Computer Glitch Causes Havoc and Losses on Nasdaq

goombah99 writes "In an illustration of how fragile the electronic stock market system is the NY Times is reporting how a tiny computer glitch rippled through the Stock Markets with buyers who bought low and sold high taking huge losses. An erroneous large sell order was entered. Many people bought at this low price, then signed options contracts to sell these at higher prices, locking in a profit. Or so they thought utill the erroneous low sell order was removed. Now to honor their options they had to buy the stock at a higher price. Since exchanges trust each other's trade prices it rippled throughout the system. There does not seem to be any way to gracefully undo such errors."

Comments

Trading has its risks

[stomv]

and while the SEC and others do their best to proteect traders, mistakes do happen. This is part of the random process of the markets, and must be accounted for when making a trade, even on options markets.

If you lost money, sorry. Unless the SEC/others can prove that somebody is liable for the initial mistaken order, you lose. Tough. Trading is risky, and sometimes the risks are completely unforseen.

Re:Trading has its risks

[wizrd_nml]

That's pure nonsense.

The markets are meant for people to invest their money in businesses they feel will make a decent return for them. Investment risk consists of inherent risk of the industry, currency risk, political risk, etc. Nowhere in that equation is there EVER risk of a glitch in the computing system factored in.

Mistakes happen because people are unethical, criminal, or just dumb managers. But mistakes should never ever happen because the system that you gave an order to buy or sell for you decided to have a glitch.

Someone IS liable. NASDAQ is liable! NASDAQ is a company and it will be sued for the losses that it caused other people. It's as simple as that.

Re:Trading has its risks

[bagsc]

You're at least half right. But this isn't a question of information about prices - if it were, NASDAQ and the ECN would be off the hook. This is a question of whether the ECN executed orders that should not have been executed and NASDAQ didn't cancel them all. That's when risk minimizers are hit.

If I market buy because the ticker says a $50 stock is selling at $40, it goes through between 10:46 and 10:58, then NASDAQ is right to cancel at 12:30: no problem. If my option straddle executes on the volatility on both sides, one before and one after 10:58, but NASDAQ cancels the options in-the-money (on Instinet) but not the options out-the-money (on another ECN), then its a problem. If I'm an idiot, and leave those options open unchecked through a halt, then its my fault for engaging in a risky behavior and getting slammed in the ensuing short-squeeze.

Other stocks in the sector were off by 10%, so it was not stupidity to think that a 20% move was legitimate. NASDAQ halted Instinet, but not other ECNs. Archipeligo already announced intentions to file suit with the SEC on the matter. And that won't be the last suit filed on it.

Hyper-transactional databases?

[mcrbids]

My first thought when reading the summary above was that this would be an easy problem if managed by a central, relational database system.

Simply "roll back" the transaction that failed, and the dependencies would cancel themselves out. But, then I realized that the current RDBMS model only allows for a single transaction - you can't nest them.

Also, transactions are private only - you cannot transact with data in the middle of another transaction.

Thus, you might have ACID compliance, but only with one level of "undo".

How hard would it be to create an RDBMS that supports infinite levels of "undo" or transaction/rollback.

Such that you commit transaction A, which affects rows 1,2,3, and 11. Then, another transaction B which affects (further) rows 2, 3, and 12.

Then, if you roll back transaction A, transaction B would be similarly affected. I dunno - the depencies may get rediculous - but it seems that this could and should be done at some point.

Bright idea? Or another noise from an unpleasant orifice?

Let me know what you think!

Re:Not always possible

[gertsenl]

I say there's a real simple way to solve this, no logistic or legal mess. Make them make good on the original sell order. They, in turn, want to sue the software developer? Let them handle that on their own time and out of THEIR pockets.

Re:The cancel probably shouldn't have happened

[LostCluster]

However, Archipelago was the first to make the decision to resume trading, so most of the people who got burned did so there. NASDAQ then was caught in a no-mans-land of decision making... their investigation hadn't yet returned an explanation, but Archipelago's actions indicated that they had already made a decision that the trades were going to stick. For a trading halt to be effective, there has to be a trading halt everywhere. The markets should have seperate regulatory divisions, but they all should be coming to the same decisons at about the same time. Archipelago clearly didn't do a good investigation here... that's the question that needs further investigation.