Market Plunge Baffles Wall Street | Hacker News Reader

Market Plunge Baffles Wall Street

online.wsj.com

47 pointsgrellas48 comments

To protect against an unexpected market crash, traders have stop-loss orders that can be executed automatically if the market starts to tank.

  Trader 1 puts in a order to sell everything if the market goes down by 4% in 1 hour.  

  Trader 2 knows that and wants to get out of the market before trader 1 does in a crash, so he puts in a order to sell everything if the market goes down by 3.9% in 1 hour.

  Trader 3, Trader 4, etc. continue this process.

  Then you have a day like yesterday when the market goes down by 2 percent and there is a lot of uncertainty about Europe, causing more people to put in stop-loss orders.  The one automatic order triggers hundreds more.  

  Now they are going to rollback some of these trades because it was an "Trading Error".   When Investors on Wall St. have trades that make money it is because of their skill and they get bonuses.  But when their own stop-loss program sells for a 70% loss they get a do over.

> But when their own stop-loss program sells for a 70% loss they get a do over.

There has to be a mechanism to undo cascade mistakes because, if someone figures out a way (and that's rather easy) to induce mistakes and to profit from them, it will shortly become the prevalent form of trading. We want exchanges to foster investment on productive companies.

I was about to agree with the GPP, but you have a really good point. This would be the sort of loophole that would be exploited.

It's a market for intangibles; over the short term it's zero-sum. The only way to profit is to induce mistakes; to get someone on the other side of a trade that's profitable to you and damaging to them.

If algorithmic trading causes vulnerability to mistakes, they need to revise their algorithms or factor in the risk of this type of loss; that's the way free markets work. The externalities of a sudden crash suck, but other than as a one-time emergency measure, rolling back all the trades isn't a good solution. Wall Street firms should've learned something since 1987.

My guess is that the algorithms are already being adjusted. My fear is that a bunch of politicians who are as clueless as most everyone else are going to try use this as yet another opportunity to grab the spotlight. Expect congressional hearings.

This is the normal operation. The "undo" key is not supposed to be used unless something abnormal happens.

if someone figures out a way (and that's rather easy) to induce mistakes and to profit from them, it will shortly become the prevalent form of trading

That sounds contrary to what Goldman Sachs is saying why they shouldn't be sued. They are using the "big boys" defense.

http://www.cbsnews.com/stories/2010/04/21/politics/washingto...

Nobody sane would trust them to write the rules on which the market operates.

The word on the Street right now is that NYSE moved some stocks from normal electronic trading to human-mediated trading at 2:40 PM. They were trying to slow down unusual trading activity in those stocks, and they thought they could do that by giving their human market makers a bit of time to consider prices and do the right thing. Unfortunately, when they stopped electronic trading for those stocks, all the orders they would otherwise have handled got sent to a bunch of smaller electronic exchanges that didn't have enough liquidity. That started to tank a few specific stocks scarily fast, which triggered a bunch of stat-arb trades and plunged the indices down. Once that got bad enough, all the liquidity providers got spooked and backed out of the market for about 10 minutes. Felix Salmon has a good writeup at http://blogs.reuters.com/felix-salmon/2010/05/07/deconstruct... .

This isn't necessarily everything that happened -- there's a ton of other rumors flying around about the yen carry trade and a lot of other random stuff.

Here is the movement of Accenture as given by WSJ:

"With Accenture, for example, 20,365 shares changed hands at around $39.98 during the minute of 2:46 p.m., then another 68,516 shares were traded at $38 per share during the minute of 2:47. But then in the 2:49 p.m. minute, 66,277 shares traded at one cent. By 2:50 p.m., the stock was back up to $39.51."

The trade at 2:50 pm is the is the one that intrigues me , i can understand how the other ones can be blamed/attributed to HFT algorithms , but how does one explain the bounce-back at 2:50 from one cent to 39.51 ?

One or more large investors believed that the correct price for Accenture was about $40. When they saw it available at prices between $0.01 and $39, they bought and bid it up.

That's my problem with this - why would i pay $40 for something that is available at $0.01.

Yes i might believe it is worth $40 but when i am getting something for much cheaper than that surely i would just buy and wait for it to go up, in which case the price rise should also be in stages, which it is not.

It's really fascinating (horrifying?) that some of these ETFs went to zero for a moment. I can't understand how they cancel all these trades though. Seems like the biggest accounting nightmare ever.

It actually seems like standard procedure. They reset the market after 9/11, too. Basically, it seems like any time something really unusual happens, it's an excuse to go outside the rules. Back when I was involved in digital gold currencies, this kind of thing seemed pretty standard across the board: banks would just reverse transactions, freeze accounts, etc, any time there was anything unusual, and it seemed like our whole business was unusual, so we were in constant fear that money we'd received and disbursed would suddenly be unreceived after the fact. It was a nightmare, though; you got that part right.

