The ranking system calculation is rather simple, as explained in the application help document and in the current article: Ranking System Engine , it parses and calculates formula, applies factor and weight values and finally assigns a value to each symbol and for each bar.
These calculated values are then assembled; each value must be assigned to a bucket. There are three different ways to accomplish this.
In the ranking system settings, there is a combo box under 'Ranking Calculation' where you can select what method to use.
Let us use an example to explain how calculation methods work.
For this example, we will choose 2 buckets.
This is a table containing different symbols with the ranking system values for each rebalance period.
The three calculation methods are:
The boundary between states is chosen so that each state is assigned the same number of positions:
With this method of calculation, the ranking system process will assemble all the generated values and sort them (16 values in the above example). It will then divide the sample size by the number of buckets (2); each sub-sample will be associated with a bucket.
Here is the result:
Values: 12 - 13 - 14 - 12 - 11 - 15 - 16 - 10 - 5 - 40 - 60 - 20 - 80 - 90 - 13 – 5
Sorted Values: 90 - 80 - 60 - 40 - 20 - 16 - 15 - 14 - 13 - 13 - 12 - 12 - 11 - 10 - 5 – 5
Sub-Sample Size: 16 / 2 = 8
Sub Sample 1: 90 - 80 - 60 - 40 - 20 - 16 - 15 – 14
Sub Sample 2: 13 - 13 - 12 - 12 - 11 - 10 - 5 – 5
Where B1 is Bucket 1 and B2 is Bucket 2.
The buckets will have the following boundaries: Bucket 1: from 0 to 50 and Bucket 2: from 50 to 100.
This is because the ranking system values must vary from 0 to 100 and we choose to have only two buckets.
On each rebalance date, sort the positions and give them a value from 0 to 100:
In this type of calculation, the ranking system processes the data for each bar independently.
It sorts the ranking system values for each bar, and then assigns a rank to each one. Because the range of a rank is 0-100 and we have only 4 symbols, the possible rank values are: 25, 50, 75 or 100.
After completing the above task, every rank will be associated with a bucket. Top ranks get the higher buckets. In our example, rank of 100 and 75 will be set to bucket 2 and rank of 25 and 50 will set to bucket 1.
The last ranking method uses the same logic as the ranking in the simulator and thus can be used to find profitable ranking systems to apply to trading systems.
NB: When a ranking system value is associated with a bucket, the symbol output value (return, return with N-Bars Stop...) for that particular bar is also associated with the same bucket.