sobota 4. srpna 2012

We are afraid, we hope, but we do not believe.


There is lot of said about the recent turmoil in Spanish Government Bond yields. After speech of nearly any European politician, Spanish Bonds are going up or down by quite a large number. For example, after the Spanish prime minister's promise to ask EU/EZ resources to buy their bonds if needed, the bonds went down to 6.84 from being nearly around 7.20, change by around 5%. But how significant  are these movements? How they tell about the price of financing the Spanish debt?

The particular movements are at some degree of very low informative power. When we look at the recent evolution of 10y Spanish bonds, we see very high volatility. In half a month, the bond can easily change the value by around a percent, depending on a particular swing in a mood, and, it seems that volatility is even growing. It means that today value does not say too much for the next bond auction. In addition, the moving averages suggest an increasing trend. I think that these two facts are really worrying. The increased levels of volatility further suggest low market's ability to predict a direction in the solution of the recent debt problems. The high level of yield is a reflection of uncertainty of the Spanish public finance (and, from a big part, also the uncertainty about the direction of the EU/Eurozone), while the growing volatility in the yields further says that even the uncertainty is uncertain.

The recent yield movements say: "We are afraid, sometimes we hope, but most of the time, we do not believe." ...or maybe the explanation is much simpler -- some big players are just speculating on volatility.

Anyway, I would recommend a following rule how to read the uncertain uncertainity: politicians should use the upper bound in yields to be motivated for a solution to the debt crisis, while people should, on the other hand, look on the lower bound to not be worried so much and hope. In reality, the previous rule is rather reversed. The politicians are enjoying a false sense of safety, while many people are losing their patience.


Spanish 10Y Government Bond Yield Index. Source: Bloomberg.

čtvrtek 2. srpna 2012

Iron(ic) Knights


Recently, the traders and commentators revealed the background information behind the error, see here and further references therein. The algorithm should execute a large block trade worth of $5bn over a 5 days period without disturbing the markets. However, the algorithm executed it immediately. Conclusion: The recent event shown us why it is important to spread block trades over a long period and do not execute them instantly.

The recent "accident" was not the first one, see this link. The error on KCG's algos disrupted trading of 30 ETFs on March 31, 2011, when these vehicles were introduced into the market. The accident forced NYSE to cease up the trading.

Finally, the market seems to punish the Knights for their error since rumours says that they are considering the bankruptcy. This 30-minutes event produced a loss exceeding the quarterly profits and they have tough times. The picture below shows the evolution of their shares. And why "Ironic" in the title? The morning increased volumes in trading with their shares evokes the previous days' trading pattern of their victims, it just lasts a bit longer... till the Knight's end, perhaps.

KCG US equity after the error.






Algorithmic Knights


In the morning, August 1st, the operations at the NYSE were disrupted by the reported bug in the algorithms of the Knight Capital Group (KCG). KCG is a big market maker, over whom flows tens of billions worth equities over a single trading day. See more information about this event and links about the KCG operations here.

To visualize what this bug means, let us look on the first picture of the three day history of market operations of the Citigroup equities as taken from Bloomberg. The lower bar with volumes clearly say a significantly increased market volume, which was very likely caused by the algorithmic bug of the market maker. The similar pattern can be seen over many other equities. Luckily, after half an hour, the bug was removed and market started to resume the normal operations and repair all the demages.


Trading with Citibank equities during the critical morning.

However, what is more striking is the speed of the market response. It is true that NYSE ceased up trading with some titles but as we see on the particular example with Citibank, the unusual market operations last for at least 30 minutes. Let us look on the trading with the KCG themselves.



Trading with KCG equities before and after the reported bug.

The shares of KCG started to fall down during the morning. This can be considered as a market evidence that it is them who are guilty for that mess and that they are not as sound as they claim to be. This can be considered as a fair response. They got punished for their fault. Their reputation was damaged as much as their shares dropped -- see the ride from 10.50 to 6.90. But let us look closer, when the fall begun.

Trading with KCG equities -- closer look.
The closer look at the trading pattern with KCG reveals a beginning of the plunge in the shares as significant break starting around 12 minutes after the opening of the markets, or, in other words, after 12 minutes of the disrupted operations. It took another 18 minutes to resume the standard trading. 12 minutes sound as a short time, but when this period is compared to the time scale of the algorithmic operations--milliseconds--this sounds like eternity. The algorithms were crazy and the rest of the market just was not able to keep pace with them.12 minutes, in this case, suggests the reaction time of the market to something nasty.

The other possibility is that human traders realized that problems and were just giving more time to Knights to repair, so the other option how to explain the 12 minutes is just a time scale of the human traders' patience. After 12 minutes, the crowd started to be angry and fire sale begun. They finally punished the market maker for their misbehavior. Only the careful analysis could decide which of two is the correct answer, or whether there is an another answer. I do not aim to answer it here.

The conclusion of this, however, is that algorithmic market makers operate on much faster time scale than humans. The human traders are not able to correct errors of the algorithms immediately and therefore the self-regulation of the markets is significantly flawed due to the technical reasons. It can be only other algorithms and technical rules, which can do this task; however, there is no convincing evidence that it is so. The Flash Crash and this event rather suggest that algorithms bring, along with their advocated advantages, also a component of the Systemic Risk since they are not endangering just their money but also money and wealth of others. The risk is obviously bigger for algorithmic market makers with an oligopolistic-like market share, who can easily hide in their codes a function title BlackSwan().

PS: The important conclusion from this event is that market data from that day are useless for any research, except the case study and I will have to remember this day for couple of next years. The only exemption in the data is the KCG equity, whose drop is an example of nice old-fashioned fire sale.