ffxiv-craft-opt-web
ffxiv-craft-opt-web copied to clipboard
doxxx
Shadowbringers (5.0) Updates
All 5.0 recipes are missing from the crafting solver, and several actions need removed or changed in their placement. For example, all of the "Brand of [Element]" and "Name of [Element]" actions have been removed along with Byregot's Brow, and Byregot's Blessing and Steady Hand II are now innate skills and no longer cross-class actions, despite showing up in the cross-class macro generation.
Further, recipes over level 70 are not present.
Other changes may include calculations for recipe difficulty over 70, although the calculations for progress/quality seem accurate when using the solver for recipes =<70 even with gear >70.
There are also new foods e.g. blood bouillabaise which are not in the database for buff purposes.
There were a few minor adjustments as well, such as Inner Quiet and Innovation having a combined control cap of 3000, all specialist heart skills being limited to five uses per recipe, and Muscle Memory and Piece by Piece being capped at 1000 progress. I'm not certain if there were any others beyond those.
EDIT: Maker's Mark was limited to 25 steps as well.
I have scraped the new recipes but I had to guess at the effective levels. However, my source for food/medicine buffs no longer works and I need to rewrite that code to use a new source.
I've uploaded the updated recipe database to the beta site. This includes things like level 71-80 recipes, the removal of elemental aspects and other new and changed recipes in the 1-70 range.
Let me know if the effective level for the 71-80 recipes seems off.
Suspect the scraper has some kind of issue, most of the 71-80 recipes are still missing. For example, in-game the Blacksmith 71-75 has 56 new recipes, like this:
Yeah, there was another change I had to make to the scraper. It's running again now but it takes a while since the Lodestone server throttles how fast it can load pages. I'll probably be able to update it tomorrow morning.
http://ffxiv-beta.lokyst.net/
On Sat, 6 Jul 2019 at 02:50, ikoyy [email protected] wrote:
I feel stupid, where can I find the beta site to try out the 5.0 algorythms?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/doxxx/ffxiv-craft-opt-web/issues/272?email_source=notifications&email_token=AAFFQNFX2UAOKD5LOIKL4NTP6A6DZA5CNFSM4H4KKKH2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZKTY5A#issuecomment-508902516, or mute the thread https://github.com/notifications/unsubscribe-auth/AAFFQNBESBKYPAEQF6XCVT3P6A6DZANCNFSM4H4KKKHQ .
I know this is a separate issue but at least somewhat related to 5.0 My level is now above 70 and I cant set my level to be above 70 in the attributes section.
Another item not yet addressed is the various 'elemental' brands and names have been consolidated into one of each, "Name of the Elements" and "Brand of Elements." Appears to be available to all job types at level requirements, so not required to be a cross-class action.
More recipes should be available now and I've increased the max level in the attributes page.
Name/Brand of the Elements poses a bigger problem. Some countries like China are not on patch 5.0 yet AFAIK (@kongspark can you confirm?), so I can't simply remove the Name/Brand of Wind/Water/etc. actions. One solution would be to introduce a setting for patch level, which is a significant change that will take some time. I could add Name/Brand of the Elements in the meantime, but the solver won't know to avoid the old Name/Brand actions.
Although, I just realized that the recipe updates have removed elemental aspects so that's broken it partially for China already. :P
For the Chinese players, I've deployed the latest patch 4.x compatible code to the main site and will continue deploying patch 5.0 changes to the beta site.
Hey @doxxx have you considered changing the source of recipes mined from the ffxiv-datamining CSV's shouldn't have any issues with parsing that and getting all that data points, could also help replace the tooltip issues. It has all the data. Alternative you could use the tools they use to extract that data directly
@unfaiyted That looks interesting, although the data is very noisy. Lots of weird incomplete entries. And it looks like it's missing a lot of the crafting actions.
I'm currently looking at using Garland Tools as a data source for actions. See #276.
Saved synths which use removed actions should now show a ? icon instead of just an empty slot.
I was thinking that now the game shows the values of progress at 100% perhaps that would be useful to include on the initial display as a double check that the base calcs are right.
I'm assuming that work needs to be done to determine the 71+ scaling levels is there anything we can do to assist determining those values?
I was thinking the panel could look like this:
So it makes it easy to compare simulation with game and detect errors.
