All Jobs Damage Simulator And Gear Sets

Finally had time to look at the code a little. If you want to go further down the math nerd rabbit hole, you might want to check out a few things:
1. Look up simulated annealing. You could make it work for you and exploration would likely be more efficient than your current method.
2. You should generally move towards dealing with distributions instead of sampling all this stuff like pdif, crits, double attack, etc. If you work with density functions, you can ultimately make one giant damage density function for the set and truncate it at 99999. This will remove all variability from your estimation and give you a far more complete picture of the damage than an estimation of an expectation. Python must have a library for this.

Don't suppose you could add raging fists to the monk WS sets?

Hmm GEO debuffs do not work very well in any content where the words "optimize damage" are important, that means frailty is generally a non-starter and stacking lots of PDL isn't very useful. SV songs, chaos and fury means we get benefit out of some PDL, but the monster isn't going to have non-existent defense from Dia II + Idris Frailty.

Isn't nerfed bolster frailty still like 20% def down? on top of dia2 + light shot and possibly a break WS, that's still pretty considerable defense down in these situations on top of like you said very high attack buffs.

Don't we know what most Ody goal NM defense is by now? thought the jpn players recorded a lot of their stats.

Not sure exactly what scenario and what buffs are considered to come up with these sets but as a ml40 run/drk my resolution set in sortie boss G was doing much better than the two stated in this topic (50-60k damage vs low 40k)

Accessories are almost the same, the main difference is that I was using adhemar head, body and gloves, meghanada chausses, lustratio leggings and both fotia.

We do sortie with no geo so our atk buffs are SV MINx3, chaos roll and dia 3.

I can't comment on this properly since I've only done some preliminary reading, but it seems like the SA optimization method would be both slower and less accurate than my brute-force swapping two items at a time method for finding the best sets. However, SA optimization seems like it might allow me to find the best conditional sets with an arbitrary number of conditions relatively quickly. I think it depends on if a neighbor state can include a set with 14 swaps at once, or if that is too much of a difference. The algorithm does not seem difficult, so I think it's worth trying out. I've still got a lot of reading to do for this, though.


Raging Fists is already available. The weapon skill list is based on what main/ranged weapon you have equipped.



I actually don't understand what you're trying to say. I used Dia2 (-15.23%) and a 20% BoG Frailty bubble (-12.54%) in my "high-buff" sets. I'm definitely not using high enough enemy defense (or VIT) to be comparable to v25 bosses, but I'm almost also* not including def- weapon skills like ByneBill mentioned and I'm not using capped Songs+, Rolls+, or even Soul Voice.



I've double checked and found that I did not apply the -15% attack penalty to Resolution (I'll update the GitHub code with this change later). However, even after this correction, the only change is swapping Adhemar path A to path B to make up some of the attack.

It doesn't help that your set is not capped accuracy in the situation I checked for ML20 RUN, unless I use Adhemar Path A. Even if I set enemy_evasion=1, then your set with Adhemar Path B is still ~4.4% behind what I've posted for the high-buff situation. This is a relatively small difference compared to the 20~30% that you're seeing in game.

What is your sample size for each set? I imagine you only have a few dozen samples total rather than the thousands required to make a statistical comparison. You're seeing such a huge difference that either my code is very wrong, or you just happen to notice the hits with more multi-attacks. Resolution is an 5-hit FTP transfering weapon skill, so each extra attack is worth around 20% more damage.

The numbers you're seeing line up with what I expect for my set not getting any MA (~40k damage) and your set getting a DA (~48k damage), a TA (~56k damage), or even a QA (~64k damage).

Not criticism towards you, good product. Basically the code is great, the "example" High sets are being passed around as BiS without context. I've had to have this conversation with three different folks, who should of known better.

It's the same issue spreadsheets had, setting everything's defense to 1 or so low as to be trivial to massively over-cap pDif, which favors a very different gearset that when not massively over-capping. Not your fault as you mentioned they are just examples and people need to adjust based on their situation.

I assume a lot of people don't fully read the text so they miss the bolded statement about these sets being references and not absolute best in slot sets. I do trust the mid-buff sets presented in the main post to be very good for most situations, but I wouldn't recommend using the high-buff sets unless you are sure that you're at or very near attack cap. Naturally, I do recommend downloading the code and playing with it to build better sets than what I presented in the main post.

