Quants and the crash | Hacker News Reader

Quants and the crash

wired.com

62 pointsMaven91140 comments

What a boring, presumptuous article ridden with hyperbole. I'm getting tired of these popular magazines picking up a cool-sounding subject, a recent mishap and selling the whole thing as if it will completely obliterate the old ways. Oh and add a few awesome words (algorithms, speed of light, 3 feet of fiber, quantum mechanics,...) that they hope will make the reader think he's in the 23rd century. Half of the article is basically repeated fancy sentences like "Faster and faster turn the wheels of finance, increasing the risk that they will spin out of control"...

Let's be serious. Value-based investment is not going away and high frequency trading is just one more avenue through which a decent number of companies, and so people, will take advantage of an opportunity and do well.

Who is supposed to be the target audience of Wired these days?

Further, many articles describing financial failures of a similar flavour will blame the "Quants", whereas in reality they have little power and merely provide the means to execute an idea. With nobody willing/able to regulate against such practices, they will continue to remain a valid avenue for value generation.

In the hand-wringing over Knight's mistake, the media seems to be missing the massive benefit to many other traders. Specifically, Knight kept buying certain stocks even as they were driving up prices of these stocks. Everyone who sold these stocks to Knight got the benefit of selling it for a higher price.

So, Knight didn't destroy wealth. It just made an unintended gift to people willing to sell them overpriced stock.

>So, Knight didn't destroy wealth

Sure it did, it destroyed its own wealth. Inefficient redistributions of money almost always destroy some wealth. The obvious way to see that in this case is that a 440M dollar loss took out 600M dollars of their marketcap.

That's perhaps not the best way to look at things. Two weeks ago Apple made a profit of $9 billion, but somehow still managed to erase $23 billion in market cap. Whose wealth was destroyed there?

The flip side of this coin is that when Knight "makes money" it isn't creating wealth either.

That's not true, because the idea that Knight's mispurchases were zero-sum isn't correct either.

All business enterprises, including stock trading, only operate when their returns can exceed the discount rate of the market [1]. This means that overall, the returns of all players in the stock market will tend to exceed 4-5% on average, and on a time-averaged basis. As a result, the S&P 500 has tended to yield an average return of 5% over the last decades [2]; it was higher before that when markets were younger.

What this means is that the stock market isn't zero-sum, but via the EMH, adds value to the economy. The first question you'll probably ask is, but how? The answer is that returns are made by making prices for assets, in this case stocks and other securities, more accurate. When the market has accurate prices it functions more smoothly [3], and this smoothness is attributed to stock investors, who make a profit from doing it.

This means that the market is benefited when prices are made more accurate--that is, when companies make profitable stock purchasing decisions. It's harmed when companies make unprofitable decisions. While other companies will buy Knight's errant stocks, their profits will ideally be less than Knight's loss, due to frictional costs. Goldman is just making a marginal profit on these purchases; Knight is taking a $440 million loss. Goldman isn't going to make $440 million on reselling Knight stock. This difference means that the transactions should ideally be a net loss to the market overall, because they were purchased in error and the time integral in the deviations caused in the affected stock prices is the factor by which the broader economy is harmed.

But yes, when the market functions correctly, it creates wealth by making prices accurate (price discovery [4]). When it functions incorrectly, a la Knight capital, prices are distorted and wealth (in this case the value of correct prices) is eliminated.

[1]http://en.wikipedia.org/wiki/Capital_asset_pricing_model

[2]http://en.wikipedia.org/wiki/S%26P_500#Total_annual_returns_...

[3]http://www.investopedia.com/articles/basics/09/the-function-...

[4]http://news.ycombinator.com/item?id=4226738

Like any trader, when they buy higher than anyone else, or sell lower the counterpartiess to that trade benefits by the margin. The fact they are trading also adds liquidity to the market, making it easier and cheaper for other traders to take and close positions more quickly than they otherwise could.

This whole process could easily be sorted by simply requiring that every quote is good for at least 1 second and that quotes are only updated every second, on the second timed from a basket of atomic clocks.

The chances of this actually happening - yes you guessed it - zero. Too many stakeholders making too much money on the status-quo.

I'd make it five minutes. If HFTs want to claim adding liquidity as legitimate value they're offering to the market, let them add it in a way that human decision-makers thinking about fundamentals can benefit from, rather than just gaming the way the market clears and scalping imminent trades. It's not as if the true value of a company can change faster than that.

You bet your bottom dollar it can. Have you ever seen activity on the CME when the non farm payroll numbers are published? Five minutes might as well be all week.

That's just the market converging on current reality, which changed gradually over the course of weeks. A company can't change its fundamentals (the goods and services it offers) continually in fractions of a second, so any such price fluctuations are noise rather than signal, and we shouldn't be using a system which amplifies and reacts to that.

A company can't change it's fundamentals, no, but the market it operates in can change in seconds. A major disaster, announcement of a government policy, or new economic information can become available that instantly changes the fate of a company, or even a whole industry. You really want to make laws that stop people acting on information? Good luck with that.

