Hedge Funds and Twitter Sourcing

The information concerning hedge funds use of twitter data is not as interesting as I thought it might be. There is one hedge fund in London who uses information generated by Johan Bollen. The company is Derwent Capitol they have not been trading using this information for very long, not quite a year so far. All looks well on the performance end thus far but let’s not get too exuberant until we have seen some substantial and consistent returns.

In the first few moths they showed a return greater than the S&P and the average return of other hedge funds so it seems they are off to a solid start.

I have to say, while I am happy that the hedge fund seems to be doing well it is really the researcher I am excited for. Johan was initially unable to get his paper published and so decided to simply put his findings online. The reaction his paper has gotten is amazing, I have seen him interviewed on many financial networks and I understand he has started his own company with his fellow researcher and is making money from his partnership with Derwent.

 

Advertisements

4/13

It’s always good to check in on data collected on a friday the 13th. April 13th, like many days in the beginning of April was not a good one for the Dow Jones Industrial Average (DJIA) it lost 136.99 points going from 12986.58 to 12849.59.

Once again I am left wondering if I am looking at the data correctly. I have tracked the top ten lists of “hot” and “not” stocks for the day and after a comparison of the lists and how they changed and how the stocks changed for that day I am left wondering if anyone is crazy enough to let their money ride based on tweet trader.

Every time I look at the data I wonder if I should be taking it from a different angle but I’m not sure what that might be. I start by copying down the lists from the website on one piece of paper next to the hour of the day.

Here is what that looks like

I then look at the lists and try to think of what is the most useful. I noticed that FAST reported earnings on the 12th so I thought that might be triggering that stock to be on the “not” list. But when I looked up the status of FAST on the 13th it turned out it was one of the few stocks that ended the day higher and it did so due to a good earnings report. Go figure.

Most of the stocks I looked at dropped on the 13th even the ones that stayed on the “hot” list all day and in a seemingly strong position.

Due to this utter lack of any telling results I decided to change my strategy a bit and look at stocks that suddenly appeared or disappeared. from the lists. I figure a stock that suddenly appears in a strong position on the hot list must behaving a good day on the market.

PHM was just such a stock, it suddenly appeared on the hot list at 3:15pm (technically after market close).

Another Article

The main article that launched this idea that twitter sentiment could make clear a kind of collective mind that could predict the moves of the stock market was found in this article.

If the article itself is more math than you bargained for then you might find the abstract is all you need to extract the conclusions from the paper. The paper cites over 87% percent accuracy when predicting the closing changes of the Dow when using specific mood indicators.

This article started a string of papers and research concerning public mood and prediction of everything from box office success to elections.

The article also led to at least one hedge fund, located in England, deciding¬† to use twitter information in some of it’s investing decision making. The hedge fund company is called Derwent Capital Markets, I will address their outcomes in a later post.

“I Hope it is Not as Bad as I Fear”

I took the quote above from an article titled Predicting Stock Market Indicators Through Twitter ‘I hope it is not as bad as I fear‘. I thought the title was clever considering it discussed the emotional content of tweets and how they predict the market. The article was written by Xue Zhang, Hauke Fuehres, and Peter A. Gloor.

What interested me about the article was that the approach was very broad. When I first heard about the link between twitter and the stock market I thought it was a link was a specific link, which might look like this: Lots of people tweeting positively about Apple = Apple’s stock price goes up. This is not unlike how it seems to be processed on the Tweet Trader website. They mechanically comb through tweets with dollar signs and ticker symbols like this $AAPL and they determine if the sentiment is positive or negative about the stock and then if enough buzz is generated about a stock in either direction it makes it to one of the lists. This approach is certainly interesting and perhaps more helpful to a trader but I like the broader approach that Zhang et al have taken too.

The broad link between emotion and the stock market is clear. People naturally tend to want to buy high and sell low which is the exact opposite of what they should be doing, and the reason they do this is emotion. It takes a strong constitution to watch your retirement account get cut in half during an economic crisis and many people sell what they have thinking they need to protect what is left. But more often than not the market rebounds and those who pulled their money out lose their chance to fully recover their losses. But these emotions are not simply present on the days where there are big crashes and rebounds. If you watch the stock market on a day to day basis you will begin to see how fickle the market is. Mr. Moody is it’s nick name for a reason.Every day the temperaments and attitudes of thousands of people determine the course of Mr. Moody.

In the paper by Zhang et. al. they analyzed twitter to get a general emotional read of the ‘masses’. They looked for key words they felt indicated positive and negative emotion such as ‘hope’ or ‘worry’ they would then keep track of the numbers of tweets with positive and negative tweets and they would be able to asses the emotional status of the masses. They found that this emotional status was a correct predictor of the Dow 24 hours in advance. What I find interesting about this is that that emotional data being collected was not stock related. They say that “…just checking on twitter for emotional outbursts of any kind gives a predictor of how the stock market will be doing the next day.” In some ways I can see ho this might be more useful as general information about broad stock market moves and seemingly more accurate than information about individual stocks.

An interesting side note; “More interestingly, we also find that the number of positive tweets is much higher than that of negative ones, more than double on average, which might suggest that people prefer optimistic to pessimistic words.”

April 10th

April tenth was an interesting day on the stock market. The Dow lost 213 points and the S&P lost 23.61. If you look at a graph of the Dow over the year to date you will see that April 10th marks the bottom of the biggest losing streak since 2011. On the tenth the Dow dropped below the 12,800 mark.

When the Dow drops over 200 points in a day, as it did on the tenth, there are usually very few, if any stocks that don’t follow suit. There are a number of reasons why this happens, sometimes it is bad news or economic data, but sometimes it is for other reasons that aren’t always evident. It is this not knowing that makes the stock market tricky for those with a short time horizon, it is also why the market is often known as Mr. Moody. And He was certainly crabby on 4/10

This day allowed for a unique test of the Tweet Traders “hot” list. Only two stocks remained on the list all day and those two stocks were the only two on the hot list that ended the day in the positive. Those two stocks were Safeway (SWY) and Supervalu (SVU). Too bad you can’t know in the morning what stocks will stay on the list all day. Supervalu, however, would have been a good guess because they announced their earnings the 10th and they beat analysts expectations. So perhaps tweet trader isn’t that smart after all. The plot thickens.

An Irritation

There is a small detail that keeps tripping me up in this project and because it is now down to the wire and I am about to freak out I thought I would write about it instead. Hopefully this will give me a chance to take a few breaths and put things back in perspective.

I am trying to look at a good chunk of data for anything note worthy and realizing that it is not easy. The variables are numerous and the despite trying to freeze time via snap shots I keep finding that everything is kind of fuzzy. Tweet trader is kinda sorta right about the stock market some of the time if you stand back and squint. At least this is what I am finding so far. I will blog about this in more depth later.

This, in and of itself is not that disturbing, nor all that shocking. In the process of looking at this data however there is a small issue that is both small enough to overlook accidentally but also damaging to the analysis if you do overlook it.

The time change between the east coast and our lovely central time is killing me. As my computer takes pictures of the websites I want, at the times I want, it registers them as central time because, well, that’s where we are. As I track the movement of individual stocks on google finance, however, the time I see is eastern time because wall street is on the east coast. This seems like a little thing but as you track stocks hour by hour you of course need them to be lined up correctly. Let’s just say that occasionally I forget, and even worse, I sometimes forget that I forgot.

Damn you time zones!