Apr

23

Some interesting stats from Easley, D., M. Lopez de Prado and M. O. Hara (2012b): Bulk Volume Classification, Working paper about E-mini trade sizes:


Most of the[..] trades are small, averaging 4.50 contracts per reported fill. Figure 1 plots the frequency of trades per trade size. […not shown…] The frequency line quickly decays as a function of the trade size, with the exception of round trade sizes (5, 10, 20, 25, 50, 100, 200, etc.).

That round trade sizes are much more common than their neighbors may be attributed to so-called ‘mouse’ or ‘GUI’ traders, i.e. human traders that send orders by clicking buttons on a GUI (Graphical User Interface). As an interesting aside, this footprint of ‘GUI traders’ could be used by machines to learn the patterns of their human competitors, and eventually anticipate them to the advantage of the ‘silicon traders’. For example, size 10 is 2.9 times more frequent than size 9. Size 50 is 10.86 times more likely than size 49. Size 100 is 16.78 times more frequent than size 99. Size 200 is 27.18 times more likely than size 199. Size 250 is 32.5 times more frequent than size 249, and size 500 is 57.06 times more frequent than size 499. Such patterns are not typical of ‘silicon traders’ who usually randomize trades to disguise their footprint in markets.


Comments

Name

Email

Website

Speak your mind

1 Comment so far

  1. Andrew Goodwin on April 23, 2012 6:40 pm

    If one uses stock execution algos provided by the banks/brokers then one will find that one achieves fills in odd lot sizes. One reason for these odd lot fills is the pinging of even lots by HFT. They are testing to see how much size is out there and dark.

    Normally if I’m watching my fills on a broker provided algo and I get an odd lot fill, I know I’ve been pinged. Then I change the algo so that they can’t get in front of me for a long time.

    Mathematics related to the fill frequencies of Odd Lots have lost their predictive ability and work in reverse according to some recent work. The Odd-Lot oscillators now work in reverse function for sentiment readings. Odd lot buying is bullish and odd lot selling is bearish, at least when reversed from extremes.

Archives

Resources & Links

Search