bellcrve | Devpost

Inspiration

We are very interested in the intersection of the financial sector and engineering, and wanted to find a way to speed up computation time in relation to option pricing through simulations.

What it does

Bellcrve is a distributed computing network that simulates monte carlo methods on financial instruments to perform option pricing. Using Wolfram as its power source, we are able to converge very fast. The idea was it showcase how much faster these computations become when we distribute across a network of 10 mchines, with up to 10,000 simulations running.

How we built it

We spun up 10 Virtual Machines on DigitalOcean, setup 8 as the worker nodes, 1 node for the Master, and 1 for the Scheduler to distribute the simulations across the nodes as they became free. We implemented our model using a monte carlo simulation that takes advantage of the Geometric Brownian Motion and Black Scholes Model. GBM is responsible for modeling the assets price path over the course of the simulation. We start the simulation at the stocks current price and observe how it changes as the number of steps increases. The Black Scholes Model is responsible for pricing the stocks theoretical price based on the volatility and time decay. We observed how our simulation converges between the GBM and Black Scholes model as the number of steps and iterations increases, effectively giving us a low error rate.

We developed it using Wolfram, Python, Flask, Dask, React, and Next.JS, D3.JS. Wolfram and Python are responsible for most of the Monte Carlo simulations as well as the backend API and websocket. We used Dask to help manage our distributed network connecting us to our VM's in Digital Ocean. And we used React, and Next.JS to build out the web app and visualized all charts in real-time with D3.JS. Wolfram was crucial to our application being able to converge faster proving that distributing the simulations will help save resources, and speed up simulation times. We packaged up the math behind the Monte Carlo simulation and deployed it as Pypi for others to use.

Challenges we ran into

We had many challenges along the way, across all fronts. First, we had issues with the websocket trying to connect to our client side, and found out it was due to WSS issues. We then ran into some CORS errors that we were able to sort out. Our formulas kept evolving as we made progress on our application, and we had to account for this change. We realized we needed a different metric from the model and needed to shift in that direction. Setting up the cluster of machines was challenging and took some time to dig into.

Accomplishments that we're proud of

We are proud to say we pushed a completed application, and deployed it to Vercel. Our application allows users to simulate different stocks and price their options in real time and observe how it converges for different number of simulations.

What we learned

We learned a lot about websockets, creating real-time visualizations, and having our project depend on the math. This was our first time using Wolfram for our project, and we really enjoyed working with it. We have used similar languages like matlab and python, but we found Wolfram to help us speed up our computations signficantly.