What do the edges in the FX pair graph really mean?
In my original post on clique analysis in Brexit FX markets, I introduced the idea of looking for changes in “trader engagement” with currency pairs in the three week window around Brexit.
Here’s what the graph looks like when the presence of a single trade is enough to define an edge
Quicktime Movie: brexit-15d-all-messages
On the other hand, here’s what happens when a threshold of 4 trades is applied:
Quicktime Movie: brexit-t4-15d
Clearly it doesn’t take much to cause a change. On the other hand, if you take away too many edges you eliminate most of the cliques.
But my first question is this: What is the best way to define the presence of an edge so that the cliques show up clearly?
Also, I’m planning to make more data available at the 1QBit QDK site, in a form that allows it to be easily imported into the iPython interactive notebooks. So any thoughts on this will also be appreciated.
A small follow up: On my Macbook, I have to download the QuickTime movies and play them locally, but they come up in separate players, so I can look at specific frames by using the slider bar. The main difference between the “no threshold” movie and the “4 trade threshold” movie is that there were only a few USD-GBP trades, and this made the GBP look more connected than it really was.
The numerical data for the blog post shows this. However, I’m also looking for ways to incorporate things like this into the visualization, so if you have viewed the movies and have some ideas, please let me know.
This is from the Eris Futures website. I’m posting it here as an example of how Brexit showed up in other types of market data. The bid/ask spread on the order book can be read as an expression of intent. The reported trades used in our clique analysis are an expression of how intent was acted on. Note how the bid/ask spread narrowed before volume started to recover in the order book.