High Frequency Traders Target Gold Miner ETF

GDX continues to exhibit curious trading behavior.

Reviewed by: Dave Nadig
Edited by: Dave Nadig

GDX continues to exhibit curious trading behavior.

I wrote an article about the Market Vectors Gold Miners ETF (GDX | B-68) in October, highlighting some odd trading on a very good day for the fund. In it, I suggested GDX was getting pushed around a bit by the large amount of shares held in the Direxion Daily Gold Miners Bull 3x (NUGT) and Direxion Daily Gold Miners Bear 3X (DUST) funds. I love writing things like that, because it drew the lines between how the ETF structure works and how the market works, and tied it up all in a nice little bow.

However, since October, the trading in GDX has stopped making all that much sense, and Tuesday was the icing on the cake. Here’s the tick chart:


This, folks, is what we call a “mini flash crash.” GDX was happily trading along all day, up just a little bit on the day, and in the last two seconds of trading—from 3:58:58 to 4:00:00—millions of shares changed hands while the price dipped as low as $17.72, before finally recovering to a rational price of $19.62.

What’s Going On Here

So what the heck is happening? Well, I have a few theories. The first thing I generally always look at is whether a fund subject to weird trading was representing the fair value of its holdings well.


Attend Inside ETFs, the World's Largest ETF Conference!


In this case, the answer is “sort of.” For most of the last two hours of trading, it was trading 10 to 15 cents over fair value (as represented but the Gold Miners Index, which it tracks, labeled GMI on the chart). That’s not a huge amount—we’re talking about a premium that never approached 0.50 percent—but it’s potentially enough that market makers could step in and book a profit by selling shares of GDX at a premium, and buying up all the underlying stocks. But it’s just not a big enough gap to pass the smell test.

Rebalancing The Culprit?

My next question would be if this was based on NUGT rebalancing. I just can’t see that. Tuesday’s 13.4 million shares traded (a value of around $190 million) is actually a light day for NUGT. It was trading 28 million a day at times in November. By contrast, GDX traded 36.4 million shares, at a value of around $700 million. I just don’t see the tail wagging the dog this time.

The next think to look at is the tape: Why would someone be selling on the close? A rational explanation would be cleaning up the book at the end of the day.

Most day traders and high-frequency shops don’t particularly want to “wear” their positions. They would rather go home to their homes in New Jersey sitting on cash, not having to worry about whether gold or Japan blows up between 4 p.m. and tomorrow at 9:30 am. So if I’ve been sitting on a gain from positions all day in GDX, why not book it by selling right before the close?

That’s when I looked at what happened on the tape:


The important thing to look at is the “Cond” column on the right side. That IS stands for “intermarket sweep,” the order type almost exclusively used by high-frequency traders (as if the tens of thousands of quotes and trades pushing through at 15:59:58 weren’t clue enough).

Here’s how an IS order (ISO) works.

Imagine there are only two places to trade: Bob Exchange and Alice Exchange. Bob has someone saying she’ll buy 100 shares at $10, and someone under her saying he’ll buy 100 shares at $9. Alice has someone saying he’ll buy 200 shares at $10, and 100 shares at $9.50.


Order Protection

Now, if I put in a regular old limit order to just sell 200 shares at better than $9, my order is protected. That means that I have to get the best price the whole market has to offer. My broker’s routing mechanism will either sell them to Alice’s guy at $10, or split the order between Bob and Alice, and I’ll still get $10. In fact, the order protection rule means there’s no way—even if I submit the order directly to Bob—that I will get worse than $10, because Alice has all that liquidity available on her side.

However, if I walk in with that same order to Bob and flag it as an IS, my order is no longer protected. It will just walk through all the books as it sees them. So I’ll sell some to Bob’s buyer at $10, and then the next 100 to Bob’s buyer at $9.

Technically, I’m sending out my order to everyone at the same time, but because I submitted it to Bob first, in the modern world of HFT, by the time Alice even knows I’m a seller, I’ve already gotten filled by Bob’s buyers.

Taking Advantage Of Liquidity

Why would I do that? To take advantage of all the liquidity I can see on Bob’s exchange on the bid side, all at once, before anyone can move the price or remove their bids. That is in fact exactly why it exists. From the SEC: “The Intermarket Sweep exception enables trading centers that receive sweep orders to execute those orders immediately, without waiting for better-priced quotations in other markets to be updated.”

And why might you want to do that? Well, the “logical” explanation is that you want to get your execution done very, very quickly. Or possibly that you believe your order is so big that if you just divvied everything up across the “best” prices, by the time you finished your order, you’d actually be worse off than just sweeping all of the liquidity out of one market center.

High-frequency traders almost exclusively use ISOs because the process of having the exchange check for better prices everywhere else would simply take much too long. They want to execute against the liquidity they actually see on a given venue.

This unintended use of ISOs has been the subject of a lot of scrutiny (and a great paper by BlackRock regarding the flash crash, with a follow-up from Manchester Business School on mini flash crashes).


Market Manipulation?

So why do HFT algorithms really do this? Well, obviously they want speed and assured execution at visible prices. That’s kind of what HFT is. But could it be an attempt at price manipulation?

Theoretically, if they sweep a book at a smaller venue and manage to print a trade that’s way off the market, they’re hoping they can spook the other exchanges and other algos where there might be additional liquidity.

That could be what’s at work here. Looking at the tape, we see a walk-down of fairly large ISOs, back to back, running the price all the way down from $19.25 to $19.03.

That’s when all hell broke loose and thousands of small trades start flying. Along the way, we start seeing more significant intermarket sweeps, all the way down the day’s lows:


What I find interesting here is some of the order pattern. The “spook” seemed to start with sweeps targeting NYSE, then spread to the smaller exchanges (K and J codes represent New Jersey-based Direct Edge, one of the smaller exchanges favored by some HFT shops).

These aren’t enormous trades—they’re all in and around $100,000 in value apiece. But they’re big enough chunks in aggregate that someone had a good day. Remember, every one of these trades has a buyer and a seller, and whoever’s algo was foolish enough to be selling $500,000 of GDX at $17.72 on Direct Edge is likely getting pistol-whipped by their boss today.


GDX End-Of-Day Trading

Trading in GDX at the end of the day has been steadily on the rise—especially since so much focus has been on NUGT since October. It’s gotten to the point where the vast majority of trading is happening in the minutes or seconds before the close, and most of it clearly coming from HFT shops. We also saw a rundown—possibly a second or two earlier—in NUGT as well. So there are really two options here:

  • A very smart algorithm is effectively spooking prices and then buying cheap.
  • A collection of very dumb algorithms is making terrible trades and selling cheap.

That’s the thing about these kind of forensics, it’s inevitably just guesswork.

The good news is that, as a rational, human investor, you can mostly ignore these kinds of shenanigans. GDX is a great fund, quietly doing what it’s supposed to do, day after day.

That it’s become the target of some predatory computer programs isn’t actually going to affect you unless you start trading with market orders, or worse, leave sleeping stops in your account to sell or buy based on triggers that seem far from today’s prices.

In other words, as long as you focus on good trading hygiene, these kinds of problems will continue to be somebody else’s.



At the time this article was written, the author held no positions in the securities mentioned. You can reach Dave Nadig at [email protected], or on Twitter @DaveNadig.

Prior to becoming chief investment officer and director of research at ETF Trends, Dave Nadig was managing director of etf.com. Previously, he was director of ETFs at FactSet Research Systems. Before that, as managing director at BGI, Nadig helped design some of the first ETFs. As co-founder of Cerulli Associates, he conducted some of the earliest research on fee-only financial advisors and the rise of indexing.