Bloomberg reported yesterday that the Derwent Absolute Return Fund was seeded with an initial 25 million pounds ($39 million) and will begin trading in February.

Its model? Following posts on Twitter, and tracking emotionally significant words to anticipate the market’s next jag up or down. A recent study concluded that short term market moves can be anticipated this way with a high degree of confidence.

I am all in favor of innovative new ways to slice data, and this one looks fascinating. However, before I jump into this with both feet, I do have some reservations. The sample set of a few short months of Twitter vs the Dow is way small — the research measured the public mood by searching Twitter posts from February to December 2008 for synonyms of and language related to six moods: calm, alert, sure, vital, kind and happy.

That period in 2008 incorporating the collapse of Bear Stearns and the September market collapse was not exactly typical. Indeed, the monkeys were unusually agitated during this time. This model will need to prove itself during periods of normal volatility and sentiment.

Still, it would not surprise me if the Twit stream reveals some consensus about the mood of the monkeys. At extremes, that could be an assist in developing a trading thesis about intermediate term highs and lows.

My working assumption is the primate factor — human emotions — is immutable, but the other factor is the changing ways the technology is applied, deployed, etc. may have an impact on the quantified sentiment reads. Already, outfits like StockTwits have a dedicated user base of active traders, and their perspectives may not parallel the public’s mood.

There have been a variety of start up attempts to capitalize on this, but so far, none seem to have found the magic formula. Going straight to Twitter bypasses all of that.

Sentiment is an important part of my model, but only at extremes. I wonder if it can be used to manage day-to-day trades. I am skeptical, but willing to consider this if it can show some trade efficacy.

>

>

Sources:
Twitter mood predicts the stock market
Johan Bollen, Huina Mao, Xiao-Jun Zeng
http://arxiv.org/PS_cache/arxiv/pdf/1010/1010.3003v1.pdf

Hedge Fund Will Track Twitter to Predict Stock Moves
Jack Jordan
Bloomberg, December 22, 2010
http://noir.bloomberg.com/apps/news?pid=newsarchive&sid=a0HeKpjwajW8

Computers That Trade on the News
GRAHAM BOWLEY
NYT, December 22, 2010
http://www.nytimes.com/2010/12/23/business/23trading.html

Category: Psychology, Quantitative, Technology

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

14 Responses to “Scraping Twitter to Find the Mood of the Monkeys”

  1. Amod says:

    AAII Bulls at 63 % and only 16% Bears….Time to Sell!!! What say Ye?

  2. low_frequency_trader says:

    Do the HFT algos have Twitter accounts? They seem to be the only ones pushing the market around these days.

  3. b_thunder says:

    where’s cause and where’s effect? i didn’t see the answer in the above post. is twittering public becomes more “bullish” and causes market to marginally move up? or is the moving higher market cause the twittering public to become more “bullish?” or is this just another variable int he “virtuous circle?”

    on the other hand if i could see the twits (or SMS or tap the phone lines) of Helicopter Ben, Lloyd, Jamie and Stevie – that would be very useful. i wouldn’t have to work another day of my life :)

  4. rktbrkr says:

    BB has done the easy part with QEX – now comes the tricky part dismounting from a galloping horse.

    I bet CNBC viewership is up too, another deadly warning.

  5. rktbrkr says:

    BR, weren’t you monitoring activity here to monitor interest in the market in general?

  6. constantnormal says:

    Amod — you don’t need the AAII survey, the very existence of this fund is an instance of the 2-minute warning bell, similar to any web company imaginable getting a huge IPO in 1999, regardless of their business model, or even whether they had sold their first product!

    The Fed has succeeded in blowing a stock market bubble out of the ruins of an economy, and things like this lead me to believe that it is just about over.

    Whether it is time to sell or not depends entirely on the individual, and their ability to tread water while waiting for the bubble to pop. It could be a long while, but my gut (and history) tells me that once the silly stuff starts appearing, valuations begin to accelerate at an exponential rate, with the (unpredictable) chaotic collapse arriving within a year.

    I imagine that there is a department within the bowels of Goldman Sachs (or one or several of the other vampire squids) that is actively working on tying automated tweet-generators into their trading bots in order to steer the markets.

    Have we seen any instances of rumors being sparked by a tweet without foundation? Rumors that moved a stock or market segment? If not, the time is ripe for some enterprising fool to flirt with a spell in an orange jumpsuit (vampire squids need not fear this, only the small fry are treated in this manner).

    What fraction, I wonder, of the stock traders (including automated trading bots) — the beings that make decisions regarding stock purchases/sales — use twitter? And what fraction of the twitter users trade stocks?

    I may be way out of my depth here, but it strikes me that the bulk of the tweeters are younger generation kids looking for a way to connect with their peers, or to slavishly follow after their idols. At least I hope that is the case, as 140 bytes of text does not strike me as a respectable building block for a body of information.

