Top/Bottom Correlations on the Topology Screen

This section describes the Top/Bottom Correlations list displayed on the Market Topology screen.

This list displays the 100 most positively and 100 most negatively correlated assets (pairs, triplets, and quadruplets) with an absolute correlation value greater than 0.1, relative to the selected instrument.

To illustrate the different modes, we will use ConocoPhillips ticker COP as an example of the asset being researched or hedged (Leg A). Additionally, for this example, the ‘market universe’ will be limited to MRO, EOG, DVN, SPY, and several sector ETFs.

In this mode, CorrGrid identifies the assets most correlated to the selected asset (in our example, COP). The candidates, in our example, are MRO, EOG, and DVN, with correlations of 0.88, 0.84, and 0.81, respectively.

Beta coefficients, serving as optimal hedge ratios, are estimated by CorrGrid to minimize the volatility of two-leg positions. Each position in the raw mode consists of a $1.00 long position in the selected asset (Leg A) and a short position of β dollars in a hedging asset (Leg B).

R² measures the volatility reduction of the two-leg position (Leg A − Leg B) relative to an unhedged $1.00 long position in COP (Leg A).

For this example, CorrGrid estimated the following betas and R² values: 0.62 for MRO (R² = 31%), 0.81 for EOG (R² = 43%), and 0.69 for DVN (R² = 42%).

These results can be displayed in two formats: “Leg A vs. Leg B” (which compares the two legs, e.g., 1.00 × COP vs. 0.62 × MRO) and “Hedge Formula” (which shows the combined position, e.g., 1.00 × COP − 0.62 × MRO ).

1.00 × COP − 0.62 × MRO is 31% less volatile than COP alone. Other Hedge Formulas include: 1.00 × COP − 0.81 × EOG (R² = 43%), and 1.00 × COP − 0.69 × DVN (R² = 42%), indicating greater and similar reductions in volatility, respectively.

Therefore, in this example, EOG provides the most effective hedge, reducing volatility by 43%. This illustrates how the Top/Bottom Correlations table, with the “Hedge Formula” option enabled, allows you to identify the best readily available, straightforward hedge for any instrument monitored by CorrGrid. Notably, instruments like VXX and other volatility ETFs, which frequently exhibit negative betas, enable hedging with two long positions.

In this mode, to obtain purified correlation metrics, CorrGrid removes the influence of the broad market, approximated by SPY, from all instruments: the target asset (COP) and the potential hedging instruments (MRO, EOG, DVN).

This is done by calculating the beta of each asset with respect to SPY and subtracting the SPY componen. For example:

• Leg A^adj.SPY: 1.00 × COP − 0.34 × SPY
• Leg B^adj.SPY (candidates): 1.00×MRO- 0.38 × SPY , 1.00×EOG - 0.40 × SPY , 1.00×DVN − 0.58 × SPY

By removing the common market factor, this mode helps reveal hidden correlations driven by factors other than overall market movements, highlighting non-obvious relationships. Therefore, SPY can be considered a source of external noise that obscures the underlying relationships between individual assets.

CorrGrid then identifies the most correlated pairs between “purified” COP and “purified” MRO, EOG, and DVN (i.e., Leg A^adj.SPY vs. Leg(s) B^adj.SPY), estimating the corresponding betas. In our example, these betas are 0.6, 0.8, and 0.7, respectively.

These betas give us Leg(s) B:

• 0.6 × ( 1.00×MRO − 0.38 ×SPY) => 0.6 × MRO − 0.23 ×SPY
• 0.8 × ( 1.00×EOG − 0.40 ×SPY) => 0.8 × EOG − 0.32 ×SPY
• 0.7 × ( 1.00×DVN − 0.58 ×SPY) => 0.7 × DVN − 0.42 ×SPY

… and Hedging Formulas:

• (1.00 × COP − 0.34 × SPY) − 0.6 × ( 1.00×MRO − 0.38 ×SPY)
• (1.00 × COP − 0.34 × SPY) − 0.8 × ( 1.00×EOG − 0.40 ×SPY)
• (1.00 × COP − 0.34 × SPY) − 0.7 × ( 1.00×DVN − 0.58 ×SPY)

Simplifying the SPY component in the previous expressions, we get “triplets” or three-leg positions:

• 1.00 × COP − 0.6 × MRO - 0.112 ×SPY
• 1.00 × COP − 0.8 × EOG- 0.02 ×SPY
• 1.00 × COP − 0.7 × DVN + 0.066 ×SPY
  

Comparing the volatility of raw COP with the volatility of the triplets, CorrGrid estimates the following R² values: 31%, 43%, and 41%, respectively.

These positions retain a residual amount of SPY. In this example, we see that after applying the ex-SPY hedge, the volatility of COP is largely attributable to sector-specific (peers-specific) exposure rather than to SPY.

This mode is similar to ex-SPY, but instead of removing the influence of the broad market (SPY), it removes the influence of sector-specific movements using anchor ETFs.

For each asset, the most relevant large and liquid sector ETF is chosen based on robust correlations with a small hysteresis to prevent jitter in the selection from day to day. These selected ETFs act as anchored benchmarks.

As a result, we analyze even finer, more subtle relationships between instruments, which are not driven by broad market dynamics or even industry specifics, but by deep, pairwise, individual dependencies or membership in narrow subsectoral clusters.

In this example:

• Leg A^adj.ETF: 1.0 × COP − 1.1 × XLE
• Leg(s) B^adj.ETF: 1.0 × MRO − 0.88 × XOP, 1.0 × EOG − 1.00 × XLE, 1.0 × DVN − 0.99 × XOP
  

The correlation between “purified” COP and “purified” DVN (purified from sector noise) was so low (< 0.1) that CorrGrid did not display it in the table.

The R² values for the remaining two multi-leg positions are 53% and 54%, respectively.

This means that the triplet 1.00 × COP - 0.80 × EOG - 0.30 * XLE (derived by combining and simplifying like with prior example) possesses the greatest hedging properties.

It is important to understand that the anchored ETFs may differ; consequently, the resulting position may be a four-legged position (quadruplet).