Talk:2010 flash crash/Archive 2

SEC/CFTC report and Kirilenko paper

This page needs to be updated to reflect the SEC/CFTC report on the flash crash and the Kirilenko paper on the event reported at the October 12, 2010 CFTC Technology Advisory Committee meeting. I'm happy to do that but I wanted to post this first - I'm a newbie here so before editing I wanted to engage the community before stepping on anyone's work.

Links for the above -

http://www.sec.gov/news/studies/2010/marketevents-report.pdf (SEC/CFTC report)

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1686004 (Kirilenko paper)

For example, the Wikipedia article says "Impact of High Frequency traders: Rejected by regulators" and that isn't the whole picture. Regulators say a large seller triggered the crash, but HFT firms exacerbated price moves. The Chicago Mercantile Exchange also takes pains to point out the "large seller" was just a small fraction of trading that day (overall 9%, and during the steepest part of the crash less than 5%). —Preceding unsigned comment added by MarketsGuy (talk • contribs) 18:49, 28 October 2010 (UTC)

Go ahead and edit the article. As long as you don't mind someone else coming in and changing stuff you've added, you've got nothing to worry about! --RegentsPark (talk) 18:52, 28 October 2010 (UTC)

@MarketsGuy The total percentage of daily volume the mutual fund's 4.1 billion dollar S&P500 futures contract short does not determine the price impact alone, but also the time period that order is unloaded unto the market. Typically traders use liquidity seeking algorithms to break large orders into smaller ones to reduce price impact and hide the size of the order from other traders who could take advantage of that information by trading alongside it. Had the order been 9% of daily volume over the course of the day it would not have been nearly as power to move the price as far as it did, what is crucial is the amount of time the order was executed in and it's sheer size. 9% is not a "small percentage" of total volume, something dramatically under 1% is, any directional order that large of a percentage is going to have major price impact, and if the order is executed without regard to consequence like the trader behind this order decided to then obviously it is going to have a large price impact, as it did. Arguably the sale was the flash crash. Also the articles cited show the explicit cause to be a mutual fund not one or even many HFT firms. Though they get involved later (and seem to lose money through market making activity due to the massive sell volume) the fact some HFT firms tried to exit the market after the massive sell volume resulting from the Waddell order is almost irrelevant. Even the official report only lists them as "likely buyers" at the start of selling (which would of slowed price declines by the way). What was not only the trigger, but the explicit cause was the 4.1 billion 20 minute rapid shorting of the S&P500. The research paper MarketGuy listed even suggests HFT do not accumulate large position or make one sided directional trades, in fact the long and short selling by HFT on that day was almost identical. Here is the reposting of the paper if anyone is interested: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1686004 To reiterate on Waddell's sale with a quote from the NYT article:

"At 2:32 p.m., a trader at Waddell & Reed placed a huge order to sell E-mini futures contracts, which mimic movements in the S&P 500-stock index. This kind of trade wasn't unusual for Waddell, which at the time managed some $25 billion, including the popular Ivy Asset Strategy Fund. As part of the fund's strategy, the firm from time-to-time places bets that the broad stock market will fall as a hedge against its individual stock holdings.

Also not unusual was that Waddell placed the trade using a computer program known as a trading "algorithm" designed to stand in for a human trader and parse out buying or selling based on different variables. Generally, traders opt for algorithms that consider trading volume, price changes and the amount of time to complete a trade.

But Waddell's desk opted for an algorithm designed to sell 75,000 E-mini contracts at a pace that would range up to 9% of trading volume—and not take into account other factors. The report details how a similar-size trade earlier in 2010 took five hours to execute, but in this case, the Waddell trade unloaded on the market in just 20 minutes." Financestudent (talk) 00:26, 31 October 2010 (UTC)

References

@Financestudent - Unless and until algorithms evolve into sentient beings, then of course there will be a trader behind them. The SEC, the CFTC and the press all attribute responsibility to the trader - who used an ALGORITHM - for starting the Flash Crash. Next, high-frequency trading firms exacerbated the selling pressure by dumping inventory when these firms hit their risk limits, just like they are programmed to do.

You miss the point and are a reductionist. Nowhere do I say "HFT caused" or "an algorithm caused" the Flash Crash. I say they are "implicated" or "directly contributed" or added to selling pressure or exacerbated price declines, all of which are amply footnoted to news reports and of course said outright in the SEC/CFTC report itself, or in Kirilenko's excellent paper. Would you like links to these? Finally, in saying that a trader did it, if your point is that human beings ultimately caused the Flash Crash, well, yes, I think we'd all agree to that. Human beings also caused all of the deaths in the sinking of the Titanic, though I think most accounts mention the ship and the iceberg too.

Financestudent took the axe to a variety of additions I made in the high-frequency trading entry, additions I made to reflect the September 30, 2010 SEC/CFTC report on the Flash Crash, claiming my additions were POV, vandalism, unsupported, unreferenced, or that the references didn't work. Those are quite remarkable claims and completely unfounded - and the revision history there shows that Financestudent deleted the references himself and then claimed he couldn't find them!