My take away from my experience in that industry was that you can't actually trust banks or other financial institutions to follow the rules; the best thing you can do is to make sure you're lost in the crowd. Don't stand out, and you won't get hammered down. These trades and this whole episode stood out, so of course they're going to retroactively change it to the detriment of anyone for whom it was a good thing.

What rules are you referring too? Are you saying that Nasdaq violated their own rules when they rolled things back?

Oh, not specifically and officially; I don't know what actual contractual or legal rules surround this, and it could well be that trading is contingent on a contract that allows them to do whatever they want. By "rules", I mean the usual rules that people believe are in force, like "Once I buy or sell, it stays bought or sold". In exceptional situations (like this one, or 9/11), it always seems to turn out that honoring obligations was optional all along, doesn't it?

On certain exchanges, cancellations happen constantly. It's complex, but it's part of day-to-day operations and after a while it stops feeling unusual.

Theory:

1) Short DJIA.

2) Pick some stocks and place buy orders at $0.01.

3) Get a friend to fill those orders for you at $0.01, taking the loss. Call it a trader error.

4) A lot of poorly written algorithms, which take into account the last traded price, start selling to cover their stoploss orders (sell if the price < X).

5) Havoc ensues. DJIA is down. Cover your DJIA short and take the rest of the day off.

Expect your trades will get reversed at the end of the day and you will not make any money.

That's the best part. Only the trades that are directly related to the stocks that dropped are being reversed. Even if your trades on the stock are reversed, you have still moved the market and executed your DJIA options.

...and then, if your scheme works (which, as others have pointed out, would take a miracle, and a very carefully chosen market that's not handled the way most of the big ones are), both you and your friend go to jail for market manipulation, which would be rather simple to prove - there's almost no reason you would ever want to sell something way below the best bid, so you're without the usual excuses that market manipulators use to justify manipulative trades ("uh, my analysis showed that the 26.5 day moving average crossed the 100 day low, so I bought, but then I noticed that the pork bellies volatility was higher than usual, so I sold a few seconds later, because that's the way my strategy works...").

Practice.

Assume the stocks you are manipulating are bid at 99.99, ask at 100.

3 - Your friend's first few sell orders at $0.01 or better are filled at about $99.99.

Your buy orders go unfilled.

4 - If your friend sold enough shares, the algorithms notice someone aggressively selling. They may undercut and sell at 99.99 or even 99.98.

The algorithms also place a few buy orders at 99.96-99.97.

5 - Your friend's trades (assuming he is placing multiple orders, and is selling a lot) execute and drive the price down to 99.96-99.97 or so.

6 - The algorithm's buy orders at 99.96-99.97, are filled by your friend's sell orders.

7 - Price back up to 99.99.

8 - Algorithms eventually sell the shares they bought at 99.96 at 99.99.

Sum total: the algorithms made a few pennies off your friend. Your DJIA short does pretty much whatever it would already have done.

Hell, you don't need algorithms to do that. This is just what market makers do. Humans are a bit slower than computers, so the price may fluctuate by a few more cents, but the exact same order will end up happening.

If you have direct market access you can match particular orders in the order book, even if they are out of the money. Your friend would fill your $0.01 order at $0.01 and move the last trade price.

This is not technically possible for most matching engines. There is simply no "fill order X bypassing price/time queue" message. In fact, NYSE doesn't even tell you that order X exists and simply aggregates all orders into "Z shares available at price Y".

Also, except in the case of certain rare events which cause high latency (e.g., yesterday), it is also illegal to play games like this. For instance, if the bid is 99 on BATS and only 50 on ARCA, and I want to sell, I can't sell on ARCA. (I'm grossly oversimplifying, of course, but my simplifications don't exclude the case of selling at $0.01 to fool the markets.)

Reg NMS does not allow a trade to happen outside the national Best Bid/Offer.

If the market is 99.99-100.00 you cannot trade even at 99.98

Thank you, but I think you need to be more explicit at stage 7.

As I understood you, algorithms notice someone dumping stock and they pick it up. Then they hope you were selling for cheaper than market value because you were in a hurry, not because you thought said stock was overvalued, and they try to pass off the stock later on.

Is this what "providing liquidity" means? It seems like a good thing, iff people generally do high-frequency trading far more than actual investment. What's providing that confidence? Experience?

Pretty interesting and weird. Some stocks lost 100% of their value in a matter of seconds. Some analysists say it was because some high-frequency traders pulled out when the market became too volatile and thus there were too few buyers.

A lot of HF traders pull out in times like this for technological reasons. For example, most quotefeeds (such as Reuters, if my memory serves) have outdated technology that handles load really badly. When quotes are 5+ seconds slow, they're essentially useless, given that HF trading occurs on a millisecond timeframe.

Most liquidity providers would love to be in the market when it's panicking, because this is a great time to make a lot of money, but can't be there if their data is bad and they're at risk of enormous losses.

Reading this just makes me feel even more that the whole HF trading thing is the biggest con going out there.

