Wednesday, May 01, 2013

Did We Contribute to the #HackCrash? Light-speed Information and the Psychology of the Herd

"Rothschild coaches careered down highways; Rothschild boats set sail across the Channel; Rothschild messengers were swift shadows along the streets. They carried cash, securities, letters and news. Above all, news---latest, exclusive news to be vigorously processed at stock market and commodity bourse."
~ Frederic Morton (1962), "The Rothschilds: Portrait of a Dynasty"

There are many stories and legends surrounding the Rothschild family, and probably most are untrue.  It is widely acknowledged that Nathan Rothschild was the first Londoner to receive the news that Wellington had defeated Napoleon at the Battle of Waterloo in 1815.  Rothschild received this information through his network of fast carrier pigeons, which one of his agents released from Waterloo when the battle was likely to be won. 

News of the outcome of the Battle of Waterloo was breathlessly awaited in the British financial markets.  According to legend, when Nathan Rothschild entered into the Exchange the day of the victory, he was already known as a master of speedy information. As such he took up his usual trading position and began quietly selling British Consols (bonds) into the market.  Traders quickly identified his selling, spurring a panic.  It was assumed by traders that Rothschild had heard Wellington was defeated.  Then, legend has it, in the midst of the panic Rothschild began stealthily buying Consols.  He purchased a sizable position before traders realized they had been fooled and drove prices higher.  Rothschild increased his profit from the victory dramatically by engineering a short-term panic and then buying into it.

This legend serves to illustrate that there are two primary ways to profit as an investor:  1) Get important information fastest, and 2) Know how the herd will react.   If you know what the investing herd believes, fears, and expects, then you know how it is likely to react to new information.

Back to point 1).  Getting relevant information fastest has always been the most straightforward key to successful investing.  And the hunt for the fastest information has spawned the term “Algo Wars” to describe the arms race among HFTs for faster servers, co-location, re-engineered routers, and other accoutrements of light-speed trading.  The time advantage HFTs use is incredibly narrow.  For example, the May 1, 2013 WSJ profiles how HFTs use a 1-10 millisecond advantage to rapidly detect order-flow and front-run larger orders.

At MarketPsych Data our clients include the largest hedge funds in the world, and we can’t say exactly how our clients are trading off our sentiment feed.  The TRMI are published minutely as a 1440 minute (24-hour) rolling average.  And several of our indices do show evidence of the #HackCrash event as you can see below.

To be clear about the title question, I doubt the minutely Thomson Reuters MarketPsych Indices (TRMI) data feed contributed to Tuesday April 23rd’s HackCrash in which the S&P 500 plunged 1% and then recovered all within the span of 4 minutes.  But I can’t be sure.  The TRMI both show fast evidence of the event and also give an excellent read on the beliefs and expectations of the investing herd.

Today’s newsletter looks at the role of information speed and herd impact.  We’ll first take a look at a famous failure related to United Airlines, then we’ll visit how our systems dealt with the hacked Associated Press tweet and its aftermath, and finally we review a few investment ideas for the week.

Pity the HFT

For those who criticize High Frequency Traders (HFTs) and their use of social and news media feeds for risk management, I’m going to step up in their defense about this topic.

Consider that the vast majority of market makers are high frequency traders.  As a market maker, every day you are buying and selling in the market for clients.  That’s your job.  But your job is very risky.  The single biggest risk to you as a market maker is being caught flat-footed – not being prepared for a dramatic event that your clients see before you and trade against you.  So you exercise risk management.  In this case, like Rothschild, you need to see the defeat or victory first.

HFTs are typically engineers and computer scientists, and many of them understand social media and text analytics well enough to construct simple systems that detect catastrophes reported in the news.  When an Associated Press tweet contains the words:  “Obama, bomb, White House, injured” – that’s an easy catch for a text analytics algorithm.

And what would a market maker do in such a case?  They withdraw their bids.  They won’t go long if a potential catastrophe is in play.  Then as selling enters the market, it causes a steep dive.  When it is clear the tweet is a “hoax” or a “false alarm” or no longer corroborated by other sources, their algorithms begin to ignore the information – “all clear” they signal.  And bids re-enter, stabilizing prices. 

This may not be a pleasant reality of the market-making system, but it is necessary for market-makers in the age of Twitter to be as fast as Twitter.  Otherwise they would not survive.
Do some high frequency traders also pursue Rothschild’s second key to profit, manipulating the herd?  Absolutely.  Look at Business Insider’s indictment of HFT’s manipulating Natural Gas prices in the low-volume seconds before inventory releases, called “Banging the Beehive.”  Anxious traders are led to believe the news has been released (or someone knows more than them), and panic ensues in one direction.  Limit orders and stops are filled, then prices reverse dramatically in the opposite direction.

