Announcing the FFXIV Crafting Optimizer

FFXIV Crafting Optimizer

This is a project we have been working on for a while now.

The aim behind the project was to create a FFXIV crafting simulator that:

• Does not require an executable download.
• Is usable by crafters and recipes below max level.
• Allows me to play with genetic algorithms.

Things still left to do:

• Include recipe data for recipes introduced in patch 2.2. In the meantime, recipe information can be entered manually.
• Update the behaviour of Ingenuity and Ingenuity II to reflect patch 2.2 changes.

It is in BETA, but I have been using it for a while successfully. Please check it out and have fun.

Thanks alot ...
I hope you'll update it to 2.2 so it become more useful

Is this using the pre-2.2 formulas for how ingenuity 1 and 2 behave? SE said they behave differently for i90 recipes in the 2.2 patch notes.

Edit: Even weirder than I expected: the solution it gave me for an i90 recipe didnt use either ingenuity 1 or 2!

Yeah, I dont think that Ingenuity considerations are in the system yet.

Um, I can say this much. With 392 craftsmanship, on an 3 star recipe, my Careful Synthesis 2 does 40 without any buffs and 70 with Ingenuity 2. Also, with 416 craftsmanship and Ing2, CS2 does 75... I think, maybe 74.

From another thread:
423 craftsmanship, Ingenuity II, , still only hitting for 95 with Standard Synthesis and 76 with Careful Synthesis II on the 3 stars. PbP hits for 146 on the first use and 98 on the second.

The big thing with the 3 star recipes is getting the progress component done... But in a reliable way.
Although you can just do Rapid Synthesis a few times, the max 80% chance really make it... unappealing.
Though I guess there are things like SH2 PbPx2 Rapid Synth, which would bring it down to 48 progress, then a non-ingenuity Standard Synth would finish it off... But you would have to do that at the end, under Ingenuity2... which wont leave you with enough CP for Reclaim if Rapid fails.
Or you have to use Ing2 twice, which is a sizeable 32 CP...

Yeah, those 3 stars are gonna be tough.
I am just thinking from a coding perspective, you would need to include a section that has 2 Piece by Piece actions under Steady Hand 1/2. And then a section which has 2-3 CS2 or similar under Ingenuity 2 in order to ensure that we actually finish the synth.
Additionally, you will also need to put Byregot's under Ingenuity 2, since the penalty for using it without Ing2 is REALLY steep.

Quote:

---

Originally Posted by Kenji1134

Yeah, I dont think that Ingenuity considerations are in the system yet.

Um, I can say this much. With 392 craftsmanship, on an 3 star recipe, my Careful Synthesis 2 does 40 without any buffs and 70 with Ingenuity 2. Also, with 416 craftsmanship and Ing2, CS2 does 75... I think, maybe 74.

From another thread:
423 craftsmanship, Ingenuity II, , still only hitting for 95 with Standard Synthesis and 76 with Careful Synthesis II on the 3 stars. PbP hits for 146 on the first use and 98 on the second.

---

Thanks for the information. It looks like the model completely falls apart on these new recipes so I'll be making adjustments to the formula. Unfortunately, I don't have my own max level crafter yet (too distracted making toys) so keep those data points coming! :)

We have pushed a bugfix for the freezing issue when running the solver on an empty sequence and made some quality of life changes to the UI. Thanks to Kenji, we have also made a first pass at accounting for 2 and 3 star items and the patch 2.2 changes to Ingenuity and Ingenuity II.

Please send more data on quality and progress gains with Ingenuity I & II so that we can refine the formula. Especially if what you're getting is different from what the site returns. Don't give up on us so easily ;)

Quote:

---

Originally Posted by Sixtail

The aim behind the project was to create a FFXIV crafting simulator that:

• Allows me to play with genetic algorithms.

---

You can mess with genetic algorithms however you like as long as you don't turn Eorzea into a zombie wasteland.

Great work! I tried doing one myself but I could never get the math right. Then this got released which was so incredibly accurate that I lost any motivation to make my own. I think the author has stopped playing through.

Quote:

---

Originally Posted by QuCont

You can mess with genetic algorithms however you like as long as you don't turn Eorzea into a zombie wasteland.

Great work! I tried doing one myself but I could never get the math right. Then this got released which was so incredibly accurate that I lost any motivation to make my own. I think the author has stopped playing through.

---

But ... but ... my dreams of being an evil zombie overlord ... T_T

Yes, I have seen that Japanese site as well and used it cross check some of my results. We use slightly different methodologies though.

Now touch actions reduce quality, so the simulator spits out a rotation without any.

