Call me pc illiterate, but how do we use this ? Do we download the zips and it creates a program ? Great job btw abnd appreciate your hard long work
You too, huh? It's too bad I'm a lazy bum and never published my work; we both built the same tool to crawl and parse the wiki. I spent a lot of time cleaning up wiki pages as well. There used to be tons of special characters used in place of quote marks, and I standardized a bunch of them to regular programmer's quotes. Are you interested in a second contributor to your project? My gear planner has a local database currently (although the format is BAD; I've been meaning to replace it with something better) and the ability to suggest entire gear load-outs automatically. It's also missing basically all the nice features yours brings, like the tools for manual searching, and having a UI at all.
I like that it shows in red where items have duplicated benefits and the double click to see which items they are. I look forward to being able to "save" the sets for quick reference while TRing and leveling back up. It should make it faster to rummage through items (especially with the changes coming in the U45 shared bank update) and get back to questing.
Thanks!
Tequira Sunrise ~ Josielynne ~ Dawnalynn ~ Litzy so Ditzy
Ghallanda: Live and Make Die
First, thanks for your efforts cleaning up the wiki; I know all too well the pain that can be. Second, gear set discovery was one of my primary goals, and v0.5 includes a first pass at it. I'm not sure the project is big enough to require a second contributor, but any comments or suggestions you may have are welcome.
I've updated the original post with the v0.5.1 release. I can't stress enough how important upgrading to v0.5.1 is, if not for the sake of memory usage, for the UI adjustments to the main window.
If you'd like, I could offer a history of my design. If nothing else, everything after the first version has been more-or-less a failure, so you might learn what not to try!
Your design for knowing which stats to select is more elegant than mine. Mine runs a brief combat simulation to determine the effect of stats. As you can imagine, this makes "scoring" a particular gear set candidate pretty expensive. I think it clocks in at about 20M sets/hour? Of course, performance will vary by hardware.
In order to find a "best" set to return as output, my original algorithm was to try random items in every slot (respecting locked slots from the start) and output whichever scored highest at the end. This yielded surprisingly good results by running it overnight. Next, I let it ignore items more than 8 levels below the target level. This helped. Then I made it pre-score "best-in-slot" lists per equipment slot, and weight the random algorithm towards "better" candidates per slot. This was SUPER helpful, and remains the best algorithm I've found to date. Since then, I've been trying to optimize its run time, without degrading quality of output. That... has not succeeded. I tried letting it fill in potential set items first (didn't help), pick an artifact first (didn't help), pick augments last (didn't help).
I'm currently working on a new underlying algorithm (which so far, does not help). The new algorithm enumerates all permutations of the remaining equipment slots, then iterates over the enumeration. At each level, it tries all items for the "first" slot that have at least 66% of the stat value as the best-in-slot for that slot, and ranks those options based on the best outcome from a recursive walk of all remaining slots. Then, back at the top level, it picks the best slot to pick and finalize first, based on the best end-results from these deep recursive crawls. Then it repeats, including deeper searches on crafted and lootgen options as it gets down to a small number of slots remaining. Surprisingly, this hasn't helped!
Way cool, thank you for sharing.
First love the program starting using it couple days ago, thank you!
I seen the Build Filters section. Was curious about a way to filter old equipment from the list so that when looking say constitution up i would get not also the old gear with the lower numbers mixed in with new gear and their higher numbers. Is there a way to filter as well by constitution 8 or any number and see items that way? Or just a way to filter out old named items that are really outdated due to new items being so much stronger when i am looking for a specific item property?
The other thing is under Item Property, there a way to have it also take you to wiki so i can know what some of them are or better a way to over above it and it tell you the description of this item property?
Thank you,
Flintridge
With the build filters, you can set the ML slider in the main window for the build to help cull out the lower valued items.
With respect to item properties, for now I'd suggest going to the item page of the item that has the property you're wondering about. if you're not sure if you can, just right click the item in the list.
I'm a little confused on how to use the Build Filters and 'Start Build' functions in the main interface. I added a lot of build parameters that I was looking for, I clicked 'Apply' and set a minimum level, but when I click 'Start Build,' the Minimum Level slider resets itself and nothing else happens. I haven't been setting any gear filters, because I don't particularly care whether a trait is given to my by a ring or by the boots.
So, what am I doing wrong? Am I just misunderstanding what that function is supposed to do?
I've updated the original post for releasing v0.5.2, which has a number of quality-of-life improvements.
Okay, I updated and tried again, and through some careful testing, I was able to find that I was able to generate builds UNTIL I added ability scores as a trait, like 'Intelligence' or 'Constitution.' If I had the same build, but with an ability trait, it wouldn't start that little window that said 'creating item sets.' If i just deleted that one trait, then it would start creating builds.
I also found that it does this with Dodge. PRR and MRR
Last edited by SpookyNoodle; 01-29-2020 at 11:06 PM.