di
diBy Tyler Durden of ZeroHedge


In the aftermath of Knight’s crushing algo-driven error and subsequent cash loss, which may well prove terminal for the business – an artifact of a broken market we have been warning and writing about since 2009 – we present some informative insights into the various eras of Wall Street trading errors courtesy of that grizzled trading veteran, the Chairman of the Fermentation Committee, Art Cashin.

From UBS:


The End Of An Error – The conversation on both the floor and the financial media was dominated Wednesday morning by a series of apparently erroneous trades.
Between the 9:30 opening and about 10:15, a series of larger than normal orders spewed onto the floor. A later post mortem would reveal that there were approximately 150 different companies involved.
The orders were so large that in some cases they represented the average volume for 10 or even 20 days. Some may have been subdivided into a series of smaller lots.
Things almost got ugly as the sharp price movements in the stocks produced sharp movements in the options for those stocks. That set up a possible feedback loop with stocks influencing options which in turn influenced the stocks again.
That could have turned into a flash crash style chain reaction.
Luckily, the system worked, as humans intervened and several circuit breakers kicked in. Some stocks were temporarily halted and the negative chain reaction was interrupted briefly but effectively. A notable victory for the humans.
Knight Trading, through whose system most of the orders appeared to have been transmitted, was said to turn away further orders, presumably to limit client risk. The market seemed to interpret that as the problem was unresolved, and leaned on the stock.
In five decades, we have seen all manner of errors. Some are simple misunderstandings. “Buy 5 million dollars worth of IBM” becomes “Buy 5 million shares of IBM.”
Some are the results of short cuts. Several decades back, when program trading was in its infancy, some traders at Kidder Peabody tried to reduce the number of key strokes. Since we only traded in round lots of 100 shares or more, the last two zeroes were redundant. So, they taught the computer that when you hit a 1, the computer should automatically add two zeros to make it 100.
That worked fine until one lazy summer day, a trader told an intern to sell 1700 shares of Boeing (BA) The unknowing kid naturally typed in 1700 and the computer added 00 making the 1700 into 170,000 shares.
It might have been caught but the computer had been taught another shortcut. To keep their orders from being caught up in a “large order net” and handed to a human, they taught the computer to slice and dice large orders into multiple orders of 500 shares.
So, on a lazy summer afternoon, the auto-routing mechanism at the Boeing post began to shoot orders out like a machine gun. Luckily, the day was slow enough that a human caught it before more than 50,000 had been executed.
Errors on Wall Street have a long history, unfortunately.



Courtesy Tyler Durden, founder of ZeorHedge(EconMatters author archive here)


The views and opinions expressed herein are the author’s own, and do not necessarily reflect those of EconMatters.

(C) EconMatters All Rights Reserved

iHHBS9mwVxE