Our data satisfies the need for both of Rothschild’s insights: 1) minutely access to low-latency Thomson Reuters news and social media feeds and 2) a quantitative measure of the fears, expectations, and consensus of the market herd.  The following example illustrates how the intersection of quick reactions to bad news (1) and herd fear (2) can drive enormous price moves. 

When Bloomberg Reported United Airlines’ Bankruptcy 6 Years Late

On September 8, 2008 a South Florida Sun-Sentinel United Airlines (UAUA) bankruptcy article from 2002 was listed onto the live Google News website as if it were current.  From Google News a hurried editor at Bloomberg republished the article onto the Bloomberg news feed at 10:53 a.m.  The below chart depicts the price action in UAUA (United Airlines) stock at the news of its repeated bankruptcy.

High frequency traders performing text analytics on the Bloomberg news wire decimated the price of UAL in minutes.  According to MarketWatch:  “Trading in UAL (US:UAUA) stock was halted at about 11:06 a.m. Eastern time after falling nearly 76% to $3 a share. Shares resumed trading at about 12:25 p.m. They closed with an 11.2% loss at $10.92.”   A dramatic daily loss occurred for UAUA despite the released information being 6 years old

News analytics has been a robust business since the mid-2000s.  Firms including Thomson Reuters, Bloomberg, and Dow Jones offer low-latency news feeds.  Thomson Reuters and Dow Jones (indirectly) offer pre-analyzed feeds with low-latency sentiment tagging.   In the past two years social media has been added into the low-latency news mix, including our own product with Thomson Reuters.

The #HackCrash

The #HackCrash is remarkable because it occurred due to social media.  Despite the prevalence of trading on social media (including our own fund, which was the first), social media communications have never before impacted markets so significantly.

This image from MarketWatch captures the drama of the tweet and the price impact on the S&P 500:

As you can see, the fall was fast and the recovery even more dramatic.

So what, if any, was our role in the event?

Tremor in the TMRI
We saw movement in several of our indices after the incriminating AP Tweet was released.  To be clear, we and most other text analytics firms have screens to detect when information being conveyed is explicitly false.  For example, when information about an event is qualified as a “hoax,” “fake,” or “false alarm” our systems ignore the indicated content of the message (it is not scored).  So when we looked at our minutely  Thomson Reuters MarketPsych Indices for the S&P 500 and the United States using combined social and news media content, we see interesting evidence of the event, but only a tiny – almost imperceptible - blip indicating that there was an attack, as it was quickly discredited.   

Below are visuals showing both the long-term and the short term impacts of the event on our relevant sentiment feeds.  In these charts, keep in mind that you are seeing a 24-hour average (1440 minute rolling average).  On a minutely level the action is much more dramatic.  Nonetheless, it can be seen here as well.

In the big picture, we note that the Thomson Reuters MarketPsych Optimism Index for the S&P 500 fell steeply following the fake AP tweet and the market price tremor.   The following chart depicts a 48-hour period in GMT/UTC time from April 23rd through April 24th.  (GMT time is currently 4 hours ahead of the U.S. East Coast time zone).  The decreased Optimism was fairly short-lived, lasting for about 24-hours. 

Over the course of minutes we saw rapid dissemination of the event reflected in our S&P 500 PriceUp index (PriceUp is a net of references to prices rising or falling).  Conversations about the price first falling and then rising were rapidly disseminated through social and news media.

As a result of the language used to describe the event, we saw the Thomson Reuters MarketPsych MarketRisk index surge.  The MarketRisk index measures emotive and speculative conversations that tend to predict riskier markets to come.  Though it surged for a few hours, it declined rapidly thereafter as concerns faded.


The above graphics display our findings around the S&P 500 and its constituents.  In our United States data feed, we noticed a rise in Anger at the United States government.  The report of Syrian hackers posting the tweet due to anger at the U.S. Government’s support for the Free Syrian Army was accurately scored.  Because the news of the alleged perpetrators and their motivation arrived late, it does not appear as a dramatic jump but rather a slow rise in the index over 24 hours.

The event was an odd event – it wasn’t an ACTUAL attack.  For example, it should not have appeared in our United States Violence index.  And it didn’t - it registered only as a tiny blip due to it being quickly identified as a fake.  That said, the price reaction did impact investor perceptions, for example the event increased pessimism significantly over 24 hours.
How was this event trade-able?  I’d say that in the 1440 minute average, it likely was not.  That said, on a raw minutely level the action was very interesting. 

