On the 23th of November 2020, the twelfth edition of the Caseday was held. With four companies from different fields there were again plenty of opportunities to meet your potential future employer. This edition marks the first ever Caseday that was held completely online.
IMC - by Duy Nguyen
For me, the morning started with the IMC case which was given by three IMC employees: a recruiter, an experienced trader and a working student. Before the case began, groups of two were formed. I got teamed up with Max, another third year student. The case consisted of three parts. First, we had to create our own trading algorithm. We encountered so many technical issues that we could not finish it, but we did have an idea. We wanted to trade based on an ‘average valuation’, which is the average of all bid or ask prices weighted on the number of bidders or sellers for each price. For the second part, we had to evaluate our algorithm. We noticed that it did not perform very well. One of the assistants told us that this method could perform poorly when the number of buyers or sellers is very unbalanced. Suppose you have a lot of buyers, then the ‘average valuation’ will go down since their bid prices are lower than the ask prices. However, as demand rises, prices go up, so our valuation would be too low.
To my surprise, our algorithm performed the best
So, in case of a lot of buyers relative to sellers or vice versa, our algorithm will use different valuation methods (best bid or best ask price). These were provided by IMC packages. Our algorithm would make a trade based on this valuation. For the final part, the performance of each group’s algorithm would be measured based on another dataset. The one who performed best, received a price from IMC. To my surprise, our algorithm performed the best!
CBS - by Martijn van Heusden
The first case I visited was the CBS case. As most of us know, CBS publishes reliable statistical information for the Netherlands. It might not be the first place to work you think about when you’ve studied econometrics, but it is certainly a job with great relevance to society. During a short presentation by Jacco Daalmans, who also conducted the case, it was also discussed that you can see your work pop up in the media sometimes. Open up any major Dutch news website, and you're quite likely to see an article about a new CBS publication.
CBS’ case was a bit like a jigsaw puzzle
As for the case, we were tasked to complete a census based on a small data sample from the population of a fictitious country. CBS actually performs the census for the Netherlands using a small sample. Contrary to this, most other countries perform their census by interviewing the entire population. It was mentioned that in China, six million employees work towards interviewing every single inhabitant to get a reliable count. You can imagine that this is a very costly process and that we should be happy that we use statistical methods to obtain a reliable census for the Netherlands!
An important lesson to learn from the case is that data is never perfect. We had to combine different datasets, a problem CBS encounters a lot. It was a little bit like a jigsaw puzzle: once you make a mistake in the beginning, it might not be possible anymore to find a solution. CBS uses, among other methods, mass imputation which feels very intuitive. Some case groups actually used this concept in their solution.
In short: CBS’ case was a clear demonstration of the type of work they do and very interesting to attend to say the least.
Optiver - by Martijn van Heusden
The Caseday would not be complete without Optiver! As Optiver is such a major market maker, again a lot of Kraketters were very excited for this case. As a market maker, Optiver sets both a buy and a sell price on a certain product and can make a profit from the difference between these prices. Since news can have a great effect on the market, speed is of utmost importance.
Making an educated guess for such a puzzle quickly turned out to be much more important than calculating the exact answer
This absolutely showed during the case which followed. The case was in the form of an online market making game where the bid and ask spread we could propose were based around the current total “Big Five'' population. The Big Five population is the total number of elephants, rhinoceroses, buffaloes, lions, and leopards currently alive. As an example, selling 100 corresponded to guessing the total is lower than 100,000, and buying 100 corresponded to guessing that the total is higher than 100,000.
Throughout the game, multiple hints were given to represent the news Optiver is affected by all day and to test are ability to react quickly. Hints were small math puzzles which could give you an idea of the population of one of the five species, so we got closer and closer to the real answer over the course of the game which resulted in an ever decreasing bid-ask spread. Making an educated guess for such a puzzle quickly turned out to be much more important than calculating the exact answer, which could cost you valuable time. Overall, Optiver offered an exciting case which worked well in an online environment.
ORTEC - by Maarten Verhoef
On the 23rd of November this year, – I am writing this on December 31st, talk about dedication – 40 eager students put on some nice clothes, opened their laptop, and joined a very special edition of Kraket’s own Caseday. Well, we did see one of our participants quickly change when he saw the outfits of the others, but we’ll count it.
How should an airport schedule its flights, and at what gates? How can I efficiently stack my truck for transport? How can I most efficiently pack the product I want to sell?
This year, four companies joined the day, and gave some very, sometimes surprisingly interesting cases. Not only did I have the honour to maintain the contact with ORTEC, I was also able to join their case in the second part of the day.
We started with a presentation about Ortec itself, as is customary. Who are they, and what do they do? They are an international company, based in Zoetermeer, providing technical OR solutions for any problem you could imagine. For example, how should an airport schedule its flights, and at what gates? How can I efficiently stack my truck for transport? How can I most efficiently pack the product I want to sell? In other words, they Optimize Your World.
To get a more hands on feeling of how this would work in practice, we were pitted against each other in a little competition in the second part of the case. The assignment: find the most efficient route for a set of mechanics, operating under time and capability constraints.
I am proud to say Ivo and I came second in both rounds. On average a victory, and that with a smaller team. I do have to add, Jonathan accidentally leaving his mic open while discussing his tactics may have helped a little, be it by revealing what we should not do.
Conclusion: a terrific case provided by two more than capable ex-Kraketters.
Sneak peak: Next week, we will post a double interview with Vivian Lodewijckx en Ron Stoop from the 38th board of Kraket and Tristan de Keyzer and Amber Politiek from the current board. Find out what Kraket was like ten years ago and how doing a board year has changed a lot in this time!