expected value of securities can be useful in determining if a trade is worth making. This question tests if you know about this concept and how to apply it in data-driven decision-making.

How to Answer

The outcomes of dice rolls fall under uniform distributions. This means we can calculate the expected value and determine the values below that are worth it to take the risk of a second dice roll. The law of large numbers can justify whether or not to roll.

Example

“No, the player should not roll the second time. Thanks to the law of large numbers, the average value of dice rolls will converge toward the expected value or their weighted average. For a dice roll, this equals ¹⁄₂ of 1 plus 6, which is 3.5. The expected value is lower than the current $4, which means the player has a poor chance of improving their earnings on the next roll.”

16. Why might overall approval rates decrease even when individual product approval rates increase or remain the same?

This question tests your ability to analyze and interpret statistical data, a critical skill in a trading environment like Optiver. Understanding phenomena like Simpson’s Paradox, which can occur in this scenario, is crucial for accurate data analysis and decision-making in financial markets.

How to Answer

Approach this question by considering Simpson’s Paradox, where individual trends differ from the aggregated trend. Analyze how the distribution of applications across products can affect the overall approval rate, even when individual rates are stable or improving.

Example

“The overall decrease could be due to a change in the distribution of applications among the products. If more applications are for the product with lower approval rates, the overall rate can be dragged down. In this case, even though Product 2’s approval rate stayed the same if it received a significantly higher proportion of applications in the second week, it could lower the overall approval rate.”

17. Between two models with 85% and 82% accuracy, respectively, which one would you choose?

Optiver values a nuanced understanding of model performance beyond just accuracy. This question probes your ability to critically evaluate models, a key skill in developing and refining trading strategies.

How to Answer

Discuss the importance of considering other metrics like precision, recall, and the specific use case. Emphasize the need to understand the context and data before deciding solely based on accuracy.

Example

“Though one model has higher accuracy, I’d examine other metrics like precision and recall, especially in a trading context where false positives can be costly. If the 82% model exhibits higher precision in predicting market movements, it could be more suitable despite its lower overall accuracy.”

18. How would you develop a model for generating respawn locations in a third-person shooter game?

This tests your ability to develop dynamic algorithms and models, skills directly applicable to creating and adjusting trading algorithms in a fast-paced, data-driven environment like Optiver’s.

How to Answer

Describe your approach to modeling a dynamic and balanced respawn system. Mention the use of machine learning or algorithmic techniques to ensure fairness and unpredictability in respawn locations.

Example

“I would create a model that analyzes past player data to determine low-conflict areas and uses this information to generate unpredictable respawn points. Machine learning could be used to adaptively learn from in-game events, ensuring a fair and engaging player experience.”

19. How would you design a podcast search engine that includes transcript and metadata analysis?

This question relates to designing a system, a crucial skill at Optiver. Although it is presented in a podcast search engine context, it demonstrates the fundamental principles of database design and machine learning system design. It highlights your ability to handle large datasets and create efficient search algorithms.

How to Answer

Discuss how you would structure the database to store podcasts, transcripts, and metadata. Explain the algorithm or ML technique you’d use to analyze and rank search results. Emphasize the importance of efficient indexing and retrieval methods.

Example

“I would design a relational database to store podcasts, their transcripts, and metadata. For the search engine, I’d use natural language processing (NLP) to analyze the transcripts and metadata, employing techniques like TF-IDF for keyword extraction and relevance scoring. The search algorithm would rank podcasts based on keyword relevance and user preferences.”

20. How long will it take for the second car traveling at 80 mph to catch up to the first car traveling at 60 mph, given it starts an hour later?

This question evaluates the candidate’s competency in using mathematical and analytical skills to solve real-world problems. These abilities are essential in a data-driven environment like Optiver, especially for optimizing strategies and performing precise calculations.

How to Answer

Use linear equations to represent the distance traveled by each car. Set up an equation equating the distances traveled by both cars and solve for time. Remember that the second car starts an hour later.

Example

“The first car’s distance over time is 60t, and the second car’s is 80(t-1). Equating these gives 60t = 80(t-1). Solving for t, we find that the second car catches up after 4 hours from its start.”

21. What would you expect to happen after a new UI is applied to all users? Will metrics actually go up by ~5%, more, or less?

This question tests your ability to analyze and interpret statistical data, a critical skill in various analytical roles. Understanding phenomena like the impact of sample representation and statistical validity is crucial for accurate data analysis and decision-making in business environments.

How to Answer

Approach this question by considering the validity of the A/B test and potential biases in the experiment. Analyzing the representativeness of the sample and understanding the confidence intervals are essential for predicting the true impact of the new UI.

Example

“The 5% increase might not translate directly to the entire user base. If the test sample isn’t representative or other variables are at play (e.g., testing only on weekends), the actual impact could be different. The metric might increase, but possibly by less than 5%, depending on these factors.”

22. What is the probability that none of the three zebras, each at a corner of an equilateral triangle and randomly running along its edges, collide when a lion attacks?

This question tests your ability to apply mathematical reasoning and probability concepts to solve hypothetical scenarios. These skills are vital in data-driven environments like Optiver, where precise calculations and strategic optimizations are crucial.

How to Answer

Consider the directions each zebra can choose and calculate the probabilities of each possible outcome. Recognize that the zebras will either run in a clockwise or counter-clockwise direction to avoid collision.

