CSI Market Data

Intermarket Correlation Analysis

Intro. Pricing FAQ Starter Picks One Vs One One Vs All Portfolio All Vs All Preferences

Correlation FAQ

Q. Where do I start?

A. It is recommended to use "One Vs. All" search as a starting point. A single commodity market (ie: HG for copper) compared with all stocks can show some interesting results about companies that use copper, and the correlation between them.

Q.When I am ready to trade, what trades and positions are recommended?

A.For highly positively correlated pairs: the logic assembed for review, and as stated below the chart displayed, says to sell the red market and buy the blue market when the absolute value of the green line is high on the current day.

Q.When do I make my trades?

A.When the absolute value of the green line is high (above your given threshold, recommended is 80%) on the current day for a pair of highly positive correlated markets, the two trades are to be executed. When the absolute value of the green line gets low (below your threshold), the two positions should be exited.

Q.What position sizes do I use for my trades?

A.The position sizes are given on the one vs. one correlation pages. The necessary values to balance full contract value and to balance volatility are given.

Q. What exactly is the green/gray signal line?

A. The Perpetual Contract® or cash series are first convered to Z-scores. The Z-score is the value rescaled by the average and standard deviation. The green or gray signal line is the Z-score of the difference or sum of the Z-scores for each of the pair of markets, respectively. Algebraically, this works out to
Where Z1 is the Z-score of the first market, Z2 is the Z-score of the second market and ρ is the linear correlation coefficient between them.
As a result of these conventions, the cummulative probability distribution is

Q. What does "Show only pairs with Correlation between" mean?

A. This study produces Correlation Coefficients that succinctly express the historical relationship between any two markets. Correlation is expressed as numerical values between -100 % and +100 %, reflecting the positive or negative extent to which members of a given pair are related. A strong negative reading suggests that one member of the pair consistently moves up while the other moves down. Conversely, a strong positive reading suggests there has been a tendency for the pair of markets to move in a similar direction. A correlation coefficient very close to 0.00 implies that the two markets have exhibited only coincidental association. The Correlation Filter is a screening device that sets the lower limit of correlation coefficient values to be reported. An absolute value, it will apply to both positive and negative results. For example, use the default correlation coefficient filter of between 0.5 and 1.0 to view all market pairs whose correlation coefficient is greater than +0.5 or less than -0.5, which would include everything from +0.5 to +1.0 and/or everything from -0.5 to -1.0. This default value would exclude market pairs whose correlation values range from zero (0) to +/-0.4999. This filter can be combined with the Signal Line filter to further screen your results. To show all results without any filtering of correlation values, enter 0 (zero).

Q. What does "Show only pairs with Signal Line between" mean?

A. The green "Signal Line" that displays on the single-pair charts reflects the Z-score of the spread between the Z-scores of the two markets being analyzed. For highly positively correlated markets, signal line values near zero indicate that the two markets are in-sync with their normal relationship. The greater the signal line value (either positive or negative), the more out-of-sync the markets are. For a normally distributed random variable, a Z-score greater than 2 should only occur 5% of the time, and 3.2 should only occur 0.1% of the time. Signal values greater than two are considered significant, highlighting fluctuations in price between the two markets that may be tip-offs for good spread trading opportunities. For strongly negatively correlated markets, the signal line behaves like the average of the Z-scores for each of the markets. If either market is reaching historical highs or lows, the Z-score of the difference will be elevated in absolute value. This is especially true when both markets are making historical moves in opposite directions. Since strongly negatively correlated markets tend to move in opposite directions, their divergence is a confirmation of significance, which is reflected in a high (positive or negative) Signal value. Since market data is rarely normally distributed and since the Z-score only corrects for the first and second moment of the distribution, this measure needs to be tempered with an understanding that the Signal value relates specifically to the markets being analyzed. For some markets, the probability of unlikely events is less than for a normal distribution. These are said to have short tails. Some markets are more prone to unlikely events than that of a normal distribution. These are said to have long tails. For a short-tail market, a Signal value of 1.5 might be very significant, and yet for a long tail market, a signal value of 2.5 might still not be worth considering. With that said, filtering your searches to markets with at least a moderate current Signal value help you to narrow down your analysis from the onset. The Current Signal Line Filter is a screening device that sets the lower limit of terminal Signal values to be reported. Like the Correlation Coefficient filter, this is an absolute value that applies to both positive and negative results. The default of +/-0.75 excludes market pairs whose most recent Signal Line value is so close to zero that no exceptional trading situation is likely. Use a higher value (perhaps 2 or even 3) to find markets whose current relationship is quite unusual. This filter can be combined with the Correlation Coefficient filter to further screen your results. To show all results without any filtering of Signal Line values, enter 0 (zero).

Q. What does "Include Macro-Economic Series" mean?

A. Macro-Economic Series include government financial reports, indicies, some cash prices, some foreign exchange rates, and some spot interest rates. Some of these series are not tradable instruments, so their primary value in correlation analysis may be as background information. Macro-Economic series tend to so highly correlated that they are abundantly (perhaps over-abundantly) represented at the top of Correlation Reports. For these two reasons, most users exclude the Macro-Economic Series by leaving this box unchecked.

