software-engineer-uses-ai-to-pit-goldfish-against-wallstreetbets-in-50k-investment-contest

Software Engineer Uses AI to Pit Goldfish Against WallStreetBets in $50K Investment Contest


Software engineer and YouTuber Michael Reeves used machine learning to pit a goldfish against the WallStreetBets subreddit in a contest to see who could make more money in the stock market.


Anybody who’s spent a few dozen hours on YouTube has probably seen at least one advertisement for an investing “guru” who can make you millions! in the span of a few days. But who really has a “green” thumb when it comes to making money off the stock market? Software engineer Michael Reeves has the answer: his pet goldfish—which was legitimately able to out-trade the WallStreetBets subreddit.

In the video above Reeves—who first went viral back in 2017 with a robot that is exceptionally good at shining a laser into one’s eye—shows how he created the tech for his experiment; walking viewers through how he “taught” his goldfish, Frederick, to trade stocks and let it decide the YouTuber’s “financial future.”

In essence, Reeves’ Frederick-driven stock picker works by using a camera to track the goldfish’s position in the tank. Then, when the stock market opens every morning, Frederick is “prompted” with two options for stocks to buy; e.g. Tesla (TSLA) and Apple (APPL). Frederick then “picks” his stock by spending more time on whichever side of the tank represents one of the buying opportunities. A computer then takes that “decision” and buys the stock.

While Reeves’ explanation of how he designed and built the rig and tracking software for the goldfish stock-picker is a thrilling ride through algorithms and more algorithms, things heat up when he builds a bot to “scrape” the WallStreetBets subreddit for another automatic fund manager to pit against Frederick.

The competition really gets going around eight minutes into the video, with Reeves and somebody else (a friend?) going ballistic as they watch Frederick compete with the WallStreetBets hive mind’s best picks. Or at least the bot’s best guess as to the Hive Mind’s best picks.

Image: Michael Reeves

In the end Frederick comes out—categorically—on top. While the WallStreetBets bot loses a total of $6,091.55 by the end of the experiment, Frederick actually makes $1,007.59. Which just goes to show you that even in a bear market, the best thing you can do is just keep swimming.

Feature image: Michael Reeves

(Visited 3 times, 1 visits today)

Accessibility Toolbar