By all means look at the relevant citations (included here) and draw your conclusions. Were algorithms and high-frequency trading implicated in the Flash Crash? Did they have a role in the price declines that day? Are they the subject of intense focus now? Are regulators considering controls on them? The answer to all of these questions is yes, but Financestudent swings away with a variety of wild accusations about my own bias without once examining his own, or his own unsavory and ham-fisted editorial behavior.

Lauricella, Tom (2 Oct 2010). "How a Trading Algorithm Went Awry". The Wall Street Journal.

Mehta, Nina (1 Oct 2010). "Automatic Futures Trade Drove May Stock Crash, Report Says". Bloomberg.

Bowley, Graham (1 Oct 2010). "Lone $4.1 Billion Sale Led to 'Flash Crash' in May". The New York Times.

Spicer, Jonathan (1 Oct 2010). "Single U.S. trade helped spark May's flash crash". Reuters.

Let's add some more -

Goldfarb, Zachary (1 Oct 2010). "Report examines May's 'flash crash,' expresses concern over high-speed trading". Washington Post.

Popper, Nathaniel (1 Oct 2010). "$4.1-billion trade set off Wall Street 'flash crash,' report finds". Los Angeles Times.

Younglai, Rachelle (5 Oct 2010). "U.S. probes computer algorithms after "flash crash"". Reuters.

Spicer, Jonathan (15 Oct 2010). "Special report: Globally, the flash crash is no flash in the pan". Reuters.

MarketsGuy (talk) 17:00, 2 November 2010 (UTC)

Still missing the point. To quote your sources the conclusion of the research paper you posted: http://www.afajof.org/afa/forthcoming/6130p.pdf

"The declining costs of technology have led to its widespread adoption throughout financial

industries. The resulting technological change has revolutionized financial markets and the

way financial assets are traded. Many institutions now trade via algorithms, and we study

whether algorithmic trading at the NYSE improves liquidity. In the five years following

decimalization, algorithmic trading has increased, and markets have become more liquid.

To establish causality we use the staggered introduction of autoquoting as an instrumental

variable for algorithmic trading. We demonstrate that increased algorithmic trading lowers

adverse selection and decreases the amount of price discovery that is correlated with trading.

Our results suggest that algorithmic trading lowers the costs of trading and increases the

informativeness of quotes. Surprisingly, the revenues to liquidity suppliers also increase with

algorithmic trading, though this effect appears to be temporary.

We have not studied it here, but it seems likely that algorithmic trading can also improve

linkages between markets, generating positive spillover effects in these other markets. For

example, when computer-driven trading is made easier, stock index futures and underlying

share prices are likely to track each other more closely. Similarly, liquidity and price efficiency

in equity options probably improves as the underlying share price becomes more informative."

>>>>>I didn't post this paper. It also doesn't appear in the "Flash Crash" entry. It does appear at "high-frequency trading" because you posted it there on August 25. Now you say I did, which is, well, very odd. MarketsGuy (talk) 21:59, 10 November 2010 (UTC)

To quote one of the articles you posted: http://www.reuters.com/article/idUSTRE69E1Q520101015?pageNumber=3

All this tough talk has spooked high-frequency traders and the exchanges that rely on their liquidity and volumes. They note that HFT was not blamed outright in the SEC-CFTC flash crash report, and argue that its short-term strategies have made trading cheaper and easier for all investors. Richard Balarkas, CEO of Instinet Europe, the Nomura Holdings Inc-owned (8604.T) agency brokerage and alternative venue operator, said winding back the clock is a mistake. "I don't think investors on the whole want to go back to a market where they all pay a tax, usually in the form of a wider spread, to a firm making monopoly profits that will in any case wave a white flag as soon as a stock has a liquidity shock," he said in an interview. "It's crystal clear why the flash crash happened: a lack of buyers, and unthinking selling. It was pure, simple supply and demand within a regulatory regime that the SEC had created."

Sort of an important fact to leave out that the report isn't directly blaming HFT, the research study you posted on HFT said it should be encouraged in the conclusion, and that it was benificial, and that the initial selling from the mutual fund who did the 4.1 billion short not only triggered the even but was the cause from a simple supply and demand standpoint, which is what I was originally asserting. Hence why I also said that the fact the order was executed at an algorithm was irrelevant, even if it was executed with the same instructions by a floor broker the result would have been the same. Financestudent (talk) 19:02, 10 November 2010 (UTC)

>>>>>Reread your quote from Reuters, the reporter is relaying the sentiments of "high-frequency traders" and certainly not relaying the conclusions of the report. Nowhere do I say the report solely "blames" HFT for the Crash; as many secondary sources concluded, the report certainly says HFT contributed to the Crash. It's very relevant the large seller was executing by algorithm - his strategy simply could not be implemented by a human being. Also, if he had been a human on a floor, no doubt at some point other traders would have walked up to him or her and asked simply "What are you doing?" MarketsGuy (talk) 19:34, 10 November 2010 (UTC)

Claim that HFT caused the flash crash is incorrect according to the sources listed

