Jun

8

Kendall and Gibbons' book, Rank Correlation Methods, is a fascinating read which highlights many possibillities to test systems in markets and life. It suggests tests for variables of interest such as relative performance among days of the week or months of the year, agreement among two time series, and trends in time series and markets.

I found it interesting that if there are m separate rankings of n items, (for example Branch Rickey ranked 20 pitchers on such things as hits divided by times at bats, earned runs divided by the sum of hits, strike outs divided by the sum of at bats, earned runs per nine innings. He did this for pitchers such as Hubbel, Dean, Grove, Alexander, Vance, Leonard, Matheson, Johnson, Gomez, Derringer and Lyons), then the maximum sum of the differences between the total rankings is (m squared ( n cubed - n))/12.

I shall give examples of how this can be used shortly, but can say for now that such a test on weeks before, during, and after a four day week (such as memorial day week) showed complete randomness when tested under these various systems over the last ten years. The research and analysis on this subject is the first contribution of Ken Sogi (son of regular contributor, Jim Sogi) to the firmament of speculators.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search