Use access key #2 to skip to page content.

May 6th Flash Crash Cause Found



June 24, 2010 – Comments (11)

 I saw this story on today, and thought it is worth bringing to attention here.  A company called Nanex has published their analysis of the May 6th "Flash Crash", and found a compelling cause/storyline.  Here is their analysis, I recommend reading it - though it is quite long and detailed. 

 The basic story is this:

  On May 6th, some nefarious HFT companies started sending 1000's of dummy orders per second for a single stock.  This would be no monetary gain or point in doing this action, except for the fact that by sending so many bogus orders into the NYSE queue, you could effectively be performing a denial of service attack.  The queue would become slowed down, and your competitor HFT companies would be slowed down.

 The side effect of this was that when the NYSE queue slowed down, it fell behind the queue's of the other stock exchanges that are electronically synchronized.  Worse, there is a problem with the system in that the orders are not time stamped from when they entered the queue but when they finally leave the queue.  SO, when the queue got overly bogged down there were old orders that were now above or below the market prices of the other exchanges - and it looked like these orders were new.

 Now, the HFT systems start seeing falling prices on the NYSE because the orders there are below the prices of the other exchanges (these are the old orders, that look new with their timestamps).  The HFT systems move towards shorting these stocks to go with an apparent momentum (non-existent, simply old orders that are being shown as new leaving the slow queues).  Their sell orders start hitting the NYSE and further slowing the queue.  These orders also leave with a delay - and a selling cycle starts because of a feedback loop with the slow queue / timestamp issue.  Many of the stocks affected were big bellweather stocks - and likely triggered specific responses for selling as well.  Further issues was that HFT companies pulled the plug on their trading completely when things looked weird.


  This is as good as I was able to get in a quick read, I may have some of it slightly wrong.

  But the general conclusion seems to be that there are now some cheating HFT companies that are purposely slowing down the queues - and that the NYSE has should have a simple fix to the particular Flash Crash cause - fix the queue timestamping.

  Companies that stuff the queues with bogus orders probably need to have the SEC come after them.


11 Comments – Post Your Own

#1) On June 24, 2010 at 4:01 PM, ElCid16 (94.95) wrote:

Interesting...thanks for posting.

Report this comment
#2) On June 24, 2010 at 4:17 PM, XMFSinchiruna (26.50) wrote:

Interesting ... thanks.

Report this comment
#3) On June 24, 2010 at 4:29 PM, rofgile (99.29) wrote:

Here is a good quote:


On the subject of HFT systems, we were shocked to find cases where one exchange was sending an extremely high number of quotes for one stock in a single second: as high as 5,000 quotes in 1 second! During May 6, there were hundreds of times that a single stock had over 1,000 quotes from one exchange in a single second. Even more disturbing, there doesn't seem to be any economic justification for this. In many of the cases, the bid/offer is well outside the National Best Bid/Offer (NBBO). We decided to analyze a handful of these cases in detail and graphed the sequential bid/offers to better understand them. What we discovered was even more bizarre and the only conclusion, outside of stupidity, is truly frightening. You can see our results in Part 3, Quote Stuffing."


Report this comment
#4) On June 24, 2010 at 4:37 PM, rofgile (99.29) wrote:

They also show that this problem with the queues deaying orders and getting out of sync happened on two other dates (1/28/2010 and 10/30/2009) - implying that this Flash Crash cause is not such a one-off event.  Neither of these times led to a crash behavior - but there were similar numbers of stocks being mispriced at the peak of each event compared to the Flash Crash, with approximately 200 stocks being out of sync.  

Makes me wonder then why the 5-6-10 incident had such completely wild affect effects compared to these two times when there was not wild price action.



Report this comment
#5) On June 24, 2010 at 4:44 PM, rofgile (99.29) wrote:

This page is really fascinating -  

 You can see the price delay effect in real time for stocks like GE when you compare the NYSE prices vs PACF and NASD.  When the price crashing is happening, GE's price on the NYSE looks time-delayed compared to the other exchanges.

 When the quick recovery happens, there is no time delay difference.

Report this comment
#6) On June 24, 2010 at 5:25 PM, rofgile (99.29) wrote:

Here's my basic graphical summary:

  There are many computerized markets in the US that are synchronized.  Normally, there isn't much of a price difference between the markets.  If there is a difference of a couple of cents, an HFT will sell on the more expensive market or buy on the less expensive market - taking a couple cent or less profit.

