Two quick updates for the beta site:
- The name of … Name of the Wind has been corrected.
- Fixed a crash in the eaComplex solver algorithm when it mutates a potential sequence that is shorter than 2 actions.
Thanks to drewlt and Thortok for the bug reports!
We have a few more updates for the Heavensward beta site:
- A button has been added on the Crafter Attributes page for automatically selecting the actions that are available at that level.
- You can import character data (class levels and attributes) from XIVSync/XIVPads on the Crafter Attributes page. Coverage is a bit spotty though. This will hopefully improve once the new XIVPads API (thanks Josh!) is available.
- An overlay showing the CP cost has been added to action icons.
- An overlay showing a class icon has been added to cross-class action icons.
- Wasted actions are highlighted in the sequence.
- A button has been added on the Simulator page to revert the sequence etc. to the last loaded synth.
- Fixed a subtle (took me how many years to find it?!) off-by-one error in the cross-over and mutation functions. This has a surprisingly significant effect on the effectiveness of the solver.
- An additional constraint to minimize the sequence length has been added.
- The penalty for wasted actions has been reduced so that sequences containing wasted actions are considered only slightly less fit than similar sequences with no wasted actions.
- The upper limits on the population and number of generations in the options has been removed. Go nuts! Just don’t blame us if you melt your CPU.
- Various tweaks to the cross-over and mutation functions to reduce the extent to which they can modify a sequence.
- Added a new solver algorithm, creatively called “eaComplex”, which uses a different method for processing the population in each generation that tries to keep highly fit individuals in the population for longer, as well as adding a new type of mutation: adjacent pair swap.
The new solver algorithm performs quite well. Please give it a try and let us know how well it works for you.
As usual, comments below or issues on GitHub are welcome!