We built a proprietary natural language processing engine which parses social media data looking for the specific "attitudes" bullishness and bearishness.
Ahhh, so proprietary sauce! I don't see any indication of when you have made changes to your engine or grammar in the graphs. This is a serious distortion from one perspective.
The components and weighting of the VIX are fixed and known. But if your metric is secret, and you make ongoing changes to it, the viewer can not know when changes were made and their impact.
One can re-play historical stock data and feed it to new algorithms without a distortion. But unless you are saving all the raw social media data and re-applying your filters to this historical data, your graphs are distorted.
Also, compared to the VIX, the volatility of the components of social media sentiment is quite high.
I am sure your system is useful and profitable, even if my criticism is accurate.
I should be more clear because there is a crush of data on the page. Look at the "10 Days Correlations Chart" de-select all but the Bull/VIX. Sorry, still working on the UI.
6 comments
[ 3.9 ms ] story [ 21.0 ms ] threadOne can re-play historical stock data and feed it to new algorithms without a distortion. But unless you are saving all the raw social media data and re-applying your filters to this historical data, your graphs are distorted.
Also, compared to the VIX, the volatility of the components of social media sentiment is quite high.
I am sure your system is useful and profitable, even if my criticism is accurate.