1
$\begingroup$

I’ve heard that a lot of HFT use so called micro price for making predictions for futures and other product. Basically it convert the LOB and order message to a single number.

I know it is kind of secret for each trader and might not be easy to get start on my own, but I wonder if there is any publication that could give me a head start, especially on pricing a contract with its own books and some related contracts books.

I am aware someone in the academic use micro price as a term for a very simple model, which only consider the top of book price and quantity. I am looking for something that would incorporate multi level of book, and execution momentum as well.

Thanks

Mat

$\endgroup$
1
$\begingroup$

First approximation: use weighted mid using bbo quantities. Next approximation: use multi-level weighted bid price and weighted ask price and take the mean of that. You can test a bunch of variations around this.

| improve this answer | |
$\endgroup$
  • $\begingroup$ yes I am aware of it, do you know any publication talk about this in a multiple product setting? Thanks $\endgroup$ – Inthematrix Dec 20 '19 at 6:09
  • $\begingroup$ By "multi-product" do you mean multiple inter-related products that move together, like ES, SPY, VOO? $\endgroup$ – Alex C Dec 20 '19 at 17:09
  • $\begingroup$ I would say something related (i.e. ES and ND), rather than something having the same underlying like (ES and SPY) $\endgroup$ – Inthematrix Dec 21 '19 at 10:56
0
$\begingroup$

As a start I would VWAP the book (not to be confused with the execution algorithm benchmark), as in multiply all the book prices by all the book volumes and divide to come up with a "volume aware" price. When there is a bias towards the buy side of the book, the VWAP will be higher than the mid price and you can use it as a signal to measure order imbalance.

Problem is that a spoofer can very easily bias the book with no chance of actually executing their volume by dumping leveraged volume on one side far away from the mid price and forcing the value much higher than is rational and delivering a false positive signal. Trade on that, and you may end up buying or selling in anticipation of a big move only to find the volume go away and the spoofer happy to take the volume off your hands at a discount. An example is documented on this wonderful High Frequency Trading blog! .

The answer may be to make a regularized average that eliminates some highly suspect volume off the extremes and not incorporate that into the metric. Find some examples of that spoofing before getting more sophisticated though, simpler is always better.

| improve this answer | |
$\endgroup$
  • $\begingroup$ I understand what you mean by spoofer problem but I don't think that is of great concern. I can always discount the book level by their distance to the mid price. A deeper level which is a few ticks from the mid will receive a heavy discount and thus will not contribute to pricing significantly. A spoofer in that case would have to put their order on top level if they wanna affect the price, which incurs significant directional risk for them. What I am looking for is some publication, that talk about how to do this in terms of 2 or more related products. $\endgroup$ – Inthematrix Dec 20 '19 at 6:06
  • $\begingroup$ I am not looking for the exact algorithm, I would like to see if there are any attempts on pricing off futures with 2 or more related product, with account of book and deal momentum. I've seen some academic try to use the word microprice (wwwf.imperial.ac.uk/~ajacquie/Gatheral60/Slides/…), but to be honest this kind of 1st level book algorithm is too simple and rarely be useful in real trading. $\endgroup$ – Inthematrix Dec 20 '19 at 6:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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