But the technology, at least in my understanding at this point, is nothing more than running your standard regression on a proprietary dataset.
It sounds like technology, because cool-sounding mathematical prediction is the selling point... But it isn't that hard to do (though with data, and time, you can surely incrementally improve it). The story here is a community-known credential. Its value comes from other people relying on it as a baseline. Like any other credential, it doesn't guarantee you'll be any good - but it is a better-than-random predictor.
Because there isn't anything like this available already, and it would be a useful thing to have, they have a good chance. For it to be used as a reference point, it doesn't need to be great - it just needs to be betterer- enough -than random (and get adopted).
A "good predictor" just means "5% better than totally random". If you have a strong team, and good connections... your chances are surely better than if you had a weak team and no connections (though exceptions occur in both ways). If you can formalize a bunch of these factors, and find what they weigh... you have a baseline predictor that's better than totally random.
If I could predict the market, I would be investing like crazy! :)
But really I'd be skeptical about data in this first batch of tests. I'd bet that YouNoodle used these widely known startups as a training set for its algorithm, or at least as a test set...I imagine if one of their unit tests came back and said "Facebook is gonna be worthless", someone would tinker with the algorithms until it didn't say that anymore. Usually you deal with this problem by setting aside some data points until the very end for you to evaluate your accuracy, but without knowledge of where they got their training set, you can't really do this yourself
Mighty oaks begin as small acorns.
To make a counterpart for startups seems interesting, and worthy. With the growing market of successful startups, the industry that supplies them with services such as this one should become viable.
If the technology is as good as is stated in the TC article couldn't they pretty rapidly build up a listing of undervalued companies on Wall Street and buy into them?