|
|
|
|
 |
Daily Speculations The Web Site of Victor Niederhoffer & Laurel Kenner Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter; a forum for us to use our meager abilities to make the world of specinvestments a better place. |
Write to us at:
(address is not clickable)
4/4/05
The Definition of Counting
One thing which has been missing from our recent discussion is a definition of counting. Undoubtedly we all think we know what it means but sometimes it is helpful to critically examine our own ignorance. It seems the markets are continually trying to teach us that lesson.
THE DEFINITION
Counting is a scientific process whereby numeric or classified data is used to test a predictive hypothesis for statistical significance in order to gain a mathematical advantage in the market.
Explanation
Essential to the above definition is the language of statistical hypothesis testing and the use of the scientific process. The hypothesis must be well defined and falsifiable. Opinions and vague visual interpretations are anathema to Counting. Usually in such circumstances the countist will seek to use the raw numbers or classification data in lieu of charts or methods which require subjective interpretation.
Classified data is also quite useful in counting and can include such events as "the CFO just left", 3 or more insiders bought, today is Tuesday, the market was up (+) or down (-) yesterday. The data which can be analyzed under the Counting rubric can be quite broad, need not be numeric and is not limited to price, volume, open interest or traditional accounting data.
Simple hypotheses are preferred in Counting because each time we add a condition we reduce the size of the sample available to test for significance. However more complex patterns are allowed. An example of a pattern event might be "two weeks with lower lows while a Republican is in office". One possible way to test such a pattern would be to classify the data into all cases where the event is true versus all cases where it is not true. Then the statistical test might be to test whether there is a difference in the means between the two cases or a difference in the proportions which are up for the two cases. Any valid statistical test can be used to test for significance.