Knight Capital, down 32 percent Wednesday, has lost about $300 million in market value in one trading session. Gads. What is the market saying? That there could be big losses from the loss of its market-making operations (they appear to be routing all orders to other parties), loss of trading opportunities, and perhaps to legal liabilities.
Does this sound like an overreaction? Maybe. But remember: yesterday UBS announced that it had lost $350 millionfrom the Facebook IPO and announced they were suing Nasdaq.
JP Morgan said in a research note that "Knight's potential hit to revenue could total $170 million, in a more pessimistic scenario."
Busting some of the erroneous trades would help Knight, of course. The NYSE has already announced that some trades in 6 stockswill be busted and that "no additional symbols will be considered for review."
I have been asked repeatedly about the NYSE's Retail Liquidity Program (RLP), which launched today. This program is designed to draw dark order flow away from competitors (like Knight). It is essentially a dark pool within the exchange.
The NYSE has said that "NYSE systems and circuit breakers operated normally" this morning, and it's likely that is true. I have also been told that the RLP operated normally. But it's also possible that there could have been a problem with Knight interacting with the RLP that was entirely Knight's fault. Orders may have been misrouted.
We don't know at this point. But given that this RLP program started today, and given that it is rare to have an order flow problem of this magnitude that happened this morning, it is worth investigating.
But let’s look at the bigger picture. Today we had a "technology issue" at Knight that caused massive volume in the opening minutes of trading. What's a technology issue? Seems like Knight misrouted trades in a group of stocks—148, to be exact. Strange number.
The common thread: All of it relates to technology issues. All three relate to the way trading systems interact with each other.
It suggests that we need improved interconnectivity between the many trading systems. It suggests the systems need more safeguards. It will also renew questions about the role that fragmentation plays in these technology issues. Specifically: Do we really need 18 exchanges and roughly 50 dark pools? I know it's tangential, but it's related.
It also suggests that it's good to have people standing around watching this stuff: For all the confusion, the NYSE floor traders reacted almost immediately to the flood of orders—most of them 100 shares repeating a "buy" or a "sell" order over and over.
The good news: The system, by and large, did immediately identify a problem and move to isolate the problems.
The bad news: No one can explain why the trading went on for so long. Here's one message I got: A trader wrote that the program "appears to sell 100 share lots about 10-15 times per second on some issues, and buy 100 share lots about 10-15 times per second on other issues...went on for 29 minute."
And it was orderly...stocks ran right up.
Aren't there circuit breakers that are supposed to kick in? Yes. A Volatility Trading Pause (VTP) kicks in when a stock moves 10 percent or more in the prior 5 minute period. But the VTP only kicks in at 9:45 AM ET, 15 minutes after the market opens. Many stocks had already run up prior to that period.
And some stocks just kept running up in a smooth line that apparently did not trip the VTP even after 9:45.
OK, if a rogue algorithm ran for 30 seconds, that I could see. But 29 minutes? Huh? How come Knight couldn't shut it off?
How about this for a technology improvement: an off switch!
—By CNBC’s Bob Pisani
Bookmark CNBC Data Pages:
Want updates whenever a Trader Talk blog is filed? Follow me on Twitter: twitter.com/BobPisani.
Questions? Comments? email@example.com