If you read the article's used as references on this matter they are not claiming HFT or even the algorithm caused the flash crash that day but that the trader at Waddell chose to unload the 4.1 billion dollar position all at once. The fact he used an algorithm at all is entirely irrelevant, almost all trades of all frequencies are not computer (AKA algorithm) executed. If it was a floor broker selling 4.1 billion dollars of the S&P500 or 9% of it's daily volume in 20 minutes the result would of been the same because of the sheer size of the sell order and the speed period of time it is sold onto the market. Here is a quote from one of the articles: "But Waddell's desk opted for an algorithm designed to sell 75,000 E-mini contracts at a pace that would range up to 9% of trading volume—and not take into account other factors. The report details how a similar-size trade earlier in 2010 took five hours to execute, but in this case, the Waddell trade unloaded on the market in just 20 minutes." A human trader clearly made a bad choice, what these articles don't say though is that HFT caused the flash crash. They use titles like algorithms gone awry but the algorithm it's self wasn't the culprit, and no where in the articles to they explicitly claim that an algorithm is the cause, also they never state the trade was an error or accident, it acted exactly as the trader intended it to, the issue was the selling of a massive order in a short period of time, which of course had huge market impact. So to conclude regulators implicated High-frequency traders, or algorithms as the cause is a massive twisting of what is actually being said and what happened. I'm guessing the person who stated that was the conclusion has some sort of axe to grind regardless of NPOV or factual accuracy. I encourage anyone interested in the validity of the claims HFT was the cause of the crash, or even curious as to wither a single article linked to as a reference contains anything stating HFT is the explicit cause of the crash to look for yourselves, because the references are not stating that. I encourage any user that cares about the article's factual accuracy to start cleaning it up. Financestudent (talk) 00:06, 31 October 2010 (UTC)

References

Financestudent is a reductionist. Nowhere do I say either Waddell or HFT was the cause. To say it's irrelevant Waddell used an algorithm is to say it's irrelevant the Titanic was a ship or the Hidenburg was a blimp. And if you dispute that analogy, there is simply no doubt that the SEC/CFTC themselves believe the use of an algorithm was the trigger for the Crash, and no doubt they believe HFT exacerbated price declines - and hence no doubt both contributed to the Crash. The pace of the selling accelerated because the algorithm was set to sell 9% of the previous minute's volume - an act itself only possible with a computer algorithm. As that volume increased, the algorithm sped up. The price declines were then accelerated because HFTs engaged in "hot potato" trading at lower and lower price levels in an uninterrupted decline. The conclusion that both market practices were hence implicated or directly contributed to the Crash is entirely factual - and in fact exactly what the regulators said. As for axes - whether to grind or whether to swing, such as Financestudent did to hours of closely written and carefully documented text in the entry for high-frequency trading, accusing me of "vandalism" because he couldn't find the references he himself had just deleted - the only axe I have to grind on this issue is to report on what the SEC/CFTC actually said in its report. The conclusions about it detailed here come directly from the secondary sources I included as references and quoted extensively. MarketsGuy (talk) 16:53, 2 November 2010 (UTC)

A point by point response: "The pace of the selling accelerated because the algorithm was set to sell 9% of the previous minute's volume - an act itself only possible with a computer algorithm." You do realize that prior to computers they would of just sold in massive blocks, or all at once, which would of had more market impact. Demonizing computer algorithms massively misses the point that it was a large irresponsible trade that caused and arguabley even was the flash crash. "The price declines were then accelerated because HFTs engaged in "hot potato" trading at lower and lower price levels in an uninterrupted decline." To state this is to completely ignore there were and are other market participants who were selling. There was 4.1 billion dollars of supply coming onto the market from one firm alone. Obviously in this instance prices will decline as supply is much higher then demand and a new (lower) equalibrium price level will be reached. In addition to the one massive trigger order just imagining the number of stop orders alone that were triggered along the decline it is easy to see how the initial order by Waddell could expodentially gain market impact as the market declined. As for the HFT trading while this was happening you have to take into account that the firms that did that would be losing money all the way down making that particular situation self correcting. Financestudent (talk) 19:14, 10 November 2010 (UTC)

It's not really relevant to the point, but prior to computers a block desk likely would have phoned around, accumulated one or more contras, and posted a large block down; prior to computers, there would have been ample time for reasonableness checks or constructive feedback to the large seller. As the report concluded and the WSJ in particular noted, the algorithm enabled rote and uninterrupted selling in an fast-increasing feedback loop that accelerated the algorithm's selling as the response to it - including HFTs - itself accelerated. Your point "it is easy to see how the initial order by Waddell could expodentially gain market impact as the market declined" is precisely the SEC and CFTC's point - and that market impact was exacerbated by HFTs, where, at one point, inter-HFT "hot potato" was 50% of volume as prices collapsed. Regardless of all of this discussion, the issue at hand is simply what the report itself concluded, and a number of reliable secondary sources concluded the report implicated an algorithmic trader as triggering events and HFT firms as exacerbating price declines. Just a fact. MarketsGuy (talk) 19:47, 10 November 2010 (UTC)