Trading Strategy

Simplified version of the qualitative trading strategy I used when trading.

Edit: This is only a part. I have the complete thing, but don't see the point in publishing it. There is a flaw in this as well, which again, I see no point in correcting publicly.

The goal of this approach is to combine many various prediction factors into one working prediction model/strategy.

Let's say there are N+M = T total factors which impact the output of a process.
The output of the process can be either 0 or 1(binary in the simple version).
The factors can point to (or predict) the output being either 0 or 1(binary in the simple version).
Examples:

1. Market either goes up (1) or down(0).
2. Football match won by Barcelona(1) or lost(0).
3. Messi scoring a goal for Barcelona(1) or not scoring(0).
4. Barcelona being the home side(1) or being the away side(0).
5. 5 day moving average being greater than 9 day moving average(1) or lesser(0)

Out of these total T factors there are N constant factors and M variable factors.
The distinction between the two is that amongst the constant factors we are more sure of their impact on the output of the random process.
Simple Example:

1. Barcelona being the home side might have a greater impact upon Barcelona winning than the chance of Messi scoring a goal (because it will be known if Barcelona is the home team and if Messi will score will not be known priori).

Now, the claim is that we can obtain a hypothetical "maximum betting point".
Imagine a time line. At some points in this time line a subset of the constant factors will point to 1(or 0) and a subset of the variable factors will point to 1(or 0).
Now, the "maximum betting point" will be the point in the time line where all the constant factors point towards 1 and all the variable factors point towards 1.

Tasks to be done:

1. Identify various factors impacting the output of the process.
2. Arrange them in order of surity of impact on the output.

Cons:

1. In practice these tasks can be hard to do. 
2. The maximum betting point may take a long while to arrive on the time line, so we have to figure out a way of betting on the suboptimal points, which can be done by arranging the factors in order of surity of impact on the output on the process and prioritizing those factors which have a more direct impact.

But, we can accomplish a simpler task using this approach, which is to combine the predictive power of many different factors into one working model. A simple example is given below combining two factors, one predicting the movement of a market using VWAP and another predicting the same via 5 period moving averages.

Let's say the set of points on the time line TL, where VWAP based factor points to market going up is {V} and the set of points on the time line TL, where the 5 day moving average factor points to market going up is {M}.

We take the intersection of these sets - {U} and bet on those points.
Alternatively we can add more weights to {U}, compared to {V} then {M} - assuming we knew that VWAP had a better predictive power than 5 day moving average.


Some Observations:

1. In itself this layout of a strategy or plan is not very impressive.
2. Even then, what's important is that if you are trading manually you try and shift yourself towards a system, after the point that you begin to have a basic grasp of reading the tape or market action.
3. This is just a method I used to help me do the same. Nothing in this blog is a secret or will explicitly help you to make money.