This is kind of a long shot, and quite an experimental indicator. I don’t know of any similar indicator to what I am going to suggest in the retail market. HotspotFX have a similar-ish function on their platform, but its used by institutional dealers for sourcing liquidity.
What I am going to suggest is another visual way of looking at market depth. It will use FX data from the CME suite and hopefully be able to incorporate other data for “market depth aggregation” purposes (I.e from several different exchanges; if they make the data freely available, that is)
Here is a screenshot from the data suite at CME. (http://datasuite.cmegroup.com/EFXweb/index.html)
You will notice that the totals of the bid and offer ladders are added together at the bottom, for a range of prices. For example, in this case we have 325 total bids and 375 total offers in that particular moment. Its widely accepted that if the bid side is significantly higher than the offer side at any particular time, then there is a chance that the market will go higher if those bids are filled. That is the extremely blunt end of the theory, but in practice its not that straight forward, obviously. What we need is a representation of the structure of bids and offers over time x, in this case, x being significantly longer than one tick!. Just because we see the bid side is higher than the offer one time, doesn’t mean we will see a run at the offers in the near future. Instead we need to analyse the build-up of bids over a particular time interval. We need to look at the distribution of bids to offers and hopefully observe a significance in the bid-offer structure.
Here is a decent suggestion; we start with 5 rows. In each row we want to determine whether the bid stack is higher or lower than the offer stack in each tick…and by how much. Is the bid greater than 1/3 the size of the offer, greater than 2 times the offer etc? For example take this data set;
(bid offer) ticks from 9:00am-9:05am
130 30
145 50
155 67
125 167
134 145
155 161
197 145
198 146
201 149
151 203
319 181
330 160
Then use;
(if bid>offer then ”b>o”= , if offer>bid then ”b>o”= )
(if (bid)>(offer*0.33+offer) then “b>0(1/3)”=) if not, then leave blank.
(if (offer)>(bid*0.33+bid) then “b>0(1/3)”=) if not, then leave blank.
(if (bid)>(offer*0.75+offer) then “b>0(0.75)”=) etc, etc.
So if we plot them it might look something like this.
Then a summary perhaps, for example;
Time 9:00am-9:15am
b>o 51 21
b>o (1/3) 40 9
b>o (0.75) 25 4
As you can see, it’s a bit like the old Windows defragmenter! I suppose one of the problems would be scraping the data from the CME site and/or other exchanges. I have tried to look for a method to get this live data into Matlab, but its beyond my capabilities at this time. What do you guys think? Fancy giving this a shot? Any feedback would be appreciated.
All the best, Aaron.