World's Largest Hedge Fund Uses Twitter For Real-Time Economic Modeling

Tyler Durden's picture

The use of Twitter and other social media to predict and trade on or, reflexively, generate interest in various assets is nothing new and has been around for years. Whether or not this strategy works is still unclear: the only "hedge fund" that traded purely on Twitter signals, Derwent Absolute Return, shut down shortly after opening (despite supposedly generating positive returns). Superficially, the thinking behind this made sense: as Venturebeat reported previously, based on research done at Indiana University, published in early 2011, there was a strong correlation between sentiments expressed on Twitter and the direction of the Dow Jones Industrial Average. According to the study, “Twitter mood predicts the stock market”, Twitter sentiment correlates with the ups and downs of the next few days on the DJIA with 87 percent accuracy. A separate study at Pace University in 2011 found that social media could predict the ups and downs of stock prices for three global brands, Starbucks, Coca-Cola, and Nike. And sentiment analysis has become an indispensible part of marketers’ toolkits, thanks to companies like Radian6 and Webtrends.

Expectedly, as more and more amateurs have piled into Twitter, the data stream has been subject to the "Yahoo Finance effect" - there is far too much noise, and not nearly enough actionable signal, especially when one tries to strip away the bias behind any given message (see "Trading Twitter: Where Noise Becomes Signal").

Yet one entity that appears to have found significant functionality in Twitter is none other than the world's biggest hedge fund: Bridgewater.

According to Bridgewater's Greg Jensen, speaking during a recent client conference call, the world's largest asset manager (except for the Fed of course) with one of the best long-term track records in history, has been using "everything that is available" online, from social media to real-time Internet prices, to model economic activity in what is effectively real-time. As Jensen said, "from Twitter and Facebook (and so on) we can capture every time somebody is saying they bought a new car. We could add those up and can compare that to the stats and be really on the pulse of what’s going on with something like auto sales or, similarly, with home sales."

Perhaps even more interesting is Bridgewater's search for equivalents of the famous Billion Prices Projects which tracks real-time prices of goods and services around the globe. Specifically, Bridgewater notes that it uses sites like the "Amazon of India" to track inflation "during a balance of payment crisis on a moment-to-moment basis" and thus confirm if any sharp currency moves have filtered down to end prices just by monitoring the internet.

Bridgewater's end goal: to be "able to track the economy on a day-to-day basis." Which in a world of high frequency trading, in which millisecond responses to stimuli is critical for alpha-generation, is paramount. It perhaps also explains why traditional periodic data releases such as inflation data, car sales and other formerly market moving macro releases, no longer pack the punch they once did when it comes to market response.

So with Bridgewater blazing the trail in real-time data monitoring, it is only a matter of time before all other macro hedge funds engage in the same strategy of near-constant monitoring of all concurrent data feeds.

At which point the next logical question is: how long until competitors begin introducing artificial and misleading noise in the data stream, and attempt to confuse Bridgewater and others' signal translators. And how soon before data analytics firm XYZ comes out with its latest offering: one million fake twitter accounts, all of which are programmed to amplify fake economic signals and confuse Twitter algos that translate signal to trades? For a very hefty price of course...

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
maskone909's picture

quick! Everyone make dummy accounts and post sell bias towards bonds and PM's! 

aVileRat's picture

What do you think the tech guys do.

Fun fact, if you use the 'fav.' button, it not only sends an alert to the account owner's email (think of it like a doorbell on bloomberg) but also triggers a dummy variable on trend tracker API's. Hence why high 'dummy' account twits like Leo D. or Justin B. will have over 2,000 fav's per tweet to 'stuff' the trend API's. 21st C. google rank hacking.



maskone909's picture

i am the master of spam.  i used to clear 4K per month on adsense!  game on!

insanelysane's picture

Still can't beat NSA's ability to "archive" emails and other communications from companies that are trading.

maskone909's picture

who do you think shorted the airlines just prior to 9 1.1.

doomandbloom's picture

Because we are in a fractal holographic universe, with each particle having information about the whole ( e.g. atom looks like a solar system) we could frankly use any point to predict anything (if we knew how).

I often joke that the pen that i am holding has information about the lottery numbers that will be come out later today. Only if I knew how to read it.


maskone909's picture

all is well until you break it down into the quantum level, quantum entanglement ect...  but i get what you are saying.  its all a computer simulation any ways :P

NidStyles's picture

Except humans have these things called brains that make it possible for them to remove themselves from the natural animal behaviors that fit your description. Human action can not be modelled and if it ever is successfully modelled it will mean state enforced slavery for all.


Take the hint and quit advocating that BS before you enslave the next generation.

insanelysane's picture

I would disagree.  Human action has the capacity to act in an independent manner at any time.  However, since humans are social animals with a large percentage having herd tendencies, human actions can be modelled with a fairly high confidence level.

Joeman34's picture

"fairly high confidence level"


It's what happens outside of the confidence interval that's important and why your premis doesn't hold...

ParkAveFlasher's picture

i can haz datumz?  k thx bbai

Non Passaran's picture

Hey, I just bought 53 tons of (edit: physical) gold today!

... waiting for the POG to go full bulltard in 3..2..1...

Sheet, no wonder Twitter's market cap is worth more than a thirld world country.

Al Huxley's picture

This alone would seem to justify a much higher price for TWTR.   OT, I have to say, I'm extremely disappointed in the market managers' performance over the past couple of days.  As an investor I expect, in fact RELY on my daily 1 - 3% bump in the indexes, with correspondingly better moves in the high-beta names.  I don't know what the fuck these guys are doing, but I hope they get it together and get these indexes green by close today.  Thank god at least Facebook's behaving properly.  Maybe they're just building suspense prior to the next big move up, so it will be even more fun and exciting...

NoDebt's picture

Market timing signals taken straight from the echo chamber.  Good luck.

gjp's picture

This from the guys who think we are deleveraging in the U.S., and that there is something 'beautiful' going on in the economic/political world.  And now Twitter-based mood research.  Total geniuses, aren't they, they deserve every bit of their billions. /s

buzzsaw99's picture

because #ASKJPM sucks cock in Hell?

roadhazard's picture

I'm shocked, again.


Anyone that puts their personal life on the interwebs and important info on their computer deserve whatever happens to that info.

monad's picture

=Bridgewater is scraping ZH

rosiescenario's picture

The model will work very well and draw in ever larger amounts of "investment money" right up until an off the chart black swan event hits wiping out, with interest, all the prior gains.


Long Term Capital comes to mind....

carbonmutant's picture

Is that why TWTR is a record highs?

analyticsnut's picture

value in twitter audiences