each trading
day for U.S. Stocks. The blue line snaking across the bottom of the chart shows the
number of trade executions. Note the scale is in billions. During the end of the Internet bull market in 2000 (not
shown), the number of quotes in a day was about 5 million, less than the height of 1 pixel
on our chart.
After Reg NMS, a form of High Frequency Trading (HFT) appeared that took advantage of
time latency arbitrage caused by differential system loads (an imbalance of quotes on
one network versus another). It wasn't long before one of these HFT figured out how
to induce latency arbitrage and the divergence between quote and trade traffic began.
There are three peaks labeled A, B, and
C that brought out the worst in HFT:
|
Date |
Event |
Quotes
millions |
Trades
millions |
|
2000 |
Internet Bull Market |
5 |
3 |
A |
10-Oct-2008 |
Financial Crisis |
1,018 |
83 |
B |
06-May-2010 |
Flash Crash |
1,112 |
65 |
C |
08-Aug-2011 |
U.S. Downgrade |
2,297 |
74 |
After the system-overloading quote volumes during August 2011, Nanex,
Themis Trading, and others who
were concerned about the long term health of our markets, became vocal about the enormous costs being heaped on those who do not benefit from HFT, yet
have to pay for the infrastructure they require. In fact, during peak times, quote
spreads are often wide and unstable.
We are paying for the costs, but reaping none
of the benefits.
Whenever a new peak in message traffic occurs, those who process market data must upgrade
systems to be able to handle the new capacity plus additional overhead for higher traffic
on high volume days. In other words, peak message rates set new, permanent capacity
requirements. You can't just flip a switch and get additional capacity when you need
it, then turn it off to reduce costs when you don't.
|
|