Flash Crash report part 4
To give some of you skeptics some assurance concerning the integrity of the electronified markets I've excerpted this item:
A number of other hypotheses regarding the causes and implications of these data delays have
been offered. One specific concern is that traders could take advantage of the timing delay
between data reported to the consolidated feed and data reported on the proprietary feeds by
buying securities at prices on one feed and selling securities at prices on the other. It generally
is not possible to do this, however, since the consolidated feeds do not reflect a separate trading
market from the exchanges. One cannot “buy” or “sell” at an exchange’s prices as shown on
the consolidated data feeds separately from the exchange’s prices as shown on its proprietary
data feed. All orders attempting to execute against an exchange quote in the consolidated data
feed must be routed to that exchange where they will be matched in real-time based on then available
quotes at that exchange. These real-time exchange matching system prices may be
different from the quotes in the consolidated data feeds if, as on May 6, the exchange is
experiencing latencies in transmitting its data to the consolidated data processors. The
exchange’s prices in the consolidated data feeds are quite literally inaccurate – they do not in
fact reflect prices that are currently available to anyone at the exchange.
This is quite a profound statement and it shows the reason why one has to approach measuring the market from a quantum mechanics viewpoint. Some years ago on this tread I went into this in some detail, but I stopped when I realized nobody who reads here would understand it.
The basic idea has to do with impossibility of simultaneity in observation which forces upon humans who are hopelessly bound to some form of classical interpretation of phenomena to arbitrarily impose a continuous picture on discrete phenomena.
We have little boxes in our representations filled with some numbers called quantum states. Then we have some other little boxes filled with more quantum numbers. We can't impose the classical view by providing a continuous transformation of one set of numbers in one little box into another, but we do know that one set of quantum numbers in one little box get to the same kind of set in another little box. We can only give correlations between two pairs of temporally evolving little boxes. This keeps our probability hypos consistent.
So, whereas, I may think I can manipulate the stream of stock atomic actions with sufficient applied force, the actual implementation of the effort creates a divergent state that works against my intent. In the stock market the divergent state is seen as latency. The more money I use to manipulate price the slower is the response to my effort. In quantum mechanics, the attempt to measure quantum numbers in one little box is inversely proportional to the strength of the attempt. In order to get information from the states to compose a correlation, a history, I can't disturb the state. Of course, any atomic action disturbs the state. It's necessary therefore to minimally act on the stream if I wish to act on the reported state and its history. Then I must keep my involvement either static or minimal.
The Flash Crash did the opposite. It utilized the unbounded force of synthetic information and tried to apply that through the machines, thereby creating a log divergence that was unbounded both below and above. In QM, these states are called "negative norm states", and they take the form of SQRT(-k). Imaginary. Fictitious. That's why the exchanges broke all those nonsensical synthetic stub executions. |