 Now, the problem is that the time stamps get set for orders when they leave the queue - not when they enter the queue.  If a queue gets bogged down, the prices will be old on one market, though they will still appear to be time stamped as if they are current prices.


 So, enter in "Bad HFT" -

 Bad HFT starts to fill GE or PG queues on the NYSE with bogus buy and sell quotes - thousands per second.  The queue of prices on the NYSE starts to get delayed from other exchanges.

 Now the HFTs all send their sell orders to the NYSE to make a couple cents profit off of the apparent price discrepency (which is actually a TIME discrepency in prices).  All their orders enter in the queue - and now you've got a big cycle started.

 HFTs will keep hammering the NYSE with sell orders until the queue gets chewed through and the time discrepency wraps up. Crash is over, The selling pressure on NYSE disappears and share prices float back up in sync on all markets.  


Report this comment
#7) On June 24, 2010 at 6:15 PM, ralphmachio (< 20) wrote:

That was not a mistake. That was a test of the system. They have been adding support for a long time, they see a day soon where they must discontinue practice, they wanted to see what would happen if they removed support on a bad day. They found out. 

Now ask yourself, "Are they likely to print that confession anytime soon?".

When the truth could be inconceivably worse than most consider fathomable, you must use 'gnosticism' in order to get the 'whole story'. In other words, I pulled this directly out my a$$, but that does not make it untrue, and I have a tougher time believing any of the other stories, but hold them as potential truths as well.

It is a major disadvantage that most investors have spent little time if any with real criminals, police, or politicians. They have no idea who they are dealing with, and it is only a question of time before this trash shows it's true colors. Hey, if a suit and some fancy jargon is all it takes to impress you, you really, REALLY have it coming! You will soon find out that the worst criminals have gravitated to positions of power. 

PS. Some policemen are really nice people, with good hearts, and mean no harm to anyone, as well as an equal cross section of criminals. Politicians are a foul species entirely different from you, and even I.

Report this comment
#8) On June 24, 2010 at 8:42 PM, rofgile (99.29) wrote:


 The evidence this group is providing, see comment #5, is that what happened was a time lag between exchanges.  The NYSE lagged in price data on key stocks such as GE.  As these stock prices on the NYSE appeared to be higher than the average price on the other exchanges, this led HFT algorithms to attempt to sell on the NYSE on those stocks to attempt to profit on the difference.

 The problem was that these stocks weren't priced differently, just that the data had lagged.  So, now all the sell orders shoot in to a lagged system and continue to put selling pressure until the queue of orders is worked through and there is no longer a delay between the exchanges.  At this point on the charts the prices of all these affected stocks quickly recover.  

 i.e. this was not the removal of some support - there is no such thing.  This was HFTs gone wild due to a bug in the NYSE where they time stamp stock prices as orders leave the queue rather than as the time the order first entered the queue.  This combined with some apparent attempt by an HFT to purposely slow down queues on stocks with bogus orders lead to the crash.


Report this comment
#9) On June 24, 2010 at 9:07 PM, russiangambit (28.71) wrote:

> So, now all the sell orders shoot in to a lagged system and continue to put selling pressure until the queue of orders is worked through and there is no longer a delay between the exchanges.

These would've been limit orders though so if there were no buyers at that price the order would've simply cancelled instead of executing. So , while I see how such a deluge of orders can bring the comuter system down or delay it I don't see how they would push market down 5% in 5 min.

Report this comment
#10) On June 24, 2010 at 11:18 PM, ralphmachio (< 20) wrote:

I watch the markets, minute by minute. There's support when it is in someone's interest, but I hold many truths as possibilities, which reduces my surprise.  You could be right.  Like I said, if it was a test, they could not admit it, because it would negate the reason for having a ppt, to keep the suckers in the market until they are to be spooked out in a controlled manner, with all the big guys having their puts in place. What, did you think they were gonna steal from Peter to pay Paul and then let a bunch of beginners and amateurs make off with the loot? Did you really think they planned to have a lasting effect with that stimulus, or was it about politics, and thievery?


Report this comment
#11) On June 25, 2010 at 6:04 PM, rofgile (99.29) wrote:

I actually emailed the SEC about the Nanex report yesterday, though I am pretty sure Nanex would do this themselves.  So far I haven't heard anything.  

We'll see if the MSM covers this story anytime soon.


Report this comment

Featured Broker Partners