Quantitative Finance Stack Exchange is a question and answer site for finance professionals and academics. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In reading about the various practices and strategies of high frequency traders, one of the most mysterious to me is "fleeting orders," or orders that are cancelled almost immediately after they are sent (see Hasbrouck and Saar (2011)). Why do HFTs use these orders? Some of those trying to explain the practice claim it gives HFTs an informational advantage. How? What information do they get and how do they use it?


I am hoping for someone with actual experience in HFT to answer and verify some of the hypotheses laid out by academics. In particular, one hypothesis is that rapidly cancelled limit orders bring forth market orders on the other side, which are then executed at a less favorable price also placed by the HFT. In other words, HFTs are gaming the system to exploit sub-optimal behavior by slower traders. Is there any evidence of this still occurring, now that the practice of fleeting orders is widespread and well known? This explanation would also imply that it is relatively simple to avoid being taken advantage of, thus weakening the policy implications. Is there some reason that slower traders prefer sending market orders only after seeing a limit order meeting their price target?

share|improve this question
-1 Flash orders are not "orders that are cancelled almost immediately after they are sent". See this diagram. Also, all the major US equities exchanges have voluntarily stopped flash orders [ 1, 2 ], though the SEC has long considered banning the practice themselves. – chrisaycock Aug 15 '11 at 14:41
@chrisaycock I have corrected the term to "fleeting orders". – Tal Fishman Aug 15 '11 at 14:53
I've undone my down vote. The paper you cite, however, should already answer your question. – chrisaycock Aug 15 '11 at 15:04
@chrisaycock well, sort of. I added an update to clarify what still mystifies me even after reading the paper (or at least the intro to the paper). – Tal Fishman Aug 15 '11 at 15:14
Emanuel Derman comes out in favor of "throwing sand in the gears" by instituting either a tax or (my favored solution) a delay in the processing of financial transactions. This effort, broadly speaking, appears to be gaining momentum. – Tal Fishman Aug 19 '11 at 17:34
up vote 14 down vote accepted

All HFTs are event driven. In the most basic sense, they have some model that is a function of order book events. For every order book event the model calculates some micro price that is the HFTs perceived fair value. This is often a function of the current bid, ask, depth, last n trade prices, inventory, etc. Given the most up to date view of fair value, the HFT will be adjust orders in the market.

As you can imagine, the rate of events for order book data is very high. This results in a very high scratch rate (cancel to fill ratio) because the HFT is adjusting their orders at a rate similar (but not equal to) the event arrival rate for a given stock.

The likelihood that there are shenanigans going on is non-zero. However, the use of said shenanigans is not wide spread, IMO.

share|improve this answer
Hi Louis, Welcome to quant.SE and thanks for answering my question. Your answer makes it sound like HFTs have "plausible deniability," but is the arrival rate of information so high that so many orders are legitimately cancelled after a tiny fraction of a second? Also, please don't be offended, perhaps only a minority of HFTs engage in shenanigans, but as a low frequency trader it is important to know to what extent my orders are being gamed (even if only by a large but vocal minority). – Tal Fishman Sep 8 '11 at 12:21
Thanks for the kind welcome! I'm not sure what you're implying they might be denying. By your logic, if the HFT is moving so fast it is unlikely that "slower" traders will be responding to all of the noise the HFT is creating. To re-emphasize, I believe that this notion that a high scratch rate indicates somtehing nefarious is going on is incorrect. – Louis Marascio Sep 8 '11 at 12:26
To be clear, I'm sure there are people and systems out there that try to move the market by displaying size very quickly and then removing it. This behavior is pretty easy to detect and has a very different signature from, say, a market making model adjusting its quotes. The Trillium case is a good example: reuters.com/article/2010/09/13/…. This has been going on for ages from when traders would bluff in the pits to now. Doing it intentionally is wrong and illegal, and hopefully the SEC will continue to crack down. – Louis Marascio Sep 8 '11 at 12:30

My favorite culprit is quote stuffing, which can be used for a lot of things, including mapping the topology of the exchange servers themselves. The general idea is to look for bottlenecks which can then be lagged with more targeted quote-stuffing to create arb opportunities.

Nanex's flash crash analysis covers this to some extent: http://www.nanex.net/20100506/FlashCrashAnalysis_Intro.html

The only thing I've heard or thought of to deal with this arms race is either built-in execution delays or going back to session-based call auctions rather than continuous. In either case, a 1 second delay or a 1 second call auction session would level the playing field in terms of speed of light (and might reduce data center rates in Brooklyn). ;-)

share|improve this answer
+1 how lagging by targeted quote-stuffing create arb to opportunities ? and where could I read about "mapping the topology of the exchange servers" by quote-staffing ? this is very interesting – Qbik Apr 26 '12 at 12:12

Based on my experience trading DJ Eurostoxx 50 (FESX), the HFTs place and delete the orders almost immediately to fool other algorithms which are based on limit order quantity at various prices. I have used a technique to scalp of a few ticks when such a phenomena occurs.

For instance if the HFt places an order for 1000 lots at 2000 and immediately it disappears, but the market price moves up and down by a few ticks. This is because other algorithms are designed to operate on certain order imbalance etc. Hence if I see a huge order on bid side, I used to sell some lots and immediately scalp off a a couple ticks. I used to do this and make some money during such hours. This stuff mostly happened during dull hours. My 2 cents.

share|improve this answer
Were you immediately selling at the offer, or at the faked bid? And then would you buy back at a lower bid? How did you get a place in the order queue to guarantee execution like this? – chrisaycock Feb 25 '13 at 12:13
No I was placing limit orders to get good execution. The fake bid always appeared 1 tick lower.For instance if some 20-30 lots were at 2001, suddenly some one would place 1000 lots at 2000 or 1999. I would place a sell order at 2002. And immediately place a buy order at 2000. The fake order would disappear soon and my order would get executed. This was only during dull hours. The quantity was less and the sudden appearance of the 1000 lot used to create some movement. But I wouldn't recommend it, as a couple of times, I was caught unaware and suffered some real loss trying to chase pennies:-) – Aravind Feb 26 '13 at 11:43

For example, high frequency market makers use all the market microstructure to take decisions of how many stocks put in each side. This microstructure can be a 5 level data (ten prices, 5 bids and 5 asks with the quote amount or quantity) so they calculate and quantify the order book imbalance or other type of measure of the supply/demand of the order book.

To modify the market they post some limit orders and send orders in different prices to try to balance the book order and also because they use this type of orders for the risk management of the strategy.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.