Hmm. Speculators choose to avoid a market which they feel is highly risky. For this, you believe they are con artists.

Similarly, any person who doesn't put their retirement fund into junk bonds and penny stocks is also a con artist. After all, they are avoiding risky markets!

The biggest con out there is government employees making salaries that are on average double than private enterprise... but I guess that's a different discussion.

Actually, a much worse con, if you want to have this discussion, is that millions of private-sector corporate employees are getting paid peanuts for their work, compensated largely with a promise of future prosperity/advancement that will never come through. A much worse con is that the economic growth since 1975 has been siphoned off almost entirely by the increasingly entrenched upper classes, with consumer debt, a housing bubble, and student-loan debt bondage replacing wage increases to fuel the growth. A much worse con is an economy now plagued by individual insolvency and a widespread lack of trust, resulting from these catastrophic social changes.

Do you really think that it's of benefit to society for some people to show up to work every day and take home Wal-Mart wages, instead of having real alternatives?

A large share of this government waste goes to military contractors employed in completely unnecessary wars such as the multi-trillion-dollar mess in Iraq.

Also, please cite your "on average double" statistic. If my memory serves me, the discrepancy is about 30% on average, mainly because private-sector pay is far more recession-sensitive. (The discrepancy would shrink, if not reverse, in a growing economy with a healthy private sector.) For the record, low-level government employees do make more while upper-level people make considerably less than they do in the private sector.

> Do you really think that it's of benefit to society for some people to show up to work every day and take home Wal-Mart wages, instead of having real alternatives?

Given the prevalence of unemployment, yes Walmart jobs are better than at least one real alternative.

I would rather have a society where people are paid unemployment, even indefinitely, than one where they work under Wal-Mart conditions for Wal-Mart wages-- a circumstance that would be considered slavery by most societies dating back to ancient Greece, although we're afraid to admit that this arrangement is such today.

It's not, at least not completely.

Arbitrage and market-making are legitimate businesses that provide liquidity to markets, making it easier for other traders to get fair trades. (Whenever you trade, the fair value is assumed to be somewhere between the bid and ask, so you're paying about half the spread on every share you trade.)

Of course, many hedge funds out there are scams that exist to take fees while taking ridiculous (and poorly-disclosed) risks with others' money, but not all of them are.

It's not technical issues that cause HF to pull out.

The exchanges are breaking a whole bunch of trades and algorithms can't predict which trades will be broken. So imagine an algorithm bought at the bottom and sold halfway up the recover of yesterday's spike. Their buy orders at the bottom get broken. Their sells halfway up do not.

To meet their obligations, the algorithm must buy again at fully recovered prices. The process of breaking trades has turned a big profit into a big loss.

I still don't understand why anyone would put in a "sell at any price" order. If you sold Accenture at one cent because you put in such an order, well, you bought yourself a lesson to not do that.

Standard trading systems have stoploss orders (sell if price < X), which don't have an option for a lowest price at which to sell.

Because they are treated differently on the market.

Market orders (sell at any price) are handled first, and then, limit orders are. You can only guarantee one thing: either execution, or a price level. And when you're dealing with stop losses, you want to guarantee the execution in most cases.

Also note that any reasonable broker has both StopLoss and StopLimit orders. Where it's either a market, or limit order that only gets triggered when the price goes to a certain level. (you can stop up too. "Buy if price goes above X", useful for closing short sales, or just getting on a bandwagon).

Ah, thankyou. That makes a lot more sense now.

At the risk of seeming too fluffy, did we just see the market equivalent of a "rogue wave"? It would be interesting to see if there are any parallels between the two phenomena.

"The entire stock market rally which we have seen this year off the February lows resembles a low volume Ponzi scheme, and formed a huge air pocket under prices.

This US equity rally was driven by technically oriented buying from the Banks and the hedge funds. There was and still is a lack of legitimate institutional buying at these price levels. This was machine driven speculation enabled by the lack of reform in a system riddled with corruption, from the bottom to the top."

Translation: there's no real buyers in the market at the current price, so when panic came, there were no buys to prop up the free falling

link: http://jessescrossroadscafe.blogspot.com/2010/05/plunge-1000...

If there were no buys to prop up the free falling, why did the markets rebound almost instantly?

government intervention via JPM trading desks and trades cancellation from the exchanges

Do you have any evidence whatsoever for that? Or is it just a case of "when something I can't explain happens, it's because the government did it"?

Because liquidity returned to the market.

The problem was the NYSE paused trading for many of the stocks in question for 90 secs because of heavy sell pressure, driving those trades to other markets which had little to no liquidity for a few minutes (and are solely electronic). Compound that with a possible trading error, and HFT algorithms, and you had a short period of time where there were no bids for those stocks on those electronic markets.

When liquidity for those stocks returned to the market, the proper forces brought the stocks back to their bid/ask prices in a liquid market.

Reply on news.ycombinator.com