Build mode

Author: admin

February 26th, 2019

Rating:
1 Star2 Stars3 Stars4 Stars5 Stars (No Ratings Yet)
Loading...

Settings that affect how strategies are generated. You can choose from two options – genetic evolution or random generation. 

For Random generation you only set when the build should stop. You can run it continually “forever” until it is manually stopped by you, or after databank is full, or after a certain number of strategies were generated.

Genetic evolution has more options:

Genetic Options

Max # of Generations
is the count of how many generations the evolution should be performed.

Islands (separate evolution)
number of islands in genetic evolution

Population size (per island)
is the number of different strategies in one population. It should be at least 50, the ideal value will be between 100 – 1000 or perhaps even more.

Then there are Advanced genetic settings available in a popup dialog:

Mutation Probability
is the probability that a strategy will be mutated, it should be around 5-15%.

Crossover Probability
is the probability that the strategy is selected for crossover operation. Every generation there is a crossover of strategies performed during which the two parent strategies “mate” and produce two children. Recommended value if 95%.

Island options

Migrate every X-th generation
how often to migrate strategies between islands. Recommended value is 5, 10 or even more. The less often strategies are migrated, the more the islands will be “separated” without exchanging information.

Population migration rate
Percentage of strategies in population that will be migrated. The recommended value is 5-10 %. We don’t want to destroy islands diversity by migrating too many strategies.


Initial Population Generation

evolution starts with initial population. It can be generated either randomly, or it could use existing strategies as a start.

Use systems from Initial population databank as evolution start
existing strategies can be used as initial population of the new genetic evolution. These strategies must be loaded to the Initial population databank. If there aren’t enough initial strategies, the rest will be generated.

Decimation coefficient
decimation means that the system generates more strategies than needed for initial population, and then chooses only the best. For decimation coefficient is 1 it means there is no decimation.

For decimation coefficient = 2 system generates 2 x population size strategies and then chooses the best half for the initial population.

For decimation coefficient = 3 system generates 3 x population size strategies and then chooses the best third for the initial population.

Filter generated initial population
when generating initial population you can specify additional settings to filter bad strategies you don’t want to use. For example, you can configure it to dismiss (throw away) strategies where Net Profit is below zero or where number of trades is below certain level.


Evolution Management

Start again when finished
if checked, it will restart the evolution when it finished (reached maximum number of generations).

Restart if fitness stagnates for X generations
if checked, it will restart if fitness stagnates for given number of generation. Fitness is the strategy score. If it stagnates for 5-10 generations it is usually a sign that evolution reached its potential or went into a dead end, so it might be a good idea to just restart it.

Leave a Reply

Please Login to comment