Cross Correlation [Kioseff Trading]

2023년 9월 4일

Hello!

This script "Cross Correlation" calculates up to ~10,000 lag-symbol pair cross correlation values simultaneously!

• Cross correlation calculation for 20 symbols simultaneously
  
• +/- Lag Range is theoretically infinite (configurable min/max)
  
• Practically, calculate up to 10000 lag-symbol pairs
  
• Results can be sorted by greatest absolute difference or greatest sum
  
• Ability to "isolate" the symbol on your chart and check for cross correlation against a list of symbols
  
• Script defaults to stock pairs when on a stock, Forex pairs when on a Forex pair, crypto when on a crypto coin, futures when on a futures contract.
  
• A custom symbol list can be used for cross correlation checking
  
• Can check any number of available historical data points for cross correlation
  

Practical Assessment

Ideally, we can calculate cross correlation to determine if, in a list of assets, any of the assets frequently lead or lag one another.

Example

Say we are comparing the log returns for the previous 10 days for SPY and XLU.

*A single time-interval corresponds to the timeframe of your chart i.e. 1-minute chart = 1-minute time interval. We're using days for this example.

(Example Results)

A lag value (k) +/-3 is used.

The cross correlation (normalized) for k = +3 is -0.787
The cross correlation (normalized) for k = -3 is 0.216

A positive "k" value indicates the correlation when Asset A (SPY) leads Asset B (XLU)

A negative "k" value indicates the correlation when Asset B (XLU) leads Asset A (SPY)

A normalized cross correlation of -0.787 for k = +3 indicates an "adequately strong" negative relationship when SPY leads XLU by 3 days.

When SPY increases or decreases - XLU frequently moves in the opposite direction 3 days later.

A cross correlation value of 0.216 at k = −3 indicates a "weak" positive correlation when XLU leads SPY by 3 days.

There's a slight tendency for SPY to move in the same direction as XLU 3 days later.

After the cross-correlation score is normalized it will fall between -1 and 1.

A cross-correlation score of 1 indicates a perfect directional relationship between asset A and asset B at the corresponding lag (k).

A cross correlation of -1 indicates a perfect inverse relationship between asset A and asset B at the corresponding lag (k).

A cross correlation of 0 indicates no correlation at the corresponding lag (k).



The image above shows the primary usage for the script!



The image above further explains the data points located in the table!



The image above shows the script "isolating" the symbol on my chart and checking the cross correlation between the symbol and a list of symbols!

Wrapping Up

With this information, hopefully you can find some meaningful lead-lag relationships amongst assets!

Thank you for checking this out (:

2023년 9월 4일

릴리즈 노트

Further improved efficiency!