    As to the study showing a high correlation between stock tweets and the “mood of the market”, it strikes me that a service begun in 2006 has not been through even a single complete market cycle, so it seems unlikely that they have sufficient data to make any sort of assessment about the “mood of the markets”. Here is another study regarding twitter, snipped from the entry in Wikipedia on Twitter:

    San Antonio-based market research firm Pear Analytics analyzed 2,000 tweets (originating from the US and in English) over a 2-week period in August 2009 from 11:00 AM to 5:00 PM (CST) and separated them into six categories:[46]
    Pointless babble — 40%
    Conversational — 38%
    Pass-along value — 9%
    Self-promotion — 6%
    Spam — 4%
    News — 4%[46]

    ===

    I’ll grant you that one can probably pick up a hint of the twitter crowd-space emotional tenor by sampling tweets, but to relate that to the emotional behavior of the markets and the players within, is a bit of a stretch for me.

    The study that BR reports on was in place between Feb 28 and Dec 19 of 2008. No other periods were sampled.

    Hardly a “rigorous” study, being over a fraction of a market cycle, and no others have repeated its findings. And yet here we find a hedge fund being started with $39M based on this stuff. Hence my hearing the 2-minute warning bell.

  7. Julia Chestnut says:

    Anyone who has ever seen a huge flock of birds wheel through the sky and turn as a group, not entirely together but eventually all with the same mind, would consider this to be a potentially interesting experiment. The nature of twitter – the regularity, length, and mundane nature of the posts – would make it tend towards unfiltered herd movement.

    But beware of the unspoken underlying assumption: what effect do normal people – or even big market participants – have on the market at large? If you believe that the market is being moved fundamentally no longer by small participants, but by large players, you need a model for how those players filter their information and make moves. Assuming they also try to leverage the herd behind those moves with propaganda. . . .

    What I would expect it to be able to very quickly gauge is commercial trends. Let’s say that you can figure out before a true consensus forms what the hot new toy is going to be this Christmas: there might be something to be done there. From a marketing perspective, I can see obvious potential. But the stock market? I’m skeptical. I feel that is totally divorced at this point from raising capital for legitimate businesses. It’s almost become its own twitter feed.

  8. louiswi says:

    A couple of thoughts here:

    Sampling tweets seems akin to sampling a couple million goose droppings in the hopes of finding the tootsie roll that fell out of your pocket on your backswing.

    Or,
    “I’ve never seen a lemming but if I ever do, I’m sure I’ll see a lot more than one.”

  9. Deflator Mouse says:

    I wonder what Hari Seldon would think of this….

  10. SINGER says:

    What real trader would post their Assessment on Twitter? Does anyone even trade stocks anymore?

  11. [...] Barry on hedge funds using Twitter sentiment.  The term "Monkeys" is employed herein.  (TBP) [...]

  12. [...] You knew this was coming.  A hedge fund using Twitter-derived signals to trade.  (Felix Salmon,  NYTimes, Big Picture) [...]

  13. Liminal Hack says:

    “where’s cause and where’s effect?”

    Isn’t that the question of the age? And of course totally unanswerable.

    A reflexive society is one in which all the output is fed back immediately to the input, only possible in an age of near instant communication and suddenly non computable network effects, and in which people do whatever they want, following their lacanian drives. Impossible to focus on the long term in such an environment. Financial deepening is all part of this macro trend.

    So what matters in this context is narratives, and particularly the narratives dreamed up by the nascent hive mind which exposes a few neurons on twitter.

    Mining twitter for trading ideas seems to me a better strategy than analysing aggregate demand, Fibonacci sequences, charts or the news flow.

  14. tweettrader says:

    Research Paper “Tweets and Trades – the information content of stock microblogs”
    Find the full paper here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1702854

    Check out the new stock microblogging forum TweetTrader.net!

    Abstract
    Microblogging forums have become a vibrant online platform to exchange trading ideas and other stock-related information. Using methods from computational linguistics, we analyze roughly 250,000 stock-related microblogging messages, so-called tweets, on a daily basis. We find the sentiment (i.e., bullishness) of tweets to be associated with abnormal stock returns and message volume to predict next-day trading volume. In addition, we analyze the mechanism leading to efficient aggregation of information in microblogging forums. Our results demonstrate that users providing above average investment advice are retweeted (i.e., quoted) more often and have more followers, which amplifies their share of voice in microblogging forums.

    What’s the difference to the work cited above?
    A few recent studies suggest that the information content of microblogs may help predict macroeconomic market indicators. O’Connor et al. (2010) have found Twitter messages to be a leading indicator for the Index of Consumer Sentiment (ICS), a measure of US consumer confidence. Both Zhang, Fuehres, and Gloor (2010) and Bollen, Mao, and Zeng (2010) find that a random subsample of messages from Twitter’s public timeline can be used to predict market indices such as the Dow Jones Industrial Average (DJIA) or the S&P 500. However, all of these studies are concerned with broadly defined data sets (e.g., all available messages or blog posts in the sample period, most without a specific reference to the stock market) and derive aggregate sentiment measures. While the correlation of these aggregate measures with macroeconomic indicators is encouraging, it does not allow us to draw conclusions about the information content of stock microblogs with respect to individual stocks. Das and Chen (2007) found the relationship between aggregated sentiment and index returns to be much stronger than the correlation for individual stocks. Therefore, our study focuses on the specific domain of stock microblogs and investigates their relationship with market prices of publicly traded companies.