Miranda is an advanced and sophisticated machine learning model that discovers actionable credit market trade opportunities that have a high probability of mean reversion.

Miranda is demanding and abuses data so don't try to construct her on a typical desktop computer using an Excel spreadsheet.

The result will be a system overload, leading to a few hours of processing (represented by a blue circle) and potential system crash.

​

An example of what Miranda does to fixed income data sets:
She will digest 2,500 bonds (and all of the data associated with those bonds) and generate 3 million possible relative value pair trade ideas.  

 

Do you know how many columns and rows are in your excel spreadsheet?

Before you ask ChatGPT or BARD let me tell you......

 

1,048,576 rows

16,384 columns

btw that comes out to be 17,179,869,184 cells

How Does She Do It

Apache Beam allows her to create her magic on the Vertex AI service on the Google Cloud Platform. 

Miranda knows several proprietary algorithms that help her read, ingest, clean and prepare price, price history and reference data from a given source. 

​

Once the data is prepared, she fires idea generate algorithm that looks at EVERY bond and pairs them together.

After pairing up all of the bonds, she is ready to analyze each possible bond pair in a given market.

​

Miranda's natural love for pattern recognition helps her identify the bond pairs with the highest correlations and dislocations (measured by the Zscore of the pair spread difference).

​

Her ability to forecast helps with her predicting the pairs with dislocations that will revert back to normal.