The Hype Window Hypothesis
Newton first started with a hypothesis about gravity. Then Einstein hypothesized about relativity. Now we've released our "Hype Window Hypothesis". Read on to find out more about how we combined gravity, relativity, and an endless stream of stonks memes to get there.
Hype-othesis
We began by creating a hypothesis that there was a way to statistically model a series of metrics that would indicate when a stock would fit the bill of YOLO. Our hypothesis was driven by the fact that the magnitude of retail traders on social platforms had demonstrated that it could arbitrage against the big players. Often the big players have more capital, axioms they adhere to, and access to more data and insights.
Our hypothesis was simple - there was a period of 2-4 weeks prior to a stock going to the moon that we could see initial rumblings. We called this period prior to the stock popping off the “hype window” which is where we derived our name.
We explored 5+ data sources, including previous attempts at doing similar activities through Shared Git repositories (think retail quant nerd herd), yahoo finance data, Alpha Vantage and several others.
Alpha Vantage offered us the most reliable and “workable” data set. As tech guys we defined workable as timely, current, accessible (like an API), format-able (no weird encoding or risk of changing interface), and license-able. Not looking to spend our free time in any SEC depositions here….
We tested close to 20 different metrics, including commonly used (e.g., price, volatility), somewhat common (I.e, momentum), and our own derivatives based on our exploratory data analysis.
Our finding confirmed our hypothesis. With the early YOLO stocks, there was ~ a two week window prior to them popping where a user could get in and ride the hype!
Hype-othesis 2.0
This allowed us to test our next hypothesis, which is that there is a series of metrics which can be used to statistically model against a predicted behavior to identify new securities that are behaving similarly to the early YOLO stocks.
In our next blog, we’ll talk through how we engineered key features in an algorithm to score, rank, and present the candidates that we think algorithmically correlate to exhibiting YOLO behavior.
Today, that algorithm includes 10 key features, and we’re looking to add more. If you have suggestions on how to improve the interface, ideas on data you want more accessible, or new metrics you want to see daily, please reach out to me at jvogel@hyperider.io!
And if you have no ideas, but still want to support, feel free to get started with our product here.
HypeTeam 6, out!