Playing the odds: Part 1

Regular readers will know that the odds of a successful trade figure prominently in my analysis. In fact, the very first thing I do when analyzing off the symbols that have given bull or bear signals in a day is to pick out those among them whose breakouts have better than even odds of producing a profitable trade.

The weakness of odds is that they're historical. They look backward. They tell what has been.

And there's no way it can be otherwise. I can't get data from the future. The past is all I've got.

A characteristic of odds, and indeed all statistical analysis, is that the more examples you have in your analysis, the greater the likelihood that your analysis is correct.

That presents a challenge. That characteristic means that the further back in time I go with my analysis, the more examples I have and the more precise my results are.

But markets are changeable beasts -- just ask the gold traders -- and the older the data is, the less likely it is to reflect the present market reality.

I'm stuck between rock and a hard place -- analyze a shorter period back and gain nimbleness in my trading, or analyze a longer period and gain precision.

It calls for a compromise, and here's how I'm handling it now.

My universe of symbols, drawn from all symbols traded on the major exchanges, consist of those with  better than even odds of a profitable signal in one or both directions within the last year. Selecting this universe out of the 9,000-plus total symbols traded takes time, so I update the universe once a week.

My daily initial screening, to identify those signals with greater than even odds of success in the direction, also relies on the odds of success for breakouts within the past year, and it is those odds, applied to the average yield for breakouts within the past year, that I use in calculating a magnitude score that combines the odds and the yield, narrowing my list still further.

Most stocks and exchange-traded funds have only a handful of signals each year, so the analysis for each is based on a fairly small, but also timely, data.

At that point, I'll usually be in a position where I can take a look at the longer-term odds, going back to early 2009, when the present post-recession recovery began. Those will be a factor in my final selection.

The procedure is the opposite of what I was doing up to this week. Previously, I gave priority to the long-term odds and yields.