Example

“Each zebra can choose to run either clockwise or counterclockwise. The probability of all three choosing clockwise is (¹⁄₂)×(¹⁄₂)×(¹⁄₂)=¹⁄₈. The same probability applies for all three choosing counter-clockwise. Adding these probabilities gives ¹⁄₈+¹⁄₈=¹⁄₄. Thus, the probability that none of the zebras collide is 25%.”

23. Create a function find_iqr to find the interquartile distance of an array of unsorted random numbers.

Given an array of unsorted random numbers denoted nums, write a function find_iqr to find the interquartile distance. The interquartile distance is defined by subtracting the first quartile from the third quartile.

24. Write a function compute_deviation to return the standard deviation of each list in a list of dictionaries.

Write a function compute_deviation that takes in a list of dictionaries with a key and a list of integers and returns a dictionary with the standard deviation of each list. This should be done without using the NumPy built-in functions.

25. How would you assess the validity of the result in an AB test with a .04 p-value?

Your company is running a standard control and variant AB test on a feature to increase conversion rates on the landing page. The PM checks the results and finds a .04 p-value. How would you assess the validity of this result?

26. How would you create control and test groups for Instagram Stories’ close friends feature to account for network effects?

You want to test the close friends feature on Instagram Stories. How would you make a control group and test group to account for network effects?

27. How does random forest generate the forest, and why use it over logistic regression?

Explain how random forest generates multiple decision trees and why it might be preferred over logistic regression in certain scenarios.

28. When would you use a bagging algorithm versus a boosting algorithm?

Compare two machine learning algorithms and provide examples of tradeoffs between using bagging and boosting algorithms.

29. How would you compare two credit risk models for personal loans?

1. Identify the type of model developed by your co-worker for loan approval.
2. Explain how to measure the difference between two credit risk models over time.
3. List metrics to track the success of the new model.

30. What’s the difference between Lasso and Ridge Regression?

Describe the key differences between Lasso and Ridge Regression techniques.

31. What are the key differences between classification models and regression models?

Outline the main differences between classification models and regression models.

Job positions at Optiver are very competitive. You’ll have to put your best foot forward to make it to the final interview and get an offer from the company. Here are some tips on how you can prepare for their interviews.

1. Practice

Practice answering as many questions as possible related to the role of data scientists. These include questions on behavior, coding, statistics, estimation, data analysis, hypothesis testing, etc.

The data science learning path and takehomes on Interview Query are valuable resources you can use to practice questions on different aspects of data science.

2. Research the Role of an Optiver Data Scientist

Tailor your resume and interview answers to the specific roles of data scientists at Optiver. You’ll have to research the work that data scientists at the company do and the tools they use.

Interview Query posts regular blogs that can tell you, among other things, what the work of a data scientist in the industry is like.

3. Show Your Thought Process

Many interview questions are designed for you to demonstrate your problem-solving skills. When answering technical questions, ensure the steps you take to arrive at your answer can be understood and your assumptions justifiable.

You can get personalized guidance and tips on how to do this when answering questions through Interview Query’s coaching services.

4. Brush Up On Mental Math

Mental calculations are heavily tested. Practice quick additions, multiplications, divisions, and fractions under time constraints.

5. Practice on Quantitative Problem-Solving

Be prepared for brain teasers, probability puzzles, and Fermi problems. Practice estimation and guesstimate style questions, as these are commonly featured.

6. Understand the Trading Domain

Familiarize yourself with basic trading concepts, regression analysis, and other relevant financial terms that might come up during the interview process.

7. Try a Mock Interview

Mock interviews allow you to see how you perform when answering questions in an interview. They can help you identify areas you will likely struggle with in the real interview.

Interview Query hosts regular mock interview sessions where you and other users are matched for mock interview sessions.

Candidates applying for data scientist positions at Optiver frequently ask these questions. Let’s discuss each one of them:

How much does a data scientist at Optiver earn on average?

What qualities and skills are Optiver looking for in a Data Scientist?

Optiver seeks candidates with strong quantitative skills, proficiency in statistical and mathematical modeling, coding expertise (particularly in Python), and the ability to solve complex problems quickly and accurately. They also value candidates who demonstrate motivation, teamwork, strong communication skills, and the ability to handle constructive feedback.

What is the culture like at Optiver?

Optiver is known for its friendly yet highly competitive and fast-paced environment. Teamwork and collaboration are vital, and the company highly emphasizes continuous learning and improvement. Feedback is crucial to their culture, and employees are encouraged to think critically and take intellectual risks.

Does Interview Query have a page discussing the data scientist role at Optiver?

No, Interview Query doesn’t have a page for discussions on the data scientist role at Optiver. However, you can still visit the discussion board to read and learn from other candidates’ interview experiences for roles in other companies or add your own.

IQ’s Slack community lets you chat with other data scientists and experts in related fields.

Are there job postings for Optiver data scientist roles on Interview Query?

No job postings for data scientist positions in Optiver on Interview Query exist. However, you can find postings for similar roles in other companies, including Oracle, Amazon, Capital One, and Workday.

Optiver offers a unique chance to thrive in the fast-paced trading and financial markets. Completing their exhaustive interview process successfully can pave the way for a fulfilling and challenging career. Good luck with your interview journey—it’s demanding, but achieving it promises a rewarding career ahead!

You can check out our main Optiver Interview Guide to find out more about what it’s like to interview for any role at this company. We’ve covered their data analyst and software engineer positions extensively, similar to this one, so be sure to check it out to gain additional ideas.

You can also explore our comprehensive guide, ‘How to Prepare for the Data Science Interview,’ and browse through our extensive list of Data Science Interview Questions for additional insights and preparation resources.