31.07.2021 Views

Ultimate Algorithmic Trading System

Using automated systems for trading in stock markets

Using automated systems for trading in stock markets

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

The results from the two tests are quite similar; results only start deviating after

the top six parameter sets. This example is not perfect for demonstrating the power

of GA, as the brute force approach only took 13 minutes, but the implications of its

power is more than evident. In many cases, a brute force optimization scheme will

be too time intensive, and the only alternative is a GA.

TradeStation makes using the Genetic form of optimization very simple. All

it requires is selecting it over the Exhaustive mode. This simple switch makes

implementation simple, but you do have the ability to override the Genetic

optimizer’s default settings. If you select Genetic as the optimization method

and then click Advanced Settings, a dialog box will open that will allow you to

override the default settings. Before changing these, you want to make sure you

fully understand what you are doing. Here is explanation of each of the Genetic

Optimization Settings.

246

GENETIC OPTIMIZATION, WALK FORWARD

Generations

This setting indicates the total number of generations each test will iterate to

come up with a solution chromosome. In our very first example, the number

of generations concluded after a single chromosome fulfilled the requirements to

solve the simple equation. The computer program and process could continue until

multiple solutions were uncovered. Trading algorithms don’t have a black-or-white

solution so you can specify to only go so far in the process. This number should

increase with larger search spaces. This setting should be ‘‘Suggested’’ by clicking

the Suggest button.

Mutation Rate

Remember how a gene in our example chromosomes could be randomly changed?

This rate defaults to 0.1 or 10 percent, and this is probably the best value for this

setting. Higher mutation rates increase the probability of searching more areas in

the search space and may prevent the population from converging on an optimal

solution. Searching a larger space takes more time and a mutation could randomly

change the gene of a good solution after selection and reproduction and turn it into

a nonoptimal solution, thus preventing convergence.

Population Size

Referring to our initial example of a simple GA, we used six different chromosomes

to solve our problem. So its population size was six. Population size grows with the

size of the search space. A small population on a large search space could lead to less

robust solutions. Initial population sizing is very important and is very difficult to

gauge. Again, let the Suggest button help here.

www.rasabourse.com

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!