Stonklytics

Prediction for a seemingly unpredictable volatile market.

Comment

Inspiration

We drew inspiration from a very recent consequence of unpredictable, volatile behaviour in the stock market. The story of Robinhood, a popular financial services app, was perhaps the biggest inspiration. In the wake of dramatic changes in stock prices, Robinhood faced a large bill, which led to a sudden buying restrictions for users of their app, leading to the legal action they now face, and the unpredictable losses that their customers had to take. We believed that, surely, we could leverage data and provide protection for retail investors and the apps they use, by allowing them to be better prepared.

What it is, and what it does

First, I must state what it isn't. It isn't a stock price predictor - that wouldn't be us trying something new. Instead, it is a predictor of volatility of a stock, based on a carefully selected range of parameters from social media. Currently, we focus on r/wallstreetbets/, the forum at the heart of the volatility that brought down Robinhood and impacted its users. We believe due to current events that this one forum alone can provide us with very valuable insights.

We display our insights in the form of a 3D mobile app, developed in Unity. It almost has a game-like feel, and displays a large volume of data effectively on one mobile screen. These are near real-time insights, with our database updating every few minutes to give as much time as possible to prepare for sudden changes.

How we built it

The first part of our hack - the analysis - was created using a python scraper, which cleans and prepares the data, then feeds it to the IBM sentiment analysis service. The processed data and sentiment analysis is stored in a Google Firebase Cloud Firestore database.

This database then links to a 3D Unity iOS app for visualisation, which reads its data from Firebase.

This app displays the data, which has 4 dimensions, on a 3D graph, with the 4th dimension depicted by the size of the data-point. We designed our data-points in Blender.

Challenges we ran into

The first major hurdle was working together virtually across different timezones. Though we did all try to adjust our sleep schedules a little, what we found most effective was regular check-ins and providing updates over Discord, especially before sleeping/after waking up. This allowed us to collaborate effectively and create a finished hack.

The second hurdle was using GCP. Several issues occurred - firstly, a webGL app did not connect easily with a Firebase database, meaning we had to rewrite our Unity scripts for iOS. Then, we realised there was no way to host this app, with the frontend, on Firebase, so we had to create effective ways to present its user-friendly nature through a recording. Later, we hit our GCP Firebase operations limit - not even the mentors could help us with this, and we did not realise how quickly we were reaching the limit to prevent this from happening. To overcome this, we quickly rescheduled our workload, so we only needed to use Firebase the following day.

Accomplishments that we're proud of

We are proud of our visualisation, especially given none of us had worked on Unity for iOS apps before, and only one had ever used Unity for any 3D development at all. The aesthetics, the colour scheme, and the ease of use overall make it a very beautiful piece of work to us.

We are also proud of how well we worked together. Given none of us had ever met before, we came from different educational institutes, and had a diverse range of skills, we still produced good code in a short span of time.

What we learned

It is clear that we learnt new coding skills - Unity has been mentioned already, we also learnt how to work with GCP for sentiment analysis, though due to time pressure we used IBM Watson in our final real-time app.

But we learnt another often ignored skill - presentation. We participated in a presentation skills workshop running during the hackathon, and learnt through 40min of trial and error how to effectively convey our idea and the market value we bring. This was something we would not normally have access to outside of this hackathon, and we greatly appreciated the opportunity.

What's next for Stonklytics

The app is ready to hit the ground running! There are a few other features we desire, such as the ability to change the general colour scheme and layout of the visualisation in the app to improve accessibility for those who need it. However given we have already laid the foundations of the UI, it is a very quick addition.

The best part is - we are using real data. So we don't need to swap out any dummy datasets. We even have a domain ready, linked below, and we have our demo videos and a demo website ready for hosting on our GitHub page! Unfortunately due to unforeseen hosting issues, we weren't able to host properly in time, but we look forward to showing you live our app live in action.

Share this project:

Updates