While we're discussing it:
Many of the high-buff sets I posted seem to be 10~20% over attack cap, which I think is not necessarily a bad thing since it leaves room for Dia2 to fall off or, for example, a worse Chaos Roll (not a Lucky roll or not Crooked).

The RUN Resolution set I posted earlier today shows 4927 attack vs the enemy's 1500*0.7223 = 1083 defense, which is 14.5% over capped with the 6% PDL from gear. Without Dia2, the enemy would have 1312 defense, which corresponds to an attack cap of 5214 for a RUN using a 2-handed weapon. This puts the RUN 6% under cap and results in a gear set mix between the mid/high buff sets I posted.

Edit: Now that I think about it, I don't think the output distribution plot reflects the player's attack bonus/penalty from certain weapon skills. So the RUN Resolution set being 14.5% over cap is actually (coincidentally?) almost exactly at attack cap since Resolution has a -15% attack penalty. Either way, the other sets that I quickly checked show similar "issues" where they are 10~20% over cap.

Something I see weird is Decimation and Mid vs High, both sets should be Sakpata even under attack cap with Sangarias +1 off hand. Decimation has no TP growth and divides all it's damage across all 4+ hits, we wend up gearing for Multi-Attack / Attack with some PDL thrown in for good measure. WSD would have almost no impact on it, so maybe it's calculating it as a single hit or something?

Not to say it isnt wrong but the mid set looks to be highly valuing atk/stat mod such as boi giving an extra 31~ via axe skill and nyame body/legs str+ final aug, the DA nyame also end up with slightly more atk assuming R30 vs the sakpata with minimal DA difference, ikenga looks to be the same, 75 atk, and a nice STR chunk vs the additional MA

first of all let me say I'm not an expert at all on WS formulas so my intention was not to challenge your findings but rather bring my anecdotal experience to better understand how it works.

For instance until yesterday I thought that in a multi-hit ws any hit has a chance of proccing a multi-attack so I thought that with enough DA and TA Resolution would consistently do 7-8 hits but it's not the case. After looking at your distribution and doing some research now I know that only the first 2 hits of a ws can proc a multi-attack (not sure if this is 100% confirmed).

Knowing that, now I understand that the high numbers I was seeing were due to some high TA proc since the set I was using has a lot more TA than the one you listed in the first post.

I don't have any data at all since again it's anecdotal evidence, however I could really feel the difference in damage over several runs. So this begs the question, if we are atk capped and we have enough STR, isn't it a good idea to stack as much multi attack chance on Resolution so that you can consistently do 6-7 hits?

If I try to max out multi attack while keeping moonshade earring and both Fotia, I would get: 50% DA (21 from gear and 29 from temper - 590 enhancing skill with max gear at ML 40), 22% TA and 6% QA (3% QA from herc boots). If only the first 2 hits of a ws have a chance to multi-attack is confirmed, according to my rusty knowledge on calculating probability and a home-made excel sheet, I would have a probability of around 75% of hitting at least 7 hits. Does that sound correct?

Regarding accuracy, I didn't mention it because I thought it was implied that we have enough buffs to cap accuracy in sortie (SV blade madrigal, HM and distract).

It's been confirmed long ago. Dual wielding follows the same principal as well, where only the first hit of the main hand and the first hit of the offhand can proc a multi attack.

The problem here is multi-fold:
-His spreadsheet isn't considering a QA DM aug on herc boots
-You are adding all the Multi-attack together and comparing that all to the stats on a single piece (75% chance to multi-attack vs 10 more STR? I'll take the MA obviously!) where you should compare piece-by-piece. Is adding another 4% DA better than adding 10 STR and 35 ATK? You still have 46% DA, and a 74% chance to hit 7 times (just making up stats here), but you now have extra STR and ATK for every one of those WS, including the ones with the MA procs.

That said, it's still possible that your set is better than his, IDK, but it's impossible to tell from feeling. Using extremely small sample sizes (10-20 WS) is not a good method to determine the proper set.

I see no issues with either set. Farsha shows up off-hand for the high-buff set now (beating Sangarius +1 R15 by <1%) since I've added Empyrean off-hands a while ago. Nariont may be correct that the code prefers to boost attack and WSC when underbuffed.

