Blogs / Analysis Blog

Beta Dispersion in Joyride

header_beta short

A few months ago, we launched a trial version of Palantir Finance available to the general public.  We call it Joyride.  If you haven’t tried it yet, you can access it at  For this particular blog post, we’ve created it entirely using the data available on Joyride (courtesy of Xignite), so we encourage you to sign up for Joyride and try out this study for yourself.

The study will compare the relative performance of high-beta and low-beta stocks adjusted for market risk.  Our goal is to create two separate indices with equivalent beta adjusted exposures — one with the low-beta stocks and one with the high-beta stocks. We cannot directly compare the low versus high beta stocks without adjusting for the indices’ average beta to the market. [1] When we are finished, we will generate a plot with both indices on the same graph. The plot will reveal an interesting phenomenon that occurred prior to the major crash in 2008.

Open Joyride

To begin, please open a browser window and point it to From there, you can run Joyride. If you haven’t done so already, create a username and password (this only takes a few seconds).   In order to understand the features used in this study, I highly recommend that you watch our video tour.  (You will need to create an account to start the tour.)  It takes about 10 minutes and provides an overview of how to analyze data in Palantir Finance.

Create Your Group

Once you have logged into the program, click on the Explorer icon (1).  This will bring up the Instrument Explorer application.  From here, we will make two baskets of stocks, one with high betas and one with low betas.   First, select S&P 500 from the “Start with” pull-down menu.  Next, click on the percentile button: 2.  At the bottom of the graph, click on the expression bar (labeled ‘Bucketing Metric’) and enter the following text: beta(SPX, 252).  By doing so, you are calculating a beta over the past 252 days for each stock in the S&P 500 Index and displaying it on a histogram.  Now select the bottom 20% of stocks with the mouse.  Your screen should look like this:


To save this group, click on the disk icon in the upper left of the screen (4) and click the ‘Save’ button.  Save your group as something descriptive, such as “Low Beta Stocks.”   So that you can more easily follow this exercise, I have saved the key groups, indices, and graphs related to this blog post and shared them with all Joyride users (my username is iancoe2) using Palantir’s document control settings.  These settings easily allow users to share work across networks and, in the case of this study, allow you to check your work.   You can access these workflows at any time in the same area where you save your results.   Palantir also has the ability to share documents via the web.  You can click here to open my saved Instrument Explorer Group from above.  (You must have username to login.  Go to the Joyride home page to create one if you don’t have one.)

Make Your Group a Dynamic Index

From here, we are going to build a dynamic index that automatically rebalances each month to contain the top 20% (high beta stocks) and bottom 20% (low beta stocks).   Open the Index tool by clicking on 5.  For those of you more familiar with the product, feel free to use the shortcut on the Instrument Explorer’s Export tab.

On the right side, you will find a menu that allows you to import an instrument group or ex6plorer group. Click on that menu item and begin typing the name of your low beta instrument group into the area that says “specify group.”

You should now see your index.  Choose monthly rebalancing.  Your screen should look like the following picture (make sure your axes are the same as mine to compare).   Once you are satisfied with your index, save it as “LowBetaIndex.”


Now make an index of the average beta of the group during this time period by clicking on “Metric value of instruments” in the lower right hand corner.  We will use this index to adjust beta exposure when plotting the final result. Enter the text: beta(SPX, 252) and choose weighted average intead of sum.   Your screen should now look like mine. You can see that the average beta of the low beta group fluctuated between 0.53 and 0.68 during this time period.  Save your index as “LowBetaStockAvgBeta.”


To summarize our progress so far, we created a group of the lowest beta stocks.  We then brought those into the Index application, which created a dynamic index that automatically rebalanced every month to contain the bottom 20% of  S&P stocks as sorted by beta.  We then created a second index that was the weighted average of the beta of our index at each point in time.   Go ahead and repeat this process for the high beta stocks and you’ll be ready to move to the final step.

Plot the Results

We now want to compare our two indices to see which had the better performance given the market risk.  To do this, we will simply plot them in the Chart application 9.  However, we can’t just compare the two indices, because they have very different beta-adjusted exposures.   The low beta index has much less market exposure than the high beta index, i.e., in an up market, we would expect the high beta index to greatly outperform the low beta stock and vice versa in a down market.   However, if we normalize the indices to have the same beta-adjusted exposure, we should expect them to perform roughly the same, except for key periods when one or both had alpha relative to the benchmark (the S&P).   In order to weight these indices appropriately, we will take advantage of the expression bar’s mathematical capabilities.

To create the normalized beta exposure graph from the high beta index, enter the expression: Index(“HighBetaStockIndex”)/Index(“HighBetaStocksAvgBeta”).  Your index should appear on the Chart application.   We can now repeat this for the low beta stocks.  Looking at the graph, it is hard to compare the two indices, because they don’t start at the same point.  To make the indices visually comparable, they both need to start with the same initial value (in this case we will move them to start at 0).

Re-enter your plotting code for both times series, but this time subtracting the initial value from both of them.  For example, your low beta plot is now: Index(“LowBetaStocksIndex”)/Index(“LowBetaStocksAvgBeta”)-158.382.  You should now see two lines that start at the same point.  Delete the non-translated lines (the ones where we did not subtract the value of the initial point) by right-clicking the labels on the upper-left of the graph and hitting delete.  Make sure that you have only one axis on the right side.  If not, highlight both of them by holding down ctrl and clicking one then the other.  Once both are selected you can right click and choose merge.  Your screen should now look like this:


As you can see, the two portfolios basically track each other, as we might have expected, except for the major discontinuity last summer.

You’ve now completed your first study in Palantir Finance.  To recap, with only a few clicks and some relatively simple calls to built-in metrics, we discovered a very interesting time period for further analysis.   With a full data set and a few more studies, one could likely isolate the causes of the big dislocation in the summer of 2008.

[1] To conceptualize this, consider an index with an average beta of .5 and another index with an average beta of 2.  To compare them fairly, one would have to invest $400 in the low beta index for every $100 put into the high beta index.

Other Blogs