How High Frequency Traders Broke, And Manipulated, The Treasury Market On October 15, 2014

We were amused to read some interpretations of today's long-awaited joint-staff report [18](prepared by the Treasury, Fed, SEC and CFTC) attempting to "explain" the flash smash in Treasury prices on October 15, 2014 when as a reminder, Treasury prices exploded and yields plunged just around 9:34 am [19] from a level of 2.20% to just over 1.95%...

... which did everything in their power to deflect attention from High Frequency Trading.

Case in point: the WSJ [22], which earlier today said that "U.S. officials concluded there was “"no single cause" of the unprecedented volatility that hit U.S. Treasury markets Oct. 15, 2014, citing instead broad changes in the structure of Treasury markets, including the growing role of high-speed trading."

To an extent the WSJ is right: the report does everything in its power to obfuscate and shift the blame away from the true culprit: "For such significant volatility and a large round-trip in prices to occur in so short a time, with no obvious catalyst, is unprecedented in the recent history of the Treasury market," the report said of the events on October 15.

To push the official position, the WSJ adds that "the staff examined nonpublic trading data before, during, and after a 12-minute window during the morning of October 15 in which the yield on a key U.S. Treasury note plummeted, then quickly rebounded. They concluded there was “no obvious catalyst” in the news that morning."

That is correct: the plunge came out of nowhere, with no catalyst at all - something market watchers see every day in equities, commodities and FX - and if it had all the telltale signs of an HFT momentum-ignition stop hunting spasm gone horribly wrong, it is because that was precisely what it was.

Which is why the WSJ promptly trotted out some of the most washed out and discredited experts it could find: such as Irene Aldridge.

Irene Aldridge, a markets expert at ABLE Alpha Trading Ltd., who sits on a subcommittee addressing high-frequency trading at the Commodity Futures Trading Commission, said her firm found “a spike in abnormal market activity” on October 15.

 

She said she didn't see much evidence that high frequency trading firms were chiefly to blame for the spike, and instead pointed to a wide array of macroeconomic forces and large cross-border trading activity that occurred that day and in the days prior.

The fact that Irene Aldridge who once upon a time, long before HFT started strategically crashing every single market in which they were introduced courtesy of Getco, Virtu and Citadel, appeared with Jon Stewart's "cash cow" to "explain" HFT... 

... is less troubling than the fact that she actually sits on a CFTC HFT subcommittee because she is perceived to be an expert.

In any event, while the WSJ did everything it could to defend HFT, Bloomberg apparently [27]was unaware that it had to pick its words carefully and do all in its power to defend Citadel's (the NY Fed's preferred arms-length trading venue) market manipulation in every asset class, and ahead of the report's release said [27]that "U.S. officials have concluded that high-frequency trading contributed to the Treasury market’s wild ride last October, a finding that will probably add to regulatory scrutiny of the industry."

Further proof that Bloomberg's Ian Katz did not get the memo that while the joint-staff report identified HFTs as the culprit behind the Oct. 15 "anomaly", such an assessment was not to be made publicly was the next paragraph:

While a soon-to-be-published government report won’t point to just one cause, it will cite speed traders as playing a key role, according to a person with direct knowledge of the study. Treasury yields plunged the most in five years on Oct. 15, 2014, before recovering, fueling a months-long debate over whether something has fundamentally changed in a $12.7 trillion market that most investors consider a safe haven.

 

* * *

While Fed officials concede Dodd-Frank has probably had some effect on price swings, they’ve joined Lew in flagging high-frequency trading as an important factor.

 

The strategy typically involves using ultra-fast technology and placing computer servers close to exchanges to react to market data as quickly as possible. Such trading has drawn increased attention from regulators since the May 2010 flash crash, when $1 trillion of value was briefly erased from U.S. stocks.

 

“Their trading patterns are different,” Fed Governor Lael Brainard said last week at an event in Washington. “As they take a preponderance of the trading activity in some markets, that no doubt also may change the patterns of liquidity resilience.”

So back to the report which indeed tries to distance itself from HFT as the primary culprit of the yield crash at 9:34 am, by saying "While no single cause is apparent in the data, the analysis thus far does point to a number of findings which, in aggregate, help explain the conditions that likely contributed to the volatility."

And an amusing sideline: after back in 2009 Zero Hedge was one of the first to use the term High-Frequency Trader, or HFT, a monicker which since stuck and has become a popular pejorative to explain away all broken markets and microstructure manipulation, the report first and foremost does what every good report seeking to defend the object it is accusing, does: change the name. Sure enough, gone is "HFT", and instead the report uses "PTF" or "Principal trading firms" instead, which is clear enough: since HFTs trade for their own account, we are happy with this new name, which unfortunately for the Modern Markets Initiative lobby which surely spent a lot of money to get HFT changed to PTF, will not stick.

So any time readers encounter PTF, just think HFT.