I'm finding the 5/5 Sakpata with Sangarius to be 8.5% worse than the midbuff set I have posted for WAR/NIN Decimation.

STR = 409, Attack = [4062, 3750] (posted set)
STR = 383, Attack = [3915, 3551] (5/5 Sakpata with Sangarius)


I think it's better if people do challenge the sets I post. Otherwise, the main post is potentially stuck showing incorrect sets. While I did write all of the code myself, it's been more of a community effort to make sure everything is correct. A lot the code is now correct because I previously posted incorrect results to a few job specific forums and the experts who play those jobs tell me that something looks weird. Your recent post helped me find that Resolution was missing its 15% attack penalty and Rune Fencer as a whole was missing DA+ from Fencer (adding 27% DA from Fencer did not change the relative difference between our sets or make significant changes to the sets I had already posted).

There could be a much larger issue with the logic in the code to produce the ~20% damage difference you're seeing, but I'll have to look into things in much more detail to find such an issue. I think it's much more likely that you're seeing a difference in MA procs, though.


Assuming my code is correct, then I think it comes down to how much you're giving up to boost your multi-attack rate.

My set has STR=410, DEX=283 (at ML20), while your set has STR=339, DEX=272 (assuming no STR/DEX on Herc and all Path B Adhemar). This is a difference of 71 STR and 11 DEX, which is 61 WSC total (85% STR and 10% DEX). My code also finds a fSTR=31.25 for my set and 13.5 for your set, assuming enemy_VIT=289. The Sortie G boss likely has much higher VIT which would lower both of these values and favor more STR gear even more. Together, this is about an 10% difference in base damage favoring the set my code finds, or ~8% difference if your set is using Lustratio Leggings +1D instead of Herc boots.

I don't think this following situation has a strong effect, but the difference between our sets was only ~5%, so this may play a small role in that:
Resolution is already a 5-hit weapon skill. If you were to proc a single quad attack, then you will not gain any benefit from another MA proc since you'd already be at the 8-hit limit. Similarly for getting a double-TA proc; the second TA will only be worth half-potency. Multi-attacks are still incredibly strong, but TA/QA are worth slightly less than usual here due to the 8-hit limit. Double Attack procs are not affected by this 8-hit limit (unless your first hit was a QA). Nyame has a good bit of DA in addition to its high base stats and attack.

I added my pair of "Accuracy+35, Attack+18, QA+3" herc boots to the code and tested them, but the code does not include them in the best set. This might be related to what I said earlier where a single QA invalidates the next MA proc due to the 8-hit limit. They would likely need a lot of STR to beat Lustratio Leggings +1D, which have 47 STR, 48 DEX (+45 WSC)

Something is up because R30 Sakpata generally has more attack, STR and multi-attack then R25 Nyame. Boii +3 head is better then Sakpata but Boii +3 hands has less attack, less strength and no multiattack. It's almost like augments on Sakpata aren't being counted.

Sangarias +1 isn't for the WS average, you can't change offhands during WS. Decimation has zero damage growth with TP, 1K or 3K it's the same, you want to get to spam at 1K and get TP as fast as possible. Sang +1 just provides a massive multi-attack boost for TP gain that's all.

Thanks for the note. I'll look into this in more detail later today. It may be related to potential issues with RUN Resolution as well.

Edit: Yeah, I'm seeing Sakpata Breastplate R30 beating Nyame Mail R25B in all relevant stats (+2 STR, +10 attack, +3 DA), but still losing to Nyame in total damage dealt (29101 vs 29215 for the midbuff set posted). The only advantage Nyame is bringing is +12% WSD, which should only apply to the first hit.

Boii earring +1 says it has 9DA when it should be 8DA

There may be something relating to the resolution issue with h2h ws, I think

Hopefully someone can correct me but generally speaking people always favored MA over WSD, yet most sets encourage nyame or bhikku body, but the latter I'm not sure if it's because of impetus or it's taking that 12% WSD. also bhikku hands which afaik only really bring pdl to the table?

idk someone call me out on this pls

depends on the WS, but DK/TK for example, I get better results with nyame pieces over ken pieces.

