Monday, December 17, 2007

Gauging Market Strength and Weakness: Buying and Selling Events as Independent Variables

Most of what we look at is the summing or averaging of buying and selling events. Price moves higher during a given time period, and it moves lower at other periods. Bar charts sum these price movements over a longer period to show us overall price movement and change. Similarly, we sum buying and selling events via such indicators as advance-decline lines, NYSE TICK, moving averages, and oscillators.

Suppose, however, we adopt a different framework and view buying and selling events as orthogonal. Instead of summing and averaging them, we treat them as independent variables. Thus, for instance, any given time period might display very strong buying *and* very strong selling; very weak buying *and* very weak selling; or any permutation thereof.

Once we conceptually separate strong/weak buying from strong/weak selling, we're then in a position to ask interesting empirical questions:

* Do weak/strong buying days have different implications for future price change than weak/strong selling days?

* Do particular combinations of buying/selling levels identify trending days?

* Can we profile average levels of buying/selling at various times of day to gauge the market's current strength/weakness?

* Do topping markets show early signs of reduced buying and bottoming markets show early signs of reduced selling?

I'm just in the early phases of this research, but results to date look promising. I hope to have concrete results to report by year's end, resulting in a useful and unique market indicator.


Trend and Trajectory

Sentiment Trends


Anatrader said...


Your post about a new research on orthogonal variables would certainly add dimension to what we have at the moment to gauge market sentiment.

Another new word added to my vocab: orthogonal - as 'statistically independent'.

Adam said...


My comment is below this line:


Brett ~

You know that I read every word of your posts every day, and it’s not my goal to argue semantic trivialities; I’m trying to understand the mathematical concepts implied. Thus,
I’ve been thinking about this post a lot; the more I think the more difficult my thinking becomes (as if it’s ever easy).

To me, an orthogonal event is one in which a pair of functions over time has a scalar product equal to zero. Thus, in buying and selling over a unit of time there would be no overhang. While there are zero-sum buy/sell markets, I’m uncertain that the NYSE qualifies.

In an orthogonal event, the variables are treated as truly independent and thus could be said to be non-correlated, or, said another way, have x,y axes perpendicular to each other. While it’s possible that I’m missing the point entirely, I’m uncertain that buying/selling on the NYSE TICK can be abstracted to the point of non-correlation and still have analytic value.

In this post, you go on to ask four important questions, the answers to which could have real implications to a carefully designed trading methodology. Each of the questions entails measuring degrees of correlation (their non-orthogonal nature) between traits of the buy/sell NYSE TICK against each other.

I suspect I am misreading your intention in writing this very interesting entry, and hope that in a future post you’ll expand upon the ideas you’re sharing here.


Brett Steenbarger, Ph.D. said...

Hi Adam,

Great question. Buying and selling are independent in the same sense that positive and negative emotion are independent in the psych research (Diener, etc.). At any one time, people either have positive or negative emotion; they are either buying or selling on balance. So in the immediate time frame, they're inversely related.

As we look over time, however, we find that people with high or low degrees of positive emotion can also have high or low degrees of negative emotion. Similarly, in any given day, we can have very strong buying *and* very strong selling (e.g., lots of volume at the bid and at the offer in Market Delta; lots of high and low TICK readings; lots of periods of broad advancing and declining) or any combination thereof.

A great example is the Dow TICK (TIKI), which is sensitive to program buying and selling. By distinguishing between positive and negative TIKI readings, we can determine the intensity of buy and sell programs in the market.

Stated otherwise, if hour number one shows very strong buying and hour number two shows very strong selling, averaged together they give the same neutral reading as a relatively flat hour one and two. Separating the buying and selling activity over the course of a day or week helps us see the vigor with which participants are acquiring or disgorging stocks.

Thanks for the excellent question.