High Frequency Trading "Quote Stuffing" Visualized | Hacker News Reader

High Frequency Trading "Quote Stuffing" Visualized

zerohedge.com

41 pointsuptown12 comments

Pretty much everyone doing "real" trades will look at things like VWAP (volume weighted average price) rather than just the pure bid/ask quotes because everyone knows bid/ask might not reflect the true market value. Indeed in many circumstances you may well be required to benchmark your trades against something like the VWAP to prove you're acting in the best interests of your client (getting good deals rather than pushing deals to your favourite brokers who give you the largest kickbacks).

Some exchanges do in-fact prohibit such actions by having fill ratio requirements (a certain percentage of your quotes have to convert into done deals) or minimum quote times (your quote has to be dealable for at least x milliseconds before you cancel it) and just kick off anyone who breaches them.

Interesting article, and some really nifty graphs, but stay away from the comments section.

More seriously, could someone with a bit of experience in this sort of thing speak to the actual import of these patterns? It's not entirely obvious whether or not there are shenanigans at work.

Actually I thought the comments were the best part. ("It might turn out like the early experiments in making meth, it works until it blows the shed all to hell. These guys are playing in real markets. Imagine what mayhem they are doing in closed-system synthetic market models. What appears like the graphs above are what escapes from the lab!")

Well, the comments are extremely mixed. There are often a few good comments, but the majority varies from unnecessary to complete crap. Zerohedge has a lot of decent articles, but they are all gloomy, which attracts a lot of conspiracy theorists. Conversely, they also have some articles that could be considered fringe/crackpot. And the majority of commenters seems to be in the last group.

I remember listening to a bunch of kooky-sounding talk in 2005-2006 about asset bubbles, inverted yield curves, credit default swaps, mark-to-market accounting, and other arcana spoken of in incredulous, worrisome tones that sounded like rejected dialogue from The X Files.

Then, pretty much everything they said was going to happen, happened, except for the usual wish-fulfillment subtexts about military coups, riots, and TEOTWAWKI.

(Of course, I've heard the same sort of talk since the mid-1970s, so its predictive power isn't all that impressive...)

In my opinion, this is almost certainly not shenanigans, but simply algorithms behaving in ways their creators did not expect. I believe this for several reasons:

1) If you wanted to deliberately manipulate the markets, you'd come up with something less obvious than this. You might even come up with a pattern that is non-periodic or even put a few calls to rand() into your code just to spice things up.

2) If you were doing anything in the markets on purpose, you'd try to be unpredictable. And by "unpredictable", I mean "hard for a human to look at a graph and guess at". Once you trade predictably, the other guy can exploit you.

3) I've come up with trading strategies myself that generate "crop circles" in simulation. It's never on purpose, and I'd certainly never be permitted to put such things into production.

Sometimes algorithms behave in ways you didn't expect. For instance, you might compute a quantity f(t), place a buy order when f(t) > 0.5 and cancel it when f(t) < 0.5. If f(t) oscillates wildly from 0.4995 to 0.5001, you will wind up doing "quote stuffing" yourself. Of course, most of the time f(t) winds up oscillating from 0.631 to 0.632 or something like that, and things work normally.

If your system is designed well, some monitoring system further down the line will prevent this. Not all systems are well designed.

  but simply algorithms behaving in ways their creators did not expect

That's possible, but if the owner of the algorithm doesn't monitor it or shut it down, someone should and the owner should be punished for the behavior of the algorithm. Currently, that doesn't seem to be happening, which means this becomes a pretty good excuse for letting your algorithm behave like this on purpose.

The owner should be punished? What for? Who is harmed by a visible "crop circle" in the trade data? Also, I don't understand why you think the owners would allow their algorithms to behave like this; a "crop circle" doesn't make money and it does expose you to risk.

Incidentally, if your algorithm spams the exchanges too much (i.e., your fill rate goes way below 1%), your broker will shut you down. So while there may be crop circles out there, no one person is responsible for more than a few of them.

Because of the first paragraph of the article:

  [..] quote stuffing by HFT algorithms in tens of stocks, in which thousands of
  cancelled quotes would reappear each second with a definitive periodicity
  and regularity, [..]. Aside from the fact that it is illegal to indicate a 
  quote without a trade intent, this form of quote stuffing is in fact 
  manipulative when conducted by HFT repeaters in specific "shapes" as it
  actually moves the NBBO actively higher or lower, in cases pushing the
  bid/offer range up to 10% higher without even one trade ever having
  occurred, [..]

Incidentally, if your algorithm spams the exchanges too much (i.e., your fill rate goes way below 1%), your broker will shut you down

Except that that doesn't happen if you essentially are the broker, as is the case with large enough institutions. JP Morgan's broker isn't going to shut them down.

All quotes on an exchange have trade intent, since if the exchange fills you, a trade occurs. You don't have the ability to back out after your quote is accepted.

If the NBBO moves around on a thinly traded stock and no trade occurs, so what?The bid ask spread on some thinly traded stock fluctuated for a few minutes, and no one bought or sold stocks? Oh noes, the horror! If we don't put a stop to this, someone might offer to sell a coke bottle on ebay for $10000 and no one will buy it!

Also, if you are a broker, the exchange imposes similar requirements on fill rates. The matching engine gains nothing if you place lots of unfilled orders, and if they allow too much of it, customers who do make trades (i.e., the people who actually pay them) are likely to switch to other matching engines or dark pools.

  You don't have the ability to back out after your quote is accepted.

But you have the ability to back out on a majority of your quotes, if you submit so many that they couldn't possibly be fulfilled in the time before you cancel them. Despite the fact that they are cancelled, they still have their effect. http://www.zerohedge.com/article/how-hft-quote-stuffing-caus...

  Also, if you are a broker, the exchange imposes similar requirements on fill
  rates

Then why aren't these algorithms being banned?

It also might be a self-organized system composed of multiple algos from different shops: i.e. like a "Life" game.

For example: one algo put a bid of X, a nanosecond later another algo steps and put a bid X+1. The first algo cancel X and put X+2, etc. Until it reaches a threshold of X+T, then all bots cancel and start again from X.

You can see this behavior with human traders in after-hours sessions or in other illiquid markets. But of course in case of humans we are not talking about nanoseconds ;)

Reply on news.ycombinator.com