Indeed it would be possible to check stage by stage even complex recipes and find bugs that previously were hard to track down by simply comparing simulator numbers with in game numbers.
A dedicated data collector could even tweak gear (using melds) by small amounts to accurately quantify impacts at different player and recipe levels.
Another option would be a way to override the quality/progress calculation by inputting the values the game gives you for a more accurate solve, until we have figured out the algorithm properly
Yes but the issue with that is the numbers can and do change after certain skills are used - eg: when you use a skill that buffs the quality. So changing the starting progress/quality numbers whilst potentially useful for debugging doesn't help understand how the figures are derived by the game and how the skills used affect it because you'd not be able to change inaccurate values mid way through a skill sequence.
This does give us a non-destructive way to collect data from the game about the relationships between the various attributes, levels, etc. On Jul 11, 2019, 7:29 AM -0400, Alexander Bisset [email protected], wrote:
Yes but the issue with that is the numbers can and do change after certain skills are used - eg: when you use a skill that buffs the quality. So changing the starting progress/quality numbers whilst potentially useful for debugging doesn't help understand how the figures are derived by the game and how the skills used affect it because you'd not be able to change inaccurate values mid way through a skill sequence. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
@ShammyLevva That is already handled, the most inaccurate thing about the solver right now is the base progress/quality calculation
@gr3ger yup hence my suggestion that with this info on the page and the info in game it would be possible to do a data collection exercise and actively test different functions solver vs game to see if there was an identifiable pattern that could lead to fixes and more accuracy.
It might also allow better reporting by users if they can say I did X and it indicated Y but in game it said Z.
To do that we will need to know the "effective crafter level" for levels 71-80. This is the current table:
51: 120,
52: 125,
53: 130,
54: 133,
55: 136,
56: 139,
57: 142,
58: 145,
59: 148,
60: 150,
61: 260,
62: 265,
63: 270,
64: 273,
65: 276,
66: 279,
67: 282,
68: 285,
69: 288,
70: 290
With the values for 71-80, I should be able to calculate the base progress & quality amounts using our current formula a display those like you suggested.
Happy to help - how do we go about determining those figures? They look a bit like the average gear level for an item at that level?
One way is to take the existing base progress/quality formula and using an actual progress/quality value from the game plug those values along with craftsmanship/control into the formula to determine the crafter to recipe level difference that would result in that value. Basically, reformulate the equation such that level difference is the output instead of the base progress/quality value.
That only applies if the coefficients for the current tier are correct for the new level range. If not, then we'll need a bunch of progress/quality values with corresponding crafter level, recipe level, craftsmanship and control, and apply the Excel solver (😄) to the data to determine the effective crafter levels and coefficients. This requires a broad set of data from different crafters to cover enough variation to achieve a good fit.
Note that the new actions have placeholder icons and are not actually distinguishable yet. I need to add simple name tooltips to replace the XIVDB tooltips.
I don't really get how you want us to fill in the table. Don't we need: crafter level, craftmanship, control, level of recipe and the in-game 100% values to be able to determine the correct values?
Yes, see my description above. On Jul 12, 2019, 3:09 AM -0400, Jonas Henriksson [email protected], wrote:
I don't really get how you want us to fill in the table. Don't we need: crafter level, craftmanship, control and level of recipe to be able to determine the correct values? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
Given the gearing I suspect the following values: 71 is 330, 74 is 370, 77 is 400, 80 is 430.
These all seem consistent with previous levels. Interpolating thus gives... 71=330, 72=343, 73=357, 74=370, 75=380, 76=390, 77=400, 78=410, 79=420, 80=430.
I've stuck that in a pull request so you can import them. I'm not sure what Ing1RecipeLevelTable & Ing2RecipeLevelTable values are for though?
If you can add the display values for 100% progress and 100% quality to the beta site as suggested above. We can then compare to see if the projected values from this interpolated values match in game.
The beta site now shows 100% efficiency values for progress and quality. These values will obviously be wrong for levels 71-80.
I've created a branch (dev/272-level-71-80-model) with the proposed effective crafter levels, which you should be able to try out with a local copy of the site.
We have a new set of formulas for progress and quality increase. These changes are available on the beta site (ffxiv-beta.lokyst.net). Please let me know if these discrepancies are resolved.