"What follows should strike you as crazy. If it doesn't, read it again, because it is."

Those words kick off a report by market technology firm Nanex LLC that attempts to explain just exactly what happened at Knight Capital last week, when a rouge trading algorithm prompted a market debacle that wildly distorted the price of hundreds of New York Stock Exchange issues, costing Knight $440 million to get out of the jam.

"Crazy" is one way to put it.

As Nanex theorizes, using the company's high-powered proprietary software to conduct the analysis, Knight Capital's algorithm was set on auto-pilot to "ping-pong between hitting the ask and hitting the bid. As if someone is buying at the offer and then, almost immediately, selling at the bid, then buying at the offer, then selling at the bid and so forth."

In layman's term, Knight was buying high and selling low, losing money on each trade. Nanex illustrates the point by showing a chart for trading in shares of Excelon, which show money losing trades occuring at a rate of 39 to 40 per second

Chart shows .025 second trading intervals for shares of Exelon Corporation (NYSE:EXC), showing an immense volume of activity over a small price range
Chart shows .001 second trading intervals for shares of Exelon Corporation (NYSE:EXC), showing an immense volume of activity over a small price range

"In the case of EXC, that means losing about 15 cents on every pair of trades. Do that 40 times a second, 2400 times a minute, and you now have a system that's very efficient at burning money," the Nanex report states.

That might all be well and good, but market-watchers may still wonder how, if Knight saw it was losing money so quickly, it allowed the rogue program to run for more than 45 minutes, an eternity in the world of high-frequency trading.

Again, Nanex delivers the answer, noting they "believe Knight accidentally released the test software they used to verify that their market making software functioned properly."

In other words, Nanex is positing Knight released two pieces of code into the market Wednesday morning, a trading algorithm to capitalize on market coditions, as well as a separate "tester"  -- used in Knight's software lab to create a "failure" scenario and see how the trading algorithm reacted. The result was the "tester" ended up wreaking havoc in a way that everyone else in the market took advantage of, saddling Knight with the huge losses.

"Since the Tester doesn't think it's dealing with real dollars, it doesn't have to keep track of its net position. It's job is to send buy and sell orders in test pattern waves. This explains why Knight didn't know right away that it was losing a lot of money. They didn't even know the Tester was running," the Nanex report states.

"When they realized they had a problem, the first likely suspect would be the new market making software. We think the two periods of time when there was a sudden drop in trading (9:48 and 9:52) are when they restarted the system. Once it came back, the Tester, being part of the package, fired up too and proceeded to continue testing. Finally, just moments before an economic news release at 10am, someone found and killed the Tester."

As hair raising as that conclusion finds, it is nothing in comparison to some of Nanex's previous reports. Below, an illustration from January of 2011 shows how the market for U.S. equity trading has absolutely been taken over by automated bot trading since 2007. Warning: the period between June 2011 and present is likely to cause sudden incontinence among the most squeamish readers.

Chart shows rise of high-frequency trading algorithms from 2007 to present.