Inspiration
Twitter's rise has minted a massive amount of people who want to glean market insight from Twitter posts. It's a great idea in concept - Twitter is a real-time unfiltered look into the psyche of its users - and should be perfect for the task. There are entire services built around the concept [http://stocktwits.com/]. The large number of people trying to mine information from Twitter has actually added volatility to the market. If the official account of say, George Soros, tweets that Coca Cola is overvalued, shares will be hit instantly as HFT algorithms pounce on the information. This process is inherently volatile.
What it does
Our project is an easily deployable stealth twitter bot designed in Python that attempts to reduce market volatility by influencing sentiment towards the mean. It is designed to run 24/7 and has a simple schedule. During non-trading / business hours the bot favorites and retweets posts by influential financial figures and institutions to try and accumulate credibility/influence. During pre-trading hours the bot tweets at these same figures/institutions the exact opposite of the movement from the previous day. I.e. if the Dow-Jones increased by 4% market, the bot will tweet that there is going to be a steep correction and that we're going to enter a bear market.
How I built it
The entirety of the bot is built in Python. Market data accessed through Quandl's open API and processed with Numpy. We used Tweepy as a fast way to handle authentication and request generation for Twitter's API.
Challenges I ran into
The major challenge is establishing credibility for an account that is clearly fake. The process of slowly favoriting and retweeting over time is meant to somewhat mirror human behavior. The inherent randomness also might help fool spam filters. The other challenge was working within the request limits that both the Twitter and Quandl APIs used.
Accomplishments that I'm proud of
Learning how to work with Twitter's API, especially the OAuth2 protocol was very gratifying. Also learning how about event scheduling in Python was really cool.
What I learned
Scheduling events in time-sensitive code completely changes the way you have to design applications.
What's next for StrattonOakmont
The sky is the limit for StrattonOakmont. We will continue to improve our bot in order to eliminate the scourge of market volatility once and for all. Perhaps a bot network is the next step.
Log in or sign up for Devpost to join the conversation.