Quote:

---

Originally Posted by RichardWagner

Now touch actions reduce quality, so the simulator spits out a rotation without any.

---

Could you please provide the crafter and recipe stat information that you put into the model? I suspect quality is not being modeled correctly on 3* items.

Unfortunately, I can't get the site to load on Opera browser. It may be my add ons. Has it worked in this browser for anyone else?

Quote:

---

Originally Posted by Sadana

Unfortunately, I can't get the site to load on Opera browser. It may be my add ons. Has it worked in this browser for anyone else?

---

Well, I ran it on a browser compatibility tool and it seems it only works with Opera version > 12.0. What version of Opera do you have?

Quote:

---

Originally Posted by Sixtail

Well, I ran it on a browser compatibility tool and it seems it only works with Opera version > 12.0. What version of Opera do you have?

---

12.16
I'm getting an error message that I can't even access the page, I should say. So perhaps it is one of my add ons.

how do you set this "solve" function to attempt to get quality up not just solve the recipe.

Quote:

---

Originally Posted by Cantontai

how do you set this "solve" function to attempt to get quality up not just solve the recipe.

---

The solver's objective is to maximize quality, but it is using a method that relies on testing random changes to the sequence and there are no guarantees. If you are not seeing a change in the quality/sequence, it is possible that just hitting solve a couple of times will find a different solution, but it is also possible that you have found the best solution already.

First off ... Thank you for your work.

Please don't take the following wrong ... I'm hoping to highlight some bugs or opportunities.

I've been playing around with the optimizer and gotten some interesting results. When solving multiple times to get a better HQ rating, I noticed you do not always get a better rating than the previous. I'm guessing that is what the previous comment regarding a setting for the highest HQ rating was pointing toward.

I've also run into multiple instances of a solution with a strange action tacked onto the end.

Ex: MM II tacked on the end.

Macro #1
*snip (appropriate block of commands)
/echo Macro step 1 complete <se.1>

Macro #2
/ac "Master's Mend II" <me> <wait.3>
/echo Macro step 2 complete <se.1>

I would just like to clarify how you're solving repeatedly, because the solver should not return a worse (in terms of quality) result than that provided in the initial guess if a valid solution is supplied.

The correct way to go about repeat solving is to click the solver button and let the solver do its business. If the solution is an improvement over the initial guess, click the "use solver result" button to replace the initial guess, then click the solve button again. Rinse and repeat.

Is this what you mean by repeated solving, or did you just hit the solve button repeatedly and sometimes the result returned was worse than a prior result?

Quote:

---

I've also run into multiple instances of a solution with a strange action tacked onto the end.

---

This is a known "feature" of GA. Essentially, if there is no change to the quality of the result, the algorithm can't tell if the solution is better or worse than another slightly different solution so it leaves in some actions that have no effect on the overall outcome.

I'm at work so cannot store an external Screenshot to link.

Example:

Alchemist
LVL 40
Craft: 214
Control: 217
CP: 289

Recipe
Smelling Salts

Below is Copy/Paste Results

Initial Guess Q:1,797 Prog:169/138 CP:1
Required Cross Class Actions
Solver Result Q:1,354 Prog:142/138 CP:5

I am still leveling all my DoH classes so perhaps I'm not understanding some skills yet. ^^; I entered my stats and let it solve for that bane of my crafting universe, Walnut Lumber. Why did it give me two Steady Hands back to back? My current method is something like: IQ, SH, (BTx2, ST on a good), MM, HT if SH still up (ST if good, then some touch skill to max quality and CSynth)... well at this point it really depends... if I've had good luck with the 'goods', I'll use touches to get max quality; if bad luck I'll use Rumination for another MM refill and just toss caution to the wind with a few HT.

/ac "Great Strides" <me> <wait.3>
/ac "Steady Hand" <me> <wait.3>
/ac "Inner Quiet" <me> <wait.3>
/ac "Standard Touch" <me> <wait.3>
/ac "Great Strides" <me> <wait.3>
/ac "Standard Touch" <me> <wait.3>
/ac "Great Strides" <me> <wait.3>
/ac "Steady Hand" <me> <wait.3>
/ac "Steady Hand" <me> <wait.3>
/ac "Standard Touch" <me> <wait.3>
/ac "Careful Synthesis" <me> <wait.3>
/echo Macro step 1 complete <se.1>

I did activate all cross class skills available to me. I have all DoH classes between levels 32-35.

Quote:

---

Originally Posted by Thum

I'm at work so cannot store an external Screenshot to link.

