Peter Lynch, the legendary Fidelity fund manager, suggested one way to find stock market winners was to look for companies that aren’t covered by Wall Street analysts. As they are discovered, the theory goes, shares will take off.

“Find a company that no analyst has ever visited, or that no analyst would admit to knowing about, and you’ve got a double winner,” Lynch wrote in One Up on Wall Street.

Howard Lindzon is trying something similar using signals from social media.

Lindzon is the founder of StockTwits, a social media platform used by traders and investors. StockTwits was launched in 2008 as a version of Twitter for people passionate about markets. The platform has 4 million users, with 500,000 active every day.

What Lindzon realized is that the volume of comments and number of traders who follow a particular stock is a proxy for interest. Companies with low levels of “watchers” are comparable to those with scant analyst coverage.

To test that theory, Lindzon has started publishing a daily newsletter edited by Riley Rosebee called Trends With No Friends. They screen for companies showing relative strength and then for companies with low watcher counts on the StockTwits platform.

I worked on a similar project at Bloomberg. That function screened for companies showing large increases in the volume of tweets with little price moves. The idea was that investor interest wasn’t yet reflected in prices.

Lindzon is also embarking on a general overhaul of the platform since he returned as CEO after a hiatus. (Full disclosure: I have been working with StockTwits as an advisor on the project.)

Another feature the StockTwits team has rolled out is price prediction surveys. One from last week asked users to predict the Reddit IPO price. More than 240 people weighed in.

Lindzon said the reception given Reddit’s IPO has made him more bullish that tapping into StockTwit’s vibrant community of traders and investors can provide insights that can be turned into content.

Reddit’s offering also highlights the opportunity to use niche content to train language models. (Google paid Reddit $60 million for model training.)

Reddit COO Jennifer Wong told Bloomberg that Reddit’s 19 years of data that has been compiled and vetted by human moderators, organized by topic and tagged for relevance is valuable for building AI applications.

Riley, the newsletter editor, said one interesting observation is that many of the companies with big gains and small followings are manufacturing or service companies, not the well-known tech firms that dominated media coverage in the past decade.

Nor are all small. Costar is a $40 billion company that sells commercial real estate data. The stock just hit an all-time high and yet only has 1,000 StockTwits followers.

Here are some of the other companies StockTwits has flagged with big gains and small follower counts.

$MMYT +50% YTD
$APPF +42% YTD
$SUZ +12% YTD
$BCC +16% YTD
$PLMR +51% YTD
$PHIN +26% YTD

It’s a reminder that there are so many reservoirs of data that investors can tap.