I've looked into it a bit more, it seems that 12% WSD (on both Nyame body R25B and Boii Mufflers +3) are a small amount better than what Sakpata brings for a job that already has a ton of DA. It's a 3-hit weapon skill, so 12% WSD is roughly +2-3% damage by itself.


Thanks for pointing this out. I'll update the code on GitHub in a few minutes.

Surprisingly, all of the JSE+1 earrings were using the +2 base stats. I'm not sure why, but I may have simply overwrote the correct version of gear.py before uploading it with +1 earrings for the first time. More likely is that I just didn't update the copy/pasted +2 base stats while focusing on updating the augments.


I've checked the code while looking into the potential Resolution/Decimation issues and found no obvious issues. Not to say that there isn't an issue.

I'm not sure about Bhikku Cyclas +3, but Bhikku Gloves +3 bring "Hand-to-Hand Skill +21," which is equivalent to some small boost to weapon damage for hand-to-hand attacks. BGWiki claims "base damage raises every 9 Hand-to-Hand skill points," so +21 skill is providing attack, accuracy, and +2~3 base damage.

The code uses average damage estimates to find the "best" sets, but simulates the damage properly when the best set is found (to build the final distribution plot). Both algorithms arrive at nearly the same average value, so it'd have to be an issue with both algorithms.

Are categorical haste caps applied? DRG/WAR should be stuck at 78.75% but the sim has them reaching 80%.

it's actually floor(skill * 0.11), i verified when i wrote my simulation. comes out the same as skill / 9 in most cases, but it can be off by 1 for some cases.

Should be. Check lines 19-22 of get_delay_timing.py to see the caps I've used and how I have each category affect delay timing:

• Gear Haste Cap: 256/1024

• Magic Haste Cap: 448/1024

• JA Haste Cap: 256/1024

I just ran a Trishula DRG/WAR TP set with capped magic haste and it returns 78.5% delay reduction:

Note that the cap on magic haste is not reflected in that 58.7%. I suppose I should fix this, but the tooltip does mention the cap if you mouse-over the value.

If you /dnc you'll have an assumed +5% JA haste from Haste Samba, which covers the remaining 1.5% to reach the 80% cap. Or, if you're dual wielding, you'll reduce your delay a bit more through dual wield.

Edit: I suppose using /SAM will also provide +10% JA haste from an assumed Hasso, but the +9.8% JA haste in the set I ran earlier was for DRG/WAR with a max level Wyvern out.

Edit2: Do you have an example set that is showing 80% delay reduction with DRG/WAR? It might be a weird edge case.


Thanks, looks like I already apply it that way on line 113 of wsdist.py

Would be awesome to have ML input on the Basic Input selection screen

I agree. The only reason I haven't added this is because I'd need to know exactly how much of each stat that each subjob provides at all MLs. I only know the values for ML20-24 with Lv53 subjobs for Hume.


If there is a good resource for finding a table of subjob stats, then I think I could make Master Level selection a simple drop down menu. The traits would be pretty annoying, but they'd just be a bunch of if-statements anyway.

Yeah, there's definitely exceptions in some of these WS, but since there hasn't really been a collection of sets posted since those that ram and gang did back in the omen days (and those were mainly kenda/MA) and those ones from the bgwiki guide that hasn't had any major updates since aug2022, then i figure i'd rather ask than make any assumptions

fair enough, it's definitely one of those situations where i'm not knowledgeable enough to come at an obvious conclusion as to why the hands would exceed the usual comparisons (ryuo hands, kenda hands, malignance hands, etc).

i appreciate you checking anyway

Super appreciate the effort being put into this. FFXI makes it difficult to know when a close set is better than another, and going by "feel" is what I've had to deal with for years. I'd love to try this out later after you've ironed out the little details.

And yes, as some said, the high sets in the OP are not "best in slot" for every scenario, but only for the stated scenario. I would love to see other scenarios shared later, like for example, sets for trust buff situations, or when you're fighting some of the hardest enemies in the game. But I know that in the short term, the important thing is getting the numbers right in the code, and I appreciate anyone giving feedback to help improve this.

Can I also suggest adding R15 Dominance Earring +1? +10 DEX and +15 Acc is pretty nice.

Thanks for sharing you are awesome !