How To: Use Automatic Skill Tree Generator
You can find the generator under the menu at Home -> Tools -> Passive tree Generator. In the window are some general settings and a tab for every different solver. "Advanced" contains the new functionality.
In the "Advanced" tab, "Attributes" is for the raw attributes, "Pseudo Attributes" is for combined raw attributes and "Settings" contains settings related to which raw attributes are calculated into pseudo attributes atm. One thing worth pointing out is that keystones will only be included in solutions if they are check-tagged (so if you check Phase Acro but not Acro and click "Run", it will give you an error because the tree is disconnected).
You can optimize an existing tree as follows:
- Skill your tree in the general UI.
- Open the skill tree generator window and make sure total points equals the number of skilled nodes (if you start it after skilling the tree and it is the first start, it should do that automatically, if not, change additional points until total points are correct or click "Reset").
- Switch to "Advanced" -> "Settings" and select weapon class, off hand and tags fitting your build and main skill.
- Switch to "Attributes", click "Load from tree" and then "Convert to pseudo attributes".
- Remove attribute constraints that are unimportant for you.
- Switch to "Pseudo Attributes" and remove constraints that are unimportant for you.
- Make sure the automatically check-tagged nodes are only nodes you feel must be included in any case (like keystones and important notables).
- Optionally check the "current tree as initial solution" check box. If similar trees are better, this will generally produce better results. If the tree needs to change heavily, it will generally produce worse results. If you keep it unchecked, make sure to reset the skill tree.
- Click "Run" and wait.
Do note that trees are very specific, so if it is already an acceptable tree, it probably won't get any better outside of pathing between nodes for which "Skill Tagged Nodes" is suited much better.
Note: I would like to thank @brather1ng for this mini tutorial and the feature itself. Pull Request #201