Our underlying sentiment feed is perfect for market-maker risk management uses.  If you’re a market maker, let me know if you’d like more information about how to use our data to quickly identify and avoid risks.

Trading Recap

None of our ideas this week are related to the #HackCrash.  We’re currently seeing a one-week sell on Trina Solar (TSL).  We also see a one-week sell on Herbalife (HLF) – a surge of excitement in that stock likely to peak this week.  SEE DISCLAIMER at end of email.  These ideas were developed using backtesting and a hedge against the S&P500, but you invest at your own risk.

Last week’s Buys on GLD, GDX, SLW, and ABX went well, with all being significantly positive on the week (I’ll quantify these later).  There were no shorts last week.  I’ll give exact stats on recommendations in another newsletter.

We are adding a monthly individual U.S. stock sentiment-based investment model and a Russell 1000 sentiment ranking system to  Users will be able to follow Ben Graham’s advice to buy from pessimists and sell to optimists (easier said than done, of course).  Please respond to this email for a free trial.

Housekeeping and Closing
Keep in mind Rothschild’s stratagem as you go forward.  If you don’t have the fastest information in the market, then you’ve got to know the herd.  What are they thinking and feeling, and how are they likely to react to the next news release or major event? 

Please contact us if you’d like to see into the mind of the market using our Thomson Reuters MarketPsych Indices to monitor market psychology for 30 currencies, 50 commodities, 120 countries, and 40 equity sectors and industries in social and news media.  This data is used by top global hedge funds to improve their investment returns.  Please let us know if you’d like to learn more.

We love to chat with our readers about their experience with psychology in the markets and with behavioral economics!  Please also send us feedback on what you’d like to hear more about in this area.

We have upcoming 2013 speaking engagements in New York, Orlando, London, Chicago, and San Francisco – we look forward to seeing our friends in those cities!  Please contact Derek Sweeney at the Sweeney Agency to book us: [email protected], +1-866-727-7555.

Happy Investing!
Richard L. Peterson, M.D. and The MarketPsych Team

This material is not intended as and does not constitute an offer to sell any securities or a solicitation of any offer to purchase any securities.
The information of the MarketPsych Report is presented free of charge.  It is no substitute for the services of a professional investment advisor.  Investments recommended may not be appropriate for all investors.  Recommendations are made without consideration of your financial sophistication, financial situation, investing time horizon, or risk tolerance.  Readers are urged to consult with their own independent financial advisers with respect to any investment.
Past performance is no guarantee of future results.  Screen and model signals and related analysis are for informational and entertainment purposes only and should not be construed as an offer to sell or the solicitation of an offer to buy securities.  Most financial instruments (stocks, bonds, funds) carry risk to principal and are not insured by the government.  Anyone using this newsletter for investment purposes does so at his or her own risk.
Data accuracy cannot be guaranteed.  Opinions and analyses included herein are based on sources believed to be reliable and written in good faith, but no representation or warranty, expressed or implied, is made as to their accuracy, completeness, timeliness, or correctness. We are not liable for any errors or inaccuracies, regardless of cause, or for the lack of timeliness of, or for any delay or interruptions in, the transmission thereof to the users.
As a matter of policy, we DO NOT act upon the investment information that this newsletter provides prior to making it available to the public.  We currently DO NOT have an actively trading investment fund.  However, it is possible that investment recommendations will be made concerning investments already present in our personal portfolios.  These are not intentional.  We do not actively discuss or post investment recommendations on social media in such as way as would influence our trading and investment signals.  We do not accept compensation of any kind from any companies mentioned herein. 
MarketPsych is not responsible for any special, indirect, incidental, or consequential damages that may result from the use of, or the inability to use, the Information contained on this newsletter whether the Information is provided or otherwise supplied by MarketPsych or anyone else. Notwithstanding the foregoing, in no event shall MarketPsych total liability to you for any and all claims, damages, losses, and causes of action (whether in contract or tort or otherwise) exceed the amount paid by you, if any, for accessing this newsletter.
MarketPsych expressly disclaims all warranties and conditions with regard to the Web sites, their Content, and the Information, including, without limitation, all implied warranties and conditions of merchantability, fitness for a particular purpose, title, and non-infringement. By using the Web site, Content, and Information, I assume all of the risks associated with their use, and I release and agree to indemnify and hold harmless MarketPsych from any and all liability, claims for damages, and losses arising from or connected with such risks.

No comments: