Craftworks Planner: Finding Better Recommendations

When I last wrote about the Craftworks Planner, I explained how it recommends which recipes deserve your available Craftworks slots.

Since then, I've been testing how well those recommendations hold up.

And I ran into something unexpected.

When fewer choices lead to better results

While testing the planner, I noticed that excluding certain recipes sometimes produced a better recommendation.

That seemed strange.

The excluded recipes weren't necessarily bad choices. The better setup should have been available all along, even without excluding anything.

The planner simply hadn't found it.

Craftworks has limited slots, but potentially many recipes to choose from. Comparing every possible combination can take a lot of work, so the planner explores promising setups without checking every possibility.

That makes generating recommendations manageable, but it also means better combinations can sometimes be missed.

I knew this was possible. What I didn't know was how much it affected the results.

Searching smarter, not just harder

I started experimenting with different search settings to see whether the planner could find better setups.

One adjustment made a noticeable difference: allowing the planner to consider more possibilities overall while being more selective about which setups it explores along the way.

In my gameplay test, the estimated manual crafting batches dropped from 2,889 to 1,411 under the Queue planning mode. That's roughly 51% less manual crafting work remaining in the recommended setup.

The planner also found a configuration that used eight slots instead of six.

Interestingly, increasing the search limit even further didn't improve the recommendation.

That taught me something useful: giving the planner more room to search doesn't necessarily help if it isn't exploring the right combinations.

Of course, these results come from my own gameplay. Different players have different quests, resources, and available slots, so I can't assume the same improvement applies everywhere.

What comes next?

For now, I'm continuing to test the planner through my own Farm RPG progress.

That's how most of the Quest Tracker has evolved: I encounter a problem while playing, build something to address it, and find out what needs improving as I use it.

I'd eventually like to test Craftworks against a wider variety of gameplay scenarios, but that isn't something I need to finish before the initial release.

There's also a performance concern I'm looking into. Some demanding calculations currently take around four to five seconds and can temporarily make the interface unresponsive. Moving those calculations into a Web Worker could help keep the interface usable while recommendations are generated.

But my priority remains the recommendations themselves.

I don't expect Craftworks to find the mathematically best setup every time. I want it to provide useful suggestions, explain why recipes were chosen, and make quest preparation easier.

v0.6.0 is still in development, with no confirmed release date.

The planner can already recommend setups and show what crafting work remains. I'm continuing to test whether those suggestions make sense during actual gameplay before releasing it.

After all, the goal was never to build a perfect optimizer.

It was to help answer a much simpler question:

What should I put in my Craftworks slots?

And I'm getting closer to an answer I'm comfortable sharing.

— kodyy