, , , , , , , , ,

A Long/Short scoring engine is at the core of our model.  I would love to call it proprietary, but it’s scoring a number of readily available ratios from free data and from that generating a unique score. Is this combination of data points unique? Maybe. But I doubt it.

While the model is replicable and other shops likely use similar ones, ours is focused on extremely liquid strategy. I don’t mind discussing the nuts and bolts of how it operates in the wild.

This was built around the Excel file made available by InvestExcel, a copy of the original file before I modified it is here.

Below you will find a snap shot of the Stock Comparison Scoring Engine. Using 40 different items to compare and contrast two like stocks, it generates a unique relative score of 1 or 0 per line compared. These are then combined to give us the relative score between any two stocks compared against each other.

Below is the score for Exxon & Chevron, as compared against each other using yesterday’s data.

The relative score will change in value as new information is available. The idea is to run the model daily and build a database of the rolling scores for each paired trade of interest.

The model will track the trend in the pairs, giving us insight into historical divergent and expected mean reversion estimates.  In addition, it generates the best timing of the pairs themselves.

While Alfred Jones, who founded the first hedge fund, had his organization track a form of volatility and weight their pairs based on a volatility-rated value, we are going to take a different approach. We will use equal dollar weighting on same sub-sector pairs.

I prefer to use what I think of as natural pairs. If they belong to the same business sector, and preferably sub-sectors, that makes a lot more sense than using cross-sector pairs. I don’t believe in the lowering of risk, just because historically if the volatility or rate of change between two stocks has been X, it will stay X.

The model uses a fundamental bias structure, so it looks to capture the change between two competitors.  We are only focused on the rate of change between two nearly identical stocks. This strategy is designed to be as uncorrelated with the index as possible.

While a relative paired score is nice to have, a baseline is needed so I’m working on adding that to the model. Below is a snap shot of the Sub-Sector Comparison Engine.

The next step will be to recode this to generate a sub-sector score for all stocks.

This helps us identify the best natural pairs in a sector. I will post on the Sub-Sector Comparison Engine, once I have it polished and generating sector-wide scores.

The model being discussed in these blog posts is an example model only. Clients of JHB Capital will see a new, and completely different set of paired trades in their own account’s once the model is in operation.

This is an example of a copy of the model only. This model has three intentionally constructed trades out of nine pairs that are designed to test the models outputs.

As such, do not invest in this model or its results blindly. You should consult  with your adviser before investing in anything you find on the internet for free. Everything in this post is for example only.

If you have any comments or thoughts about how I can improve on this design please join me in the comments section of this or the first article in the series.

Jack H Barnes Jr
CEO & Founder
JHB Capital LLC