Trading Volatility Halts and the HFT 4% Boogie

halt

Yesterday morning just after the market’s open we noticed a series of trading halts – over 50 in all –that were triggered in the markets. All of these halts were triggered with a Reason Code of “M” – the volatility trading pause when securities experience a price change of over 10% within a 5-minute period. All of these halts occurred  in ETFs and ETNs, and fairly illiquid ones at that. And the halts were triggered by algorithmic quoting activity that did not result in trades. This is curious; we thought volatility trading pauses were triggered with trades, and not just quotes. Here is how NYSE ARCA describes how the trading pauses work:

If a security falls/rises 10% in a 5-minute time period a market-wide trading pause will occur in the security for a full 5-minutes. To determine the high and low volatility thresholds NYSE Arca will calculate the continuous 5-minute high and 5-minute low for each security on the pilot list.

1) The Low Volatility Threshold will equal 5-minute high minus 10%

2) The High Volatility Threshold will equal 5-minute low plus 10%

When a trade triggers a trading pause, NYSE Arca will send the indictor to the single plan processor. This will result in the trading halt reason code and quote condition code “M” disseminated by CT/CQ to all data feed recipients. Pauses will last at least 5-minutes and end with an auction on the primary market similar to those held at the open beginning and close of each trading day.

This is apparently not so. Looking at the activity in one of these ETFs – ONN, which is the Gartman “Risk-On” ETF, there were no trades that triggered the halt. In addition, the events seemed to be set in motion when two things happened:

1)      At precisely 09:45:01 the ETF appeared to lose all offers. There ceased being a two-side market – even a wide one. At 09:45:00 the bid-ask was 29.28 by 30.35, and then all of a sudden, poof – just bids.

2)      At precisely 09:45:15 “The 4% Algo” commenced to quote strangely on the CBSX. The algo started entering orders, and cancelling them in a laddered way. In a period of 5 seconds, this algo started quoting bids at $28.11, cancelling, and subsequently entering new bids 4% lower, until it quoted ONN at a bid of a tenth of a penny. Likewise it started quoting offers at $31.53, then cancelling and entering subsequent offers each 4% higher than its previous one until it quoted ONN at $7,069 offered. See the NANEX chart and page ahead to chart #36 to see the ski jump in detail.

ONN

Ok. So a whole mess of ETFs and ETNs were ravaged by some “4% algo” set in motion by some participant on the CBSX. We still do not know why this caused a trading pause when no trades took place. After all, we see this type of activity daily in thousands of stocks.

Is this activity quote stuffing? What is the intention of this “4% algo”, specifically? Is it nefarious? How does this contribute to market quality, tighter spreads, and more liquidity? Was it related to Limit Up Limit Down (LULD) implementation? How?

We hope the exchanges and regulators are looking at yesterday’s events closely. While these ETFs and ETNs are not liquid to begin with, and yesterday was a quiet August Monday, we do not like the sound of untested, not-understood, and possibly nefarious algos clogging the plumbing and impacting trading in any security.

What if this algo caused halts in SPY, IWM, or QQQ?  What if yesterday was a stressed high volume/ high volatility day? How would you all feel about trading be ceased in those instruments? What would have been the correlation spillover effect to the individual securities comprising those ETFs?

These halts were a no-harm no-foul freebie for the regulators; without investor harm they have the opportunity now to get a glimpse into how some algos think and act, and hopefully prevent something more impacttful down the road.