Google Has Own $26.5bil Trading Floor, Predicts Market With Search Data? | Hacker News Reader

Google Has Own $26.5bil Trading Floor, Predicts Market With Search Data?

ozcarguide.com

50 pointscl3m22 comments

There is a story told, I think in Ken Auletta's book, by Eric Schmidt from the earlier days of the company where Sergey proposes starting a hedge fund because they have so much better data than anyone else. Schmidt cautions against the notion due to perhaps significant legal implications.

Something that would perhaps counter this are complaints on zerohedge and elsewhere that Goldman's privacy policy on some of its data services leave open the possibility for them to use usage data for prop trading.

There must be people within Google wondering that if others are getting away with this, why can't they? There must also be people worried about the company's dependence on a single source of revenue...

Aren't they kicking off the trading floor to manage their cash, rather than to becoming a hedge fund/investment bank ?

I remember reading an article in the recent past where they have hired some ex-IB traders (Goldmans, Lehmans, etc) to run their Treasury department.

The new hires were due to the fact that they were changing their risk profile of their portfolio. I read they were looking for some bond traders as they're planning on moving towards corporation bonds rather than treasuries/govvies.

I need to look for that article.

http://news.ycombinator.com/item?id=1385741

http://www.businessweek.com/magazine/content/10_23/b41810335...

After reading "The Black Swan" by Taleb and finally understanding some of the issues with predictability. These kind of claims just sound like astrology to me.

brown9-2-2 is right the latter part of the headline is pure speculation.

Careful. "The Black Swan" is a pop non-fiction book, and reading it won't give you an intimate understanding of markets. Taleb has taken his share of criticism (http://www.efalken.com/papers/Taleb2.html) and his ability to capitalize on the deep insights he'd like you to believe he has is somewhat questionable (http://www.businessinsider.com/did-nassim-talebs-universa-ge...)

Let me put it like this. If the market where predictable we would live in a very different world.

So the proof is in the pudding. No one predicts markets unless they somehow control them.

The latter part of this headline is nothing but speculation in the article.

As to the question of whether Google could exploit their search data to predict the stock market: http://messymatters.com/searchpred (summary: not really)

Slightly less short summary:

In some things search can add a slight benefit to predictive analysis (from the current baseline strategy), in others it doesn't really help at all and even in the cases it does it's only marginally useful, "we find search volume on its own is predictive of future outcomes, but search is nevertheless outperformed by baseline models trained on publicly available data; combining search and baseline models generally leads to modest improvements."

Is Google following RedHat?

http://saviorodrigues.wordpress.com/2009/03/09/is-red-hat-a-...

Hmm I remember reading somewhere before about Google Venture's acquisition of (or rather, investment in) a startup company who was working on a product that could predict future events. Maybe with this Google can predict markets as well. link: http://www.nydailynews.com/money/2010/05/04/2010-05-04_googl...

Could Google be using their search data for...

equals

easiest way to drive traffic to your site.

Quite frankly, data that is scraped and analyzed by the algos and quant funds is probably much more relevant, but it does not mean Google could not define a strategy based on search data; however, it is pure speculation and link bait to just throw the assertion out there willy nilly.

Agree 100%, without actually showing how search data would be used it's kind of pointless. At first though it seems OBVIOUS that they'd be using search data, but when you actually think about it, what would the application be? Would the market effects in search show up fast enough to be relevant to the market? And for HFT it's useless.

Would be more that interested in being proven wrong.

This is nonsense - I know people in quant funds who have tried to do things like this. All the boring stuff that Google doesn't care about, like SEC filings and Quarterly Reports, matter far more than aggregate attention. I'm sorry but if you're making decisions for a multi-million dollar fund, and you tell investors that your investments are based on the twitterings of the unwashed masses, you'll be sent home before you've had your lunch.

Even if there is value in their data (and I do imagine enough PhD's could find some reasonable strategies) - there would be such a public outcry that it wouldn't be worth it from a PR perspective.

In terms of them predicting markets with search data lots of commentators are saying no way, but I strongly disagree.

Google can predict flu-like illnesses within 92% accuracy of CDC data which is dervied from labratory testing.

Furthermore trading stocks seems to me to be a behavior that would be easy to predict through Google's "intent-based" algorithms used for serving advertising.

EG - For instance knowing how many searches there are a day for "is AAPL overpriced" vs. "is AAPL underpriced" you'd have data that would begin to correlate to market value - with enough data points you could make some pretty darn educated guesses about what a given stock would do.

Toss in a hundred PHDs, a few dozen traders from the greediest depths of wallstreet and the question shouldn't be CAN they do it, but WHEN will they do it.

Reference:

http://pagingdrgupta.blogs.cnn.com/2010/05/18/how-accurate-i...

> For instance knowing how many searches there are a day for "is AAPL overpriced" vs. "is AAPL underpriced" you'd have data that would begin to correlate to market value - with enough data points you could make some pretty darn educated guesses about what a given stock would do.

You may be right but the only thing that really moves the market is the large institutional funds. None of these funds are going to do a google search for "Is apple overpriced" and then say OMG we've got to sell apple now.

Fact: some non-US companies explicitly forbid their people to search using google, especially patents through google's service.

Interesting fact, happen to have a source?

NDA, but remains a fact, nevertheless...

Fact: if you provide no documentation and use hand-wavy terms like "some non-US companies," your speculation can sound much more like factual information.

You're right of course, institutional traders are not searching Google or Yahoo Answers for "should I buy or sell AAPL?" However - strong analytics could tell you some interesting habits if you analyze search strings coming from IP blocks known to be used by said institutional traders.

For example, I imagine institutional traders do a lot of research on companies before taking a position in them - this research might show up in search analytics data and Google traders could place a position before the institutional traders do.

Would this be considered illegal by the SEC? I have no idea, but I could definitely see Google doing this.

The search data is probably less useful than all the other data they have regarding how much money individuals and corporations will pay for various words. Or all the private email they process for "other purposes relating to offering you Gmail".

Reply on news.ycombinator.com