Nothing new here - the problems with computer aided stock trading go back two decades:

This was from an article in the NY Times back in 1989:

http://www.nytimes.com/1989/11/09/business/market-place-asse...

"Money managers engaged in the version of program trading known as tactical asset allocation, which involves using computer models to signal when to shift assets among stocks, bonds and cash equivalents like Treasury bills, concede their trades may be disruptive. But they disagree with Mr. Phelan that the technique is as potentially destructive as portfolio insurance."

I suppose what is new is the speed of execution.

The title is a bit misleading. What crash does it discuss?

I think the title refers to the Knight Capital Group incident which is referred-to in the Editor's note:

  One of the most interesting things about the catastrophe
  at Knight Capital Group—the trading firm that lost $440
  million this week—is the speed of the collapse.

I didn't realize the market crashed as a result of the Knight Capital incident. Hmm, let me go look at a stock chart:

http://www.google.com/finance?q=INDEXSP%3A.INX

I see a few 10-20 point losses at various times in the past month, but nothing that looks like a crash.

Yeah, "the" crash means Knight's fail.

The headline made me think is was the 2008 crash - caused by the same issues...

I'm disappointed by your lack of outrage, comrade.

Perhaps it refers to the crash of Knight's own stock price. I do vaguely remember a FT headline that day (Wednesday) about market prices falling due to the problem.

It appears this article was written and slated for a September publication before the Knight incident. While the Knight incident is mentioned in the leading editor's note, the crash most prominently dicussed in the article is the May 2010 'flash crash'. The title as submitted is vague and misleading.

I remember that. I was working in HFT at the time, so I certainly noticed when it happened.

But I decided to try an exercise. I went to google finance and graphed the S&P. From the graph (weekly data points) I can't tell when the flash crash occurred.

It's almost as if the flash "crash" was just a flash in the pan...

What I'm confused by is the way this article talks about lower-latency intercity links as valuable for HFT activities. Surely if you're an NY fund you just buy servers in Chicago or wherever. Can anyone explain why these low latency links could be valuable? (Over and above general infrastructure?)

They are actually trying to shave off milli and nano seconds in their trades being registered compared to the competition, right?

The exchanges are correlated so if you're fast enough you may be able to execute on arbitrage, I would assume.

Because futures and options are traded in Chicago and equities are traded in NYC (well north jersey) there are opportunities to trade equities and their derivative counterparts. For instance SPY (S&P 500 ETF) in NYC and ES (S&P 500 futures contract) can be traded against each other as they move. However the guy who gets their first will get most or all of the profit. So you need servers on both sides and the lowest latency possible links between them. For instance a microwave network is being built which shaves a couple of milliseconds off the existing special fiber cable that traders have been using.

Why don't you let the "quants" speak for themselves outside of pop-science news articles, they've built their own HN clone: http://quant.ly

Surely, unrealistic assumptions about rationality were the biggest factors leading to the crash - the Fed ignored all the hazards because it would be "irrational" for anyone to make a mistake.

Quants are mostly not big believers in rationality. They use it sometimes ("all other things being equal, the market is probably about right"), but if they didn't think people were making mistakes they would be trying to make money off those mistakes.

Seems like a lot of the R&D that HFT funds are interested in could have great side-effects for society. Nutrino messaging through the centre of the Earth; fleets of drone relays over the Atlantic; super fast internet infrastructure; research into trends of vast historical datasets; cross country wireless networks...

I'm loathe to listen to the advice to invest in the market for the long term when the HFTs (who are making lots of money) have driven the average hold time down to under 30 seconds.

It make me feel like I'm being encouraged to put by money in the market so ensure that there's money for the HFTs to steal.

If you buy and hold, the hft leeches won't have much opportunity to steal your money...

if they manipulate the market downwards my investment. is worth less. I'd buy a stock long term for dividends. But since dividends are apparently rare these days (something about the evil of double taxing), we're forced to invest in the hopes that somebody will believe it's worth more later. What used to give stocks some worth was they value of the dividend, that you'd reinvest, to compound over the long term.

Without dividends, I think you should just invest in currencies. It's just as volatile, and you don't have to lie to people about the nature of the marketplace.

The obvious solution is to take a short position so then you profit when the market is manipulated downward.

That's why they fail: they are running their programs on batteries instead of servers.

http://www.wired.com/images_blogs/business/2012/08/ragingBul...

Can someone explain why trades are done continuously in real time? Wouldn't it make more sense to have a turn-based exchange with a turn of one minute or an hour?

What do you do if 10 people try to buy a stock at a certain price when there is only one on sale?

Lots of possibilities, but the simplest way to modify our existing sysyem is probably to process requests in random order within a turn.

I think the market makers (who are often just skimmers) and Investment Banks have cooked their goose. How many 'muppets' have now left the market never to return because of all the outrageous revelations, scams and bad news?

Reply on news.ycombinator.com