What are the report's main findings:

• An analysis of transactions shows that, on average, the types of firms participating in trading on October 15 did so in similar proportions to other days in the sample data. Principal trading firms (PTFs) represented more than half of traded volume, followed by bank-dealers. Both bank-dealers and PTFs continued to transact during the event window, and the share of PTF trading increased significantly.
• The trading volume of PTFs and bank-dealers in the cash and futures markets is highly concentrated in the most active firms. In the cash market, for instance, the 10 most active PTFs conducted more than 90 percent of the trading activity of all PTFs on October 15, while the 10 most active bank-dealers accounted for nearly 80 percent of the trading activity of all banks. The concentration findings were generally similar for the futures market.
• A review of position changes shows sizable changes in net positions by different types of participants following the retail sales data release. However, during the event window, only modest changes in net positions occurred, suggesting that changes in global risk sentiment and associated investor positions may help to explain a portion of the price movements during the day, but do not appear to explain the round-trip in prices during the event window itself.
• During the event window, an imbalance between the volume of buyer-initiated trades and the volume of seller-initiated trades is observed, with more buyer-initiated trades as prices rise in the first part of the window, and more seller-initiated trades as prices fall in the second part of the event window. Such imbalances are common during periods of significant directional market moves. Both bank-dealers and PTFs initiate these liquidity-removing trades, though PTFs account for the largest share. At the same time, strong evidence suggests that PTFs, as a group, also remained engaged as liquidity providers throughout the event window, implying that more than one type of PTF strategy was at work.
• Several large transactions—though not unusual in size relative to other sample days— occurred between the retail sales release and the start of the event window. Some coincided with a significant reduction in market bid and offer depth—both during this interval and at the start of the event window itself. But during the event window, the analysis does not suggest a direct causal relationship between the volatility and one or more large transactions, orders, or substantial position change.
• The significant reduction in market depth following the retail sales data release appears to be the result of both the high volume of transactions and bank-dealers and PTFs changing their participation in the cash and futures order books. During the event window, bank-dealers tended to widen their bid-ask spreads, and for a period of time provided no, or very few, offers in the order book in the cash Treasury market. At the same time, PTFs tended to reduce the quantity of orders they supplied, and account for the largest share of the order book reduction, but maintained tight bid-ask spreads. Both sets of actions prompted the visible depth in the cash and futures order books to decline at the top price levels.
• The time required by the futures exchange to process incoming orders, or “latency,” increased just prior to the start of the event window. This latency was associated with a significant increase in message traffic—in this case elevated due to order cancellations. Transaction data also show a higher incidence of “self-trading” during the event window. For the purpose of this report, self-trading is defined as a transaction in which the same entity takes both sides of the trade so that no change in beneficial ownership results. Although self-trading represented a non-trivial portion of volume, this activity also appears on days other than October 15 in the sample. Any causal connection between the unusually high level of cancellations or the self-trading and the event window at this time remains unknown.

Which brings us to the report's conclusion:

In sum, record trade volumes, a decline in order book depth, changes in order flow and liquidity provision, and notable and unusual market activity together provide important insight into the factors that may have contributed to the heightened volatility, decreased liquidity, and round-trip in prices on October 15.

In short: a lots of words to show just one chart which is the only one that mattered.

What does this chart show? Recall from the top chart that the peak of the chaotic yield plunge (and price surge) took place at precisely 9:34 (and 3 second). Here is the official explanation of what the chart shows:

Figure 3.29 shows the message rate and latency build-up within a single second around 9:34 ET at millisecond resolution, illustrating how a peak message rate of around 40 messages per millisecond results in a gradual slowing down of the response time of the matching engine. Once the messaging rate fell, trading platform latency quickly returned to previous low levels. While the message cancellations observed very near the beginning of the event window were not a direct cause of price movements at the time given their distance from the market price, the associated latency would have affected the trading speeds of other market participants by increasing the time lag between initial order entry and possible execution on the platform. As some market participants monitor latency and include it as a variable in their trading strategies, sudden changes in latency would cause them to adjust their behavior.

Still unclear: here's more. "Given the finite capacity of any matching engine to simultaneously process messages and execute matches between buyers and sellers, extremely high message rates appeared to cause trading platform latency to temporarily jump higher."

[28]

This is what happened: a massive burst of quote stuffing (seen with absolute clarity on Figure 3.29 above) in the form of a surge in message, resulted in a burst of accumulated order latency, which in turn was the catalyst to send the price soaring from 129 to over 130 in the span of 5 minutes, and then sliding back down again once the quote stuffing effect was eliminated.

Seem familiar? This is precisely what we, in collaboration with Nanex, said was the reason for the Flash Crash of May 2010 - not some scapegoat Indian trading out of his parent's basement in a London suburb, but either a malicious, premeditated rogue algo or else a freak algorithm whose programmers lost all control over. For those who have forgotten, we urge re-reading: "How HFT Quote Stuffing Caused The Market Crash Of May 6, And Threatens To Destroy The Entire Market At Any Moment [29]."

Whichever case, it was quote stuffing that single-handedly destabilized the market, first in stocks in May 2010 and then in Treasurys in October 2014.

Confirming that it was all PTFs, pardon HFTs, is a chart showing the burst in volume by firm/order type which not surprisingly was all HFT...

[30]

... coupled with the collapse in market depth as HFT added tons of volume and eliminated all the market debt, also known as liquidity.

