The Sealed Dagger: A Ninja Guide
The Sealed Dagger: A Ninja Guide
Is the nin guide up to date?
No, check the other two gear guides.
The 2021 guide i just found ty. But whats the other one?
[+]
Do keep an eye on this topic when it pops up, usually has discussion about other aspects of ninja strats that are useful, aside from specific BiS gear builds. Good place to ask about midgrade gear questions also, while you are working your way up to BiS sets.
[+]
I am looking for suggestions on my Kamu set
This would be for underbuffed/mid buff situations. Also, why do I see so much variation in this particular WS? Capping currently at like 15k with trusts, with dips in 2-5k range.
Also, curious about the AM3 set, assuming its similar to the tizona one for BLU? Something like this?
Code
sets.precast.WS['Blade: Kamu'] = {ammo="Seething Bomblet +1",
head="Hachiya Hatsu. +3",neck="Fotia Gorget",ear1="Lugra Earring +1",ear2="Brutal Earring",
body="Tatena. Harama. +1",hands="Adhemar Wrist. +1",ring1="Gere Ring",ring2="Regal Ring",
back=gear.hybrid_jse_back,waist="Sailfi Belt +1",legs="Mochi. Hakama +3",feet="Mochi. Kyahan +3"}This would be for underbuffed/mid buff situations. Also, why do I see so much variation in this particular WS? Capping currently at like 15k with trusts, with dips in 2-5k range.
Also, curious about the AM3 set, assuming its similar to the tizona one for BLU? Something like this?
Code
sets.engaged.AM = {ammo="Seki Shuriken",
head="Malignance Chapeau",neck="Ninja Nodowa +2",ear1="Telos Earring",ear2="Dedition Earring",
body="Malignance Tabard",hands="Malignance Gloves",ring1="Chirich Ring +1",ring2="Chirich Ring +1",
back=gear.tp_jse_back,waist="Kentarch Belt +1",legs="Malignance Tights",feet="Malignance Boots"}
Missing mainhit lows and multi procing highs, watch your tp returns
Asura.Aller said: »
I am looking for suggestions on my Kamu set
Based on the sets you posted, I assume you do not have access to max-augmented Odyssey gear yet. You mention AM3 sets so I'll assume you have an R15 Nagi for main-hand as well R0 Odyssey gear in general. If you do have access to R20 Odyssey gear, then check out the sets I present in my thread. Running my Python3 code with the above restrictions, the best sets I'm finding are:
Low Buff (Dia2 + Grape Daifuku):
Midbuff (Dia2 + Grape Daifuku + COR + BRD):
Both sets are about the same in terms of damage, but it looks like the low buff set barely does not cap attack against the target I used, so it does not fully benefit from swapping in PDL ammo and hands. Since Blade: Kamu provides an effective 3.0 attack multiplier, it is very easy to cap attack, especially with Dia2 or Dia3. I recommend just sticking with the attack capped set here until you get R20 Nyame and Mpaca, then swapping to the set I present in my Python3 thread.
This was a decent consistency check for my code when comparing my output against the Blade: Kamu sets presented in the pre-Odyssey 2021 Ninja Gearsets thread. It's nice to see them agree as well as they do.
I have not started working on TP sets yet, but your proposed STP-stacking set seems fine to me. I'd swap out the rings for some multi-attack, but somebody else is likely more qualified to talk about TP sets right now. I may work on TP sets soon, but I am more curious about DRK TP sets at the moment.
Asura.Aller said: »
Also, why do I see so much variation in this particular WS? Capping currently at like 15k with trusts, with dips in 2-5k range.
Blade: Kamu uses an FTP damage multiplier of 1.0 (without multi-hit replication). Additional hits (from off-hand and multi-attack hits) happen to also use FTP=1.0. Since Blade: Kamu is a single hit weapon skill with an off-hand hit for two total hits, missing one of these hits will roughly half your damage. Obtaining a single