Example:

Alchemist
LVL 40
Craft: 214
Control: 217
CP: 289

Recipe
Smelling Salts

Below is Copy/Paste Results

Initial Guess Q:1,797 Prog:169/138 CP:1
Required Cross Class Actions
Solver Result Q:1,354 Prog:142/138 CP:5

---

Thank you for clarifying, we will look into it since that is not supposed to happen.

Quote:

---

Originally Posted by Sadana

Why did it give me two Steady Hands back to back?

---

Glad to see you got it to work in Opera.

This is related to the second half of my reply to Thum.

Quote:

---

Originally Posted by Sixtail

Glad to see you got it to work in Opera.

This is related to the second half of my reply to Thum.

---

Sorry, I have to use it in Firefox. ^^;; Still no luck with Opera.

I will re-read that comment, then. Thank you.

Sorry about the terse replies. I'm not very good at replying on my iphone :x

Quote:

---

Originally Posted by Thum

I'm at work so cannot store an external Screenshot to link.

Example:

Alchemist
LVL 40
Craft: 214
Control: 217
CP: 289

Recipe
Smelling Salts

Below is Copy/Paste Results

Initial Guess Q:1,797 Prog:169/138 CP:1
Required Cross Class Actions
Solver Result Q:1,354 Prog:142/138 CP:5

---

We accidentally fixed it and are unable to replicate the problem. If you do still have that screenshot it would be greatly appreciated.

This simulator is not accurate. The Quality/Control as well as the Progress/Craftsmanship formulas are not calculating correctly.

I plugged in a rotation for a Rose Gold Cog, double checking every stat and number. Your calculator says it completes with 100% HQ, but in the game, I am left with 84% HQ and I need one extra Careful Synthesis II.

Please fix this.

First off, awesome tool!

It isn't quite accurate from a craftsmanship/progress POV, though. The difference in numbers is very small (1 at most in the 10 recipes I tried), but it was enough to cause two synths to fail for me. Here is a very basic spreadsheet with the data from testing a few CUL recipes.

In other news, I submitted a GitHub issue for ya. You can contact me there or PM me if you want more data. I have 3 DoH in the 20-30 range right now.

Not sure if I'm using this right!

Genetic Algorithm Result
========================
# Action DUR CP EQUA EPRG WAC
0 40 345 0.0 0.0 0
1 Inner Quiet 40 327 0.0 0.0 0
2 Great Strides 40 295 0.0 0.0 0
3 Ingenuity 40 271 0.0 0.0 0
4 Advanced Touch 30 223 476.1 0.0 0
5 Great Strides 30 191 476.1 0.0 0
6 Advanced Touch 20 143 1021.5 0.0 0
7 Great Strides 20 111 1021.5 0.0 0
8 Ingenuity 20 87 1021.5 0.0 0
9 Steady Hand 20 65 1021.5 0.0 0
10 Advanced Touch 10 17 1706.5 0.0 0
11 Standard Synthesis 0 2 1706.5 114.0 0
Progress Check: true, Durability Check: true, CP Check: true, Tricks Check: true, Reliability Check: true, Cross Class Skills: 1, Wasted Actions: 0

It says this is 100% reliable, yet Advanced Touch failed because there was no Steady Hand to correct it to 100% chance of success (from 90%). Pretty sure those steps won't get very far regardless. Could be wrong though!

Great idea for a tool though, I hope you get it to 100% HQ ;)

Hi. First off, thanks for this tool, it will be very useful!

Secondly, a stupid question and perhaps i've missed a post on it somewhere , but does the simulation assume hq or nq base materials?

Quote:

---

Originally Posted by Divine_Intervention

... does the simulation assume hq or nq base materials?

---

The simulation assumes you entered a "Start Quality" in the Synth section. If you didn't enter one, it uses 0 initial quality, which is the same as NQ mats.

FYI for lower leveled recipes it often suggest using the same action twice in a row due to lack of CP (I think).

Example, on a steel elmo it suggested I use steady hand, then steady hand again, for no reason other than to use as much CP as possible during the craft.

Quote:

---

Originally Posted by Loony_BoB

...It says this is 100% reliable, yet Advanced Touch failed because there was no Steady Hand to correct it to 100% chance of success (from 90%). Pretty sure those steps won't get very far regardless. Could be wrong though!...

---

Quote:

---

if (progressGain > 0) {
reliability = reliability * successProbability;
}

---

From the code, it looks like reliability refers to the success of progress-increasing actions only. It starts as 100% and is modified by the success chance of any progress-increasing action taken. Quality steps are tricky due to the variability of condition (e.g. Poor, Normal, Good, and Excellent), which is likely why they're not used to calculate the reliability.