"We risk becoming the best informed society that has ever died of ignorance"
- Rubén Blades

"You can't make up anything anymore. The world itself is a satire. All you're doing is recording it"
- Art Buchwald

"It's getting exciting now, two and one-half. Think of everything we've accomplished, man. Out these windows, we will view the collapse of financial history. One step closer to economic equilibrium"
- Tyler Durden

"It is your corrupt we claim. It is your evil that will be sought by us. With every breath, we shall hunt them down."
- Boondock Saints

Tuesday, September 20, 2011

It's Official: HFT Breaks The Speed of Light

On September 15, 2011, beginning at 12:48:54.600, there was a time warp in the trading of Yahoo! (YHOO) stock. HFT has reached speeds faster than time itself. Up to 190 milliseconds into the future, or 0.19 fantaseconds is the record so far. It all happened in just over one second of trading, the evidence buried under an avalanche of about 19,000 quotations and 3,000 individual trade executions. The facts of the matter are indisputable. Based on official exchange timestamps, there is unmistakable proof that YHOO trades were executed on quotes that didn't exist until 190 milliseconds later!

Millions of traders depend on the accuracy of exchange timestamps -- especially after bad timestamps were found to be a key factor in the disastrous market crash known as the flash crash of May 2010. We are confident the exchange timestamp problem has been completely addressed by now: the SEC would have made sure of it. Adding accurate timestamps is not exactly rocket science; it's not even considered to be a difficult problem. Based on recent marketing materials, the exchanges are practically experts on measuring time. And with hundreds of millions in annual data feed subscriptions paid by the same subscribers expecting quotes with accurate timestamps, there is no shortage of funds to make it happen.

So we can be certain the exchange timestamps were accurate, which means that HFT has truly entered the era of the fantasecond.

But let us suppose for a moment that in reality, quotes became queued (delayed) and were timestamped after leaving this queue. After detailed analysis of the UQDF data feed (see chart below) that transmits this information to traders, we find that the traffic rate for all output lines and specifically multicast line #6 which carries YHOO, were well below peak rates. So it doesn't appear there were any capacity problems which have always been an excellent indication of feed delay.

This raises a few thorny questions.

Does this mean there are far more delays than previously thought? Is there a delay every time we see an explosion of quotes in one stock? Because that sort of thing happens. All the time.

Regulation NMS is pretty clear that direct exchange feeds are prohibited from having a speed advantage over the UQDF data feed. UQDF computes the NBBO after all. So how does one ensure trade-through price protection if the price being protected hasn't even occurred yet? The NBBO lies at the heart of Regulation NMS (Reg. NMS) and is the key concept that assures investors are getting the best price when buying or selling stocks.

Maybe it would be better to just fantasize about fantaseconds after all.

Zoomed in detail of above chart in 2 millisecond intervals. 

The next chart is the same as the one above, but includes trade executions which are plotted as dots or squares and are sized according to trade size. Each reporting exchange uses a unique color and shape. An exchange code table appears at the end of this page. What is unusual about this chart is that trades are reported ahead of quotes. The dots/squares should trail after the quotes. Up to the point in time that is labeled "A", and after the point labeled "B", trades and quotes were in sync. Between these two points is where the quote timestamps began falling behind.

By plotting quotes and trades from just one of the active exchanges, we can easily measure how far in time the trade messages came before the quotes and therefore estimate the minimum amount of time the quotes were delayed. Below is a 1 millisecond interval chart of YHOO showing only Nasdaq trades (black circles) and the Nasdaq bid-ask spread (gray vertical bars).

Exchange Code Legend:

1 NQEX Nasdaq Exchange
4 AMEX American Stock Exchange
8 CINC National Stock Exchange
9 PHIL Philidelphia Stock Exchange
11 BOST Boston Stock/Options Exchange
57 NQNX NSX Trade Reporting Facility
59 NTRF NYSE Trade Reporting Facility
60 BATS BATS Trading
63 BATY BATS Y Exchange
64 EDGE Direct Edge A
65 EDGX Direct Edge X

The chart below shows quote message rates for UQDF and multicast line #6 which is the line that carries YHOO quotes. The time of the event is shown at the lower left. Although traffic from YHOO was a significant percentage of all traffic on UQDF, it was still not high enough to indicate any problems. Note that the surge on the right side of the chart is much higher, and there weren't any known problems at that time in YHOO.