Q. What does "Include LME Series" mean?

A. LME Series are forward contract futures trade on the London Metals Exchange. Although they are tradeable in London, one must understand that a 90-day forward copper contract today is an 89-day forward copper contract tomorrow. Traders have sometimes made the mistake of thinking that they had large paper profits when infact they had large brokerage losses. You may also wish to exclude the LME series; because, the different forward terms are so highly correlated that they are abundantly (perhaps over-abundantly) represented at the top of Correlation Reports. For these two reasons, most users exclude the LME Series by leaving this box unchecked.

Q.What does "Color Negative and Positive Correlation Coefficients Differently" mean?

A. In the Portfolio view, the correlation coefficients are color coded according to how strong they are. Correlation coefficients near 0 are very weak. Correlation coefficients near +1 or -1 are very strong regardless of sign. Thus some customers want to see only what coefficients must be treated carefully (strong correlations) without distinguishing between positive and negative. If you want the positive coefficients to be in red and the negative coefficients to be in orange, then check this box.

Q. What does "Show Negative/Positive Correlations" mean?

A. Positive correlation reflects a pair of markets that tend to move in the same direction. The stronger the positive correlation, the greater the tendency of the markets to move in tandem. Negative correlation reflects markets that tend to move against each other. To help you to find the market situations of most interest, CSI reports on the markets which are positively or negatively correlated, or both, according to your selection. You can view the opposite perspective by running the study a second time and selecting the alternate choice.

Q. What does "Type: Stocks, Commodities, or Commodities and Stocks" mean?

A. When running an "All vs. All" or "One vs. All" analysis, the "Type" refers to the segment(s) of the database to be included in the screening. You can focus your search be choosing either "stocks" or "commodities," or, for the most comprehensive search in a single run, select "commodities and stocks."

Q. What does "Max Results To Return" mean?

A. The massive CSI database could produce Correlation Reports that hold perhaps millions of results. In order to generate a useable output that can be downloaded in a reasonable amount of time, use this option to limit the number of market pairs reported. The default, 2,000, is a practical setting for users with a high-speed Internet connection. We recommend limiting this number to gain maximum benefit from the results provided.

Q. What does "Commodity Symbols:/Stock Symbols:" mean?

A. These are the exchange symbols for the markets you wish to view. This entry is not case sensitive, so you may use upper- or lower-case letters. Entries are separated by commas or spaces. In the "Commodity Symbols" list, enter all of the commodities/futures that you wish to include in your analysis. Similarly, in the "Stock Symbols" list, enter all of the stocks/indices that you wish to include. All pairs of markets, including Futures pairs and Stock pairs will be included in the analysis.

Q. What does "View Mode" mean?

A. When viewing a matrix of selected markets through the "Portfolio" correlation screen, you have the choice of displaying an HTML-friendly matrix or a square matrix. The HTML-friendly matrix shows one column of all selected markets with a maximum of six columns of correlation coefficients for related markets. Additional matrices are displayed as needed to show each successive six columns of data. The "Square" matrix will be as wide as necessary to display all selected symbols as they relate to all other selected symbols in a single table, but the browser scroll bar may be needed to see the full report.

Q. What does "Period" mean?

A. Correlation coefficients can be calculated over various time periods ranging from ten years to 40 years. The longer the time period selected, the fewer results will be available because many commodities and stocks simply do not have long histories. In addition to correlation coefficients for the specified time periods, the website also shows the correlation value of the time period displayed in the chart, which may be less than 10 years. It is useful to examine the correlation coefficients for all time periods that display above the "One vs. One" charts to determine if the correlation values are consistent over time. The less consistent the correlation coefficient is historically, the more likely that it is a coincidence, the less likely that it will hold true going forward, and the less likely that the markets will respond as expected.

Q. What does "Symbol, Type:" mean?

A. The "Type" classifies the market symbol or name you have entered. It eliminates the confusion that can occur when the same symbol represents both a commodity and a stock. For example, if you have entered the symbol C, selecting the type "Stocks" would indicate you want the stock "Citigroup;" however, if you select the market type "Commodities," the symbol C would indicate the commodity "CBT Corn." If you entered a name, then your "Type" choice of Stock or Commodity will determine the appropriate search list to use.

Q.What parameters are used to calculate the standard deviation in order to equalize the contracts for volatility?

A.One standard deviation of corn price movement over time vs. One standard deviation of soybeans movement over time will give rise to a ratio of corn volatility to soybean volatility. Not corn price to corn price.

That same ratio measured in corn price volatility to soybean price volatility in standard deviation units will give rise to the volatility ratio of the product values. This would be the "volatility" reading that is reported in the contract sizing recommendations on our website.

The corn price to soybean price example is simply price valuations over time for the respective products from day-to-day, and is based solely on price.

* Short History Disclaimer and Trading Risk
[ Expand ]

All information and data on this website are © 2015 Copyright by Commodity Systems Inc. (CSI) All rights are reserved.
800-274-4727 | 561-392-8663 | 561-392-7761 (Fax)