Just when I thought algos couldn’t get any weirder…
Late last year, I posted an interesting paper about an algorithm designed to trade the Dow based on calmness levels on Twitter. Naturally, we were very skeptical of it:
Data mining. Correlation vs. causation. “87.6 percent accurate?” Yeah, I’m sure it is.
Not long after though, LIBOR posted about the The Derwent Absolute Return Fund, a fund set up to use that same algorithm and at the time, was just getting ready to launch.
I scoffed, I sneered, I thought it the weirdest thing ever… or at least until the Japanese caught wind of it.
Well, enter the Japanese…
“A Japanese banker and professor have created the Pluga AI Fund, which uses a web-mining system to track some 20 million blogs written in Japanese. The model, designed by the University of Tokyo's Yutaka Matsuo, seeks out keywords that may affect the Nikkei 225 Index.”
Shit, I think this one might actually work.
Targeting returns of 30% per annum (they returned 46.8% in simulated trading from November 2006 through July 2010), the Pluga AI Fund debuted in Japan last August and is now in preparations for overseas marketing come June.
What keywords the model is programmed to look out for was not discussed, but think about it; how many times have you read about GLD, SLV, AAPL, or whatever other security before and during their big moves? And as far as gauging market sentiment is concerned, I feel that blogs are a far better medium to use compared to twitter, so they really may be on to something here.
That said, I certainly won’t be putting any of my money in it though.
How about you monkeys? Any of you plunking down your hard earned cash on the blog fund?
Have a good one WSO.
This is really interesting Jorgé. I will be looking into it.
I'm thinking the algorithm makes a lot of sense if it's based on 20M+ blogs and has a 4 year return of about 50%. This return was over Nov 2006 - July 2010, which I think proves that it can produce returns through tumultuous times (which means the algo is very accurate in predicting the market). With the explosion in the number of retail investors (who absolutely love scavenging the internet for "stock picks") opening accounts with discount brokerages , I think this fund might prove very worthwhile.
My .02
First time post. AI is really starting to pick up, there's been a lot of usage on financial statements and whatnot, but more and more I see it embedded in all kinds of derivative models. I predict many similar offerings in the future.
I am currently working on an AI model that trades based on Jim Cramer's Mad Money. Expected annualized return = -50%
But seriously, intuitively the Japanese model makes sense. Blogs serve as the closest link between mind and media, "you think it you tweet it". Besides market sentiment and subsequent market moves are a self-fulfilling prophecy. So enough people are talking about something they're either already doing it or they're going to read about everyone else doing it and do it too. With enough data its bound to work.
I wonder what the algo will buy when everyone starts talking about the Royal nuptials - stock in the companies that make the souvenirs?...might even work lol
Google will rule the world one day, probably do already. First they steal this dude though from Univesity of Tokyo.
I can see such systems being exploited in pump and dump scenarios.
Couldn't someone just start a bunch of fake blogs full of all the trigger keywords and wreck the fund?
Yeah kinda like what happened to P&G and the DOW on 05/06/'10. If you had some sort of program that could automate fake blog entries, the Algo would read it and act on it. Is the Web 3.0 available to everyone or is it some sort of rarified bandwith?
haha i wonder if the anti-linkfarming idea from Google applies in this case
What about botnets ? They are relatively cheap and could be much more effective... program the bots to do google search requests for the companies you want to monkey around with and you could make all the algo's programmed to track such trends go haywire... Considering the money to be lost/made if big banks start heavily investing in such black boxes, im sure some eastern european computer geeks will consider creating such a system to make a quick buck.
+1 for the mention.
Interesting to see how this plays out. Although the term "web 2.0" is thrown around a lot, this technology mentioned above is an example of web 3.0, also known as the
The semantic web is a web in which all data has meta-data, which is machine readable. Thus, a blog post will have a machine readable format, which will allow it to be read and interpreted by machines.
Ullam iusto ut omnis a vitae. Odio animi ex officia consequatur dolores. Minus nisi eaque aliquam rerum corporis consectetur molestias. Sed vitae dolorem provident recusandae suscipit molestiae.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...