[31]

As it turns out HFTs not only don't add liquidity, they actively eliminate it! How ironic that this is precisely the opposite of what the HFT lobby claims it does. And this time it is not some tinfoil hat allegation: it is documented in a Fed, Treasury, SEC, CFTC paper.

Furthermore, it wasn't just quote-stuffing with order message blasts, it is also a surge in order cancelations and self-trading - all telltale signs of rogue HFT algos.

Analysis of transaction and order book data during the event window revealed two notable patterns in activity on October 15, high levels of cancellations and self-trading, but whether this activity contributed to the rapid price movements is unknown.

 

The cancellation activity witnessed in the invisible futures order book also resulted in a highly volatile total order book depth (including visible and invisible orders) in the futures market (Figure 3.30). In the futures market, the portion of the order book that is not visible to market participants (that portion that rests at levels outside the top 10 best bid and offer price levels) can represent anywhere from 50 to 90 percent of total market depth—witnessed both on October 15 and the control days.

Ok cancellations we have covered extensively in the pastt but self-trading, as in the HFT firms "buys" and "sells" from itself? Why yes:

A second notable aspect of trading on October 15 was the heightened level of self-trading during portions of the event window. Self-trading, for the purpose of this report, is defined as a transaction in which the same entity takes both sides of the trade so that no change in beneficial ownership results. Self-trades appeared in both cash and futures market data at varying levels across firms and time periods.28 In the cash market for 10-year Treasury securities, for example, self-trading represents 5.6 percent of the total activity on control days, and 4.2 percent on October 15 (Table 3.9).

How is this possible, or better, legal? Simple: this is precisely what HFT is all about - creating the illusion of activity to force out real orders and frontrun them.

The bulk of self-trading in cash and futures markets was observed among PTFs, perhaps due to the fact that such firms can run multiple separate trading algorithms simultaneously. For instance, one of these algorithms could specialize in placing buy or sell limit orders at the top of the order book while another could specialize in initiating trades given specific conditions in that market, potentially leading one algorithm to end up being matched with another algorithm from the same firm. In addition to PTFs, the cash data also showed a very small amount of selftrading by bank-dealers and hedge funds, some of which are also known to trade algorithmically.

Sure enough, in addition to everything else noted previously, self-trading was a dominant feature during the flash smash period:

During the event window, the data showed that the share of overall transactions resulting from self-trading was substantially higher than average. At the 10-year maturity, it reached 14.9 percent and 11.5 percent for cash and futures, respectively, during the move up in prices in the event window (Figure 3.31). During the retracement, when the price moved back down rapidly, the share of self-trading declined to 1.2 percent and 4.8 percent in cash and futures, respectively. Moreover, the concentration of self-trading volume among PTFs was very high in both markets during the event window. Another aspect of self-trading flows during the event window was its directional nature (Figures 3.32 and 3.33). For example, between 9:33 and 9:39 ET, the cumulative net aggressive buyer- minus seller-initiated self-trade volume increased by around $160 million in the cash 10-year note, accounting for close to one-fifth of the total imbalance between buyer and seller initiated trades observed over that time interval

So... say you want to sent the price of a security, whether cash or derivative, soaring, what do you do? Why you just buy and sell from yourself at a furious pace, lifting the bid ask ever higher to draw out any organic, hidden order flow. And one you have achieved your goal, or fail to draw out momentum piggy backers, or halt self-trading and the price tumbles back to pre-manipulation levels.

And there you have it: this is precisely how HFTs, pardon, PTFs, caused the bond market shock of October 15. But don't worry, according to both the report, the WSJ, and Irene Aldridge, there was not "one single cause."

* * *

All of this is known to our readers: we have covered virtually every aspect of fragmented, broken markets and price manipulation via HFT algorithms for over six years. We are delighted to see the biggest market regulators and supervisors admit, if tacitly, that our "conspiracy theories" have been right all along.

But what is surpising is that unlike the SEC's Flash Crash report which was a travesty and blamed the crash on Waddell and Reed, to be followed by another travesty of a report, one which has sent an innocent trader behind bars, this time HFT is explicitly, if not deliberately, singled out.

Which in our opintion sets the stage. The stage for what? Why blaming the upcoming market crash on HFTs, of course.  As Bloomberg commented, these findings "will probably add to regulatory scrutiny of the industry."

The reality is that regulators know very well what is really going on in the markets, and now that HFTs have been exposed as the catalyst for the bond market crash, when the inevitable stock market crash - a crash that will be the result far more of the ruinous decisions of central planners around the globe - it will be the HFTs, pardon, PTFs that will be the first to blame, while the central bankers do their best to quietly slip out to a non-extradition country.

Just look at China: the government is so terrified of losing control over its own stock market bubble and the potential for violent, social conflict that would result, that it will throw everything at the market to support it. In the US, the regulators are already one step ahead: they know a crash is inevitable, and the only thing they need is the scapegoat to blame it on when it all comes crashing down.

Nameless, faceless algos would be just the perfect scapegoat.

h/t Eric Hunsader

Source: Joint Staff Report: The U.S. Treasury Market on October 15, 2014 [18]