Log in

No account? Create an account

The · Psychohistorian

Catching falling knives

Recent Entries · Archive · Friends · Profile

* * *
There's a phrase among stock traders: "to catch a falling knife". This refers to buying a stock while it is still falling, in the hopes that one is buying at or close to the bottom and will benefit from the entirity of the subsequent recovery. As the phrase implies, it's a dangerous thing to do, as more often the stock continues falling. It's generally considered safer to wait until the stock has bottomed and started recovering before buying.

Of course, if no one ever tries to catch a falling knife, then once a stock starts declining it will go all the way to zero. This points up the critical role knife catchers play in the stock market: they are the ones who keep stock market slides from becoming stock market crashes, by stepping in to buy while the market is still declining.

That's why the exchanges' plans to cancel orders that went through today at particularly low prices is a bad idea. Days when knife catchers catch the hilt instead of being sliced by the blade are rare enough already. Usually the knife catchers' assets continue to decline - and cancelling their profits when they do make money just discourages them further. That could mean they won't be there to buffer the next crash - and that next crash could well have happened today without them.

Image showing how the popularly reported 'precipitous drop' today was just the culmination of hours of continuous deterioration:
* * *
* * *
[User Picture]
On May 7th, 2010 07:34 am (UTC), izmirian commented:
Yeah, canceling orders just because they went through at low prices seems like a bad idea.
* * *
[User Picture]
On May 7th, 2010 12:40 pm (UTC), jaedian commented:
that sets a bad precedent. What is their justification for this? Was there some kind of glitch? Otherwise, I can't see how they can do that - the sellers agreed to sell at that price.
[User Picture]
On May 8th, 2010 02:48 am (UTC), psychohist replied:
The exchanges have broad latitude to cancel trades, though this is probably pushing things. I expect they'll get away with it primarily because it's easier for the cancelled buyers accept the cancellation and modify one's future behavior than to fight it. Also, people who lose a lot of money on a bad trade that doesn't get cancelled get a lot more sympathy from politicians and the press than people who stood to gain a lot of money on a good trade that does get cancelled; the good traders tend to be characterized as "greedy" if they complain, while the bad traders are portrayed as victims.

The trades being cancelled appear to be legitimate; this kind of thing can easily happen when a significant decline starts, because the decline triggers stop loss orders to sell, which can outnumber the limit orders in the "order book" to buy at less than the current price. The markets cancelling the orders - NASDAQ and NYSE - say their software was working fine.

There have been some rumors about trading errors but they seem to be just rumors. 14,000,000,000 shares of S&P500 futures changed hands in a short period of time, leading to rumors that a Citigroup trader accidentally placed a sell order for 14,000,000,000 shares instead of 14,000,000 shares; however, all of Citigroup together actually only traded 9,000,000,000 shares of those futures for the whole day, so that wasn't the case. When values change rapidly, volumes are usually large anyway due to the aforementioned stop loss and limit orders.

There was what could be considered a computer glitch, but it didn't have to do with trading. The computer that calculates the DJIA got behind due to the large volume. It then got overloaded, and the backup computer had the up to date number. The switchover could have looked like a sudden movement, but the number was correct; a slightly less sudden but earlier drop might have led to less of a subsequent drop, but it might also have led to more of a subsequent drop.

Now, there's no doubt that some of the stop loss and sell orders were the result of computerized trading, but they were the result of computerized trading working as the traders had programmed, which I personally wouldn't characterize as a 'glitch'.

The truth is, the graph above seems extremely predictable to me. The market has been sensitive ever since 2008, and the "stalagtite" seen in graph is exactly what one might expect: as the numbers deteriorated over the afternoon, more people - and computers - decided to sell, and more stop loss orders were triggered, accelerating the decline. Then when the decline hit the psychologically important number of 10,000, it bounced due to strong support at that number.
[User Picture]
On May 8th, 2010 02:56 am (UTC), harrock replied:
From this, it sounds like at least some of the trades were actually full-bore glitches. But the article contains too many pesky generalizations to be sure what they're saying.

I think that what that article is saying is that when the NYSE paused trading in some stocks, some sell orders were automatically transferred to exchanges where no buy orders existed, which caused two things to happen:

1) It sounds like they're saying that some system on NASDAQ entered some default bids at $0.01 per share. (On behalf of whom? The market makers is the only thing that makes sense to me, unless I'm misunderstanding their explanation.)

2) Anyone who realized what was happening could then enter whatever bid they felt like on NASDAQ for these stocks, and they would be filled. Presumably some of them entered bids at a penny, and others entered bids at middling values.

Presumably the transferred orders were "market sell" orders, which work approximately as you'd expect in a liquid environment, but have this obvious failure mode when there are no buy orders except the ones which you consider to be crazy-low. (Lesson: use limit orders if you aren't truly willing to sell stock for a penny a share, or buy it at infinity per share...)

I think that before I draw conclusions, I want to know a little more about what actually happened and what people had reason to expect. On one hand, I expect anyone who buys and sells stocks to understand what "market order" means. On the other hand, if I'm a seller, I may reasonably be surprised to find out that the fine print of my broker agreement says "in the event of a liquidity disruption, your order may be sent to the Elbonian Stock Exchange, where prices of buy orders may be denominated in yak hair, and where you will have no recourse if quoted prices are set to zero for no apparent reason." (Damn. Excuse me...now I have to go look at my broker agreement...)

NASDAQ said they're cancelling trades that were more than 60% lower than the 2:40pm price yesterday. Nobody has published any stats on what dollar value of trades were wiped out as a proportion of those that went through during the blip. Some very big stocks dropped by a lot, but didn't drop by 60%, so all of those trades went through. I suspect that a ton of money was made by people who were paying attention at the time, even with the cancellations.

60% seems pretty arbitrary; I have no idea why it was chosen. (And apparently, companies whose stock was heavily affected don't know either...they read it in the news like everyone else.)
[User Picture]
On May 8th, 2010 03:38 am (UTC), harrock replied:
As psychohist points out above, I've neglected to consider stop loss orders, which are the normal method for this kind of thing to happen.

Nonetheless, unless someone tells me that there were no buy orders above $0.01 for these stocks on NYSE while trading was paused, then it's pretty clear that the penny trades went through not because there was no demand, but because the demand was locked out. Not exactly a natural market phenomenon.

Really, what we need here is for some economic conservative to point out that this is a good real-world example for how simple and well-intentioned rules can screw up a market by interacting with a complex reality. Circuit-breakers sound like a fine idea, but it sounds like in practice, suspending trading on the NYSE just shoved activity into other exchanges that were (momentarily, at least) much less liquid.
[User Picture]
On May 8th, 2010 03:54 am (UTC), psychohist replied:
I read that article differently from you.

First off, NYSE didn't transfer any trades. NYSE delayed trades, causing traders who wanted fast execution to switch to NASDAQ to place their trades instead. The traders switched voluntarily; they weren't switched involuntarily.

Secondly and more importantly, the traders that the article focuses on weren't humans; they were high powered computers. These are the "computers that found no bids" mentioned in the article. There's a technique that some trading computers use to probe the order book by placing orders and then cancelling them near instantaneously, in the couple of seconds allowed for cancellation. With the NYSE slowdown, these computers would have found a much thinner order book, resulting in their "panicking". So basically the people being bailed out here are people who can afford computers faster than the stock exchanges' computers - big guys like Goldman Sachs.

Thirdly, while that explains the NASDAQ cancellations, it doesn't explain the cancellation of trades that were actually done on the NYSE.
* * *
[User Picture]
On May 17th, 2010 08:18 pm (UTC), psychohist commented:
Here's an article on someone who made a panic sell of his entire portfolio, and feels gypped because he lost $20,000 when his sale of his Proctor & Gamble stock went through at 39 instead of 60. His trade didn't get cancelled, but it's an example of how it's poorly thought out trades that tend to lose money in this kind of situation:


Adding to the irony is the fact that he was manager of operations at GMAC's mortgage operations unit for six years while they were originating $75,000,000,000 of a mortgage portfolio that contributed to the housing meltdown. I'm guessing he's not offering to help out the underwriters of any of those mortgages that went bad.
* * *

Previous Entry · Leave a comment · Share · Next Entry