Stuy Team: How We Won Our Way to the International Math Tourney Finals

Stuyvesant High School juniors Katie Wong and Andrew Zhang prepare for the Mathworks Math Modeling (M3) Challenge. Photo courtesy of Mathworks

Editor’s note: Katie Wong and Andrew Zhang, members of a Stuyvesant High School math team that recently won a coveted place in the finals of a highly competitive international math competition, wrote this piece for the Trib about the team’s experience during the semi-finals. The team, which emerged as one of only nine teams out of 770 to reach the finals, lost in the last round on April 27, an in-person presentation of their paper to a panel of PhD mathematician judges. They did, however, win a technical computing award.

In March our team of five Stuyvesant High School juniors—Andrew Zhang, Cyrus Yau, Daniel Li, Jayden Kim and  Katie Wong—spent 14 consecutive hours writing a 20-page paper for the Mathworks Math Modeling (M3) Challenge, a competition with 770 submissions from the U.S. and United Kingdom, each competing for $100,000 in scholarships. Our work was chosen as one of nine best solutions, booking us a spot in the finals.

We heard about this competition from our math teacher and modeling advisor, Patrick Honner. Our team had participated in modeling competitions before, but M3 was very different. 

While the previous competitions had taken place over multiple days or weeks, this time we had only half a day to produce a comprehensive paper on a subject we had no prior knowledge about.           

At some point, everyone in a math class wonders: “Where will I ever use this?” For us, the answer was here: Working to quantify and analyze the scale and consequences of online sports gambling. That was the challenge presented to us on that Sunday morning.

The time limitation proved to be our first hurdle. Although we all agreed that meeting in-person would be better than an online call, some of us had research competitions or violin lessons that weekend. So we compromised and agreed to meet on Sunday from 9am to 11pm, with Andrew joining later at 6pm.

There was a lot of discussion in the initial planning stages of those 14 hours. When you have four people simultaneously proposing idea after idea, it’s pretty hard to form a single, coherent plan of action. The google doc we used to gather ideas filled up pretty quickly, and by the end of our first brainstorming session, it was a jumbled mess of bullet points and half-formed thoughts.

Instead of trying to perfect our plan, however, we moved forward with what we had. After identifying what our main goals and focuses would be and identifying key variables such as geographical location and demographics like age and gender, we began testing possible approaches—how these variables could be represented in a mathematically simplified solution.

We then split up responsibilities, which is key for a competition like this. There’s a lot more than just math that has to go into a math modeling paper. Of course, the model itself has to be developed to turn real-life behavior like loss-chasing into numbers that we could compute, and code written to implement, fine-tune, and run it in order to generate simulated data. However, there’s also research, writing, and organizing that has to be done to introduce and analyze the model.

By lunchtime, we had a decent first draft of our paper for the first section and model. Although the progress was encouraging, we were already pretty exhausted, so the next two sections felt daunting, and, admittedly, we procrastinated a little before starting work on the next section.

When we resumed work, however, we ran into more conflict. Our group was split into two different opinions on how to handle it, and we proposed two different potential models. One major point of contention was how to handle the time frame of one year of gambling loss that the model would have to simulate. Breaking time into chunks—a discrete model—would make the model easier to implement but might not capture smaller behavioral changes. We also debated how much we should break up these “time chunks” into, trying to find a balance between a realistic, accurate model, and something achievable within our time constraints.

Similarly, we had different ideas for the very open-ended third and final section, which asked us to quantify “impact.” We debated whether this should be represented as direct financial loss, or recovery time, or opportunity cost, among other ideas.

We resolved both of these conflicts through debate, each of us supporting our position and challenging each others’ until we came to agree on the models we ended up using as the most feasible and accurate options. After that, work on the two sections continued throughout the afternoon and evening. 

As the day began to draw to a close, the pressure of the looming deadline intensified. By the evening, we were exhausted, running on snacks and determination. In the final minutes, we rushed to check for mistakes, fix formatting, and make sure the paper read cohesively.

The final moment we submitted felt anticlimactic. Instead of a dramatic celebration, we sat in quiet relief after the long, mentally exhausting day.

All of us were pretty surprised when we received the news that we qualified for the MATLAB Technical Computing award a few weeks later. None of us had ever used MATLAB (a computing program by MathWorks) prior to this competition, so learning and using it under time pressure, and then getting recognized for it, made the experience feel very meaningful. We’re super grateful to M3 for providing us with the resources to access the software and the incentive to learn how to use it and integrate it into our solution and paper.

The next step for us was to present to a live panel of Ph.D level mathematicians at the final event. These are the highest standards that we had to meet, and we put countless hours into preparing our presentation.

Looking back, it wasn’t just the results of the semi-final competition that made it worth it. The process itself—making and executing decisions quickly in a group setting, defending and debating ideas under pressure, and turning hastily gathered and incomplete information and ideas into a coherent, workable paper—is what really stood out. Math modeling competitions, especially M3, helps us approach math as more than just a set of problems with a fixed answer, but a dynamic tool that can be used and thought about in many different angles