Some more Insights into the Genetic Evolution in SQX

Here’s a breakdown of the best settings for refining existing strategies versus exploring または testing lots of different strategies. These modes require different configurations, as one focuses on improving known strategies, while the other seeks to explore a wider range of possibilities.
<h3>Key Settings for Refining Strategies:</h3>

1. Max # of Generations: Keep this value moderate (20–50) to give the algorithm time to refine existing strategies. More generations help to polish the strategies without wasting time exploring too many random possibilities.
2. Population Size: Use a larger population (50–100) to give the genetic algorithm more material to work with, but without too much diversity.
3. Crossover Probability: Increase crossover probability (0.8–0.9) to refine good traits from successful strategies.
4. Mutation Probability: Lower mutation probability (0.01–0.05) since you want to make small adjustments rather than introducing major random changes.
5. Islands: Fewer islands (1–2) are best for focusing on refining strategies, ensuring that effort is concentrated in a smaller search space.
6. Migrate Every Xth Generation: Less frequent migration (every 20–30 generations) keeps the island focused on refining local optima without disrupting it too often.
7. Replace X % of the Weakest Strategies: Set a low replacement rate (5–10%) to carefully weed out the worst strategies while refining the rest.
8. Generated Decimation Coefficient: Keep this low (1–2), as the focus is on refinement and you don’t need to waste time generating many extra strategies.
9. Filter Generated Initial Population: Use more stringent filters for the initial population to ensure only higher-quality strategies enter the refinement process.
10. Detect Same Strategies and Replace Them: Keep this option off or at a low level to avoid introducing too much randomness into the refinement process.

<hr />

<h3>Key Settings for Testing Lots of Different Strategies:</h3>

1. Max # of Generations: Lower (5–20) to prioritize broad exploration of different strategies, restarting frequently to cover more ground.
2. Population Size: Use a smaller population (10–30) to quickly explore a wide range of possibilities.
3. Crossover Probability: Use a moderate crossover probability (0.5–0.7) to balance blending strategies and creating new ones.
4. Mutation Probability: Set mutation probability higher (0.05–0.1) to encourage greater diversity and exploration of new strategy ideas.
5. Islands: Use more islands (3–10) to test different strategies across separate populations, increasing diversity.
6. Migrate Every Xth Generation: More frequent migration (every 5–10 generations) helps cross-pollinate ideas between islands and unlock stuck populations.
7. Replace X % of the Weakest Strategies: Set a higher replacement rate (20–50%) to quickly cycle out weaker strategies and explore new ones.
8. Generated Decimation Coefficient: Increase this (2–5) to ensure a larger pool of strategies is generated for exploration.
9. Filter Generated Initial Population: Use more relaxed filters to allow a wider variety of strategies into the initial population.
10. Detect Same Strategies and Replace Them: Keep this option オン to ensure that duplicate strategies are replaced with new ones, promoting diversity.

<h3>Refining Strategies:</h3>
When refining a set of known strategies to improve their performance, it’s important to prioritize settings that promote stability and fine-tuning over wide exploration. These settings will help you hone strategies that already show promise:

1. Max # of Generations: Set to around 20–50 generations to focus on incremental improvements to the strategies without going too far. This range allows sufficient evolution while keeping the process efficient.
2. Population Size: Use a relatively large population size (50–100) to give the genetic algorithm more strategies to work with. This larger pool helps the algorithm refine solutions more thoroughly.
3. Crossover and Mutation Probability: Set crossover high (0.8–0.9) to blend good traits effectively, while keeping mutation low (0.01–0.05) to avoid introducing too much randomness.
4. Islands: Use 1–2 islands to keep the process focused on refining local optima, rather than exploring radically different strategies. With fewer islands, the algorithm can concentrate on the refinement process.
5. Migrate Every Xth Generation: Set migration frequency low (every 20–30 generations) to allow strategies to evolve within their islands without too much interference.
6. Population Migration Rate: A lower migration rate of 1–2 strategies (10–20%) is recommended to maintain a balance of diversity and focus on fine-tuning.
7. Replace X % of the Weakest Strategies: Set this to a lower rate (5–10%) to carefully replace the weakest strategies without drastically disrupting the population.
8. Generated Decimation Coefficient: Set this low (1–2), as generating too many extra strategies won’t help in the refinement phase. Instead, focus on improving the current population.
9. Filter Generated Initial Population: Use stricter filters to ensure only quality strategies are allowed to enter the refinement process.
10. Detect Same Strategies and Replace Them: This should be off or set low to maintain focus on fine-tuning the existing strategies rather than introducing new ones.

<hr />

<h3>Testing Lots of Different Strategies:</h3>
For broad exploration or testing many different strategies, the focus is on diversity and speed rather than fine-tuning. These settings encourage the genetic algorithm to explore a wide range of possible strategies, improving the chances of finding novel solutions:

1. Max # of Generations: Keep this low (5–20) to prioritize fast exploration of many strategies. Restart frequently to cover more ground instead of over-evolving a small set of strategies.
2. Population Size: Use a smaller population (10–30) to allow for quick testing of different strategies. This keeps computational costs lower while increasing the breadth of exploration.
3. Crossover and Mutation Probability: Set crossover to a moderate value (0.5–0.7) and mutation higher (0.05–0.1) to encourage diversity and exploration of new ideas.
4. Islands: Use more islands (3–10) to test a wider range of strategies across different, isolated populations. This increases the variety of strategies being explored simultaneously.
5. Migrate Every Xth Generation: More frequent migration (every 5–10 generations) ensures that successful strategies from one island can influence others, preventing stagnation.
6. Population Migration Rate: Use a higher migration rate (20–50%) to promote greater cross-island diversity and introduce fresh ideas to each population.
7. Replace X % of the Weakest Strategies: Set this to a higher rate (20–50%) to quickly cycle out poor-performing strategies, keeping the population dynamic and open to new possibilities.
8. Generated Decimation Coefficient: Use a higher decimation value (2–5) to generate more strategies initially, ensuring the best ones are chosen to start the exploration.
9. Filter Generated Initial Population: Use looser filters to allow for greater diversity in the initial population.
10. Detect Same Strategies and Replace Them: Keep this option オン to prevent duplicates and encourage exploration of new strategy possibilities.

<hr />

By tuning the settings in these two modes, you can either focus on refining and perfecting existing strategies or test a wide range of different strategies to explore new opportunities. Each approach requires a different balance between diversity and focus, and the settings above will help guide you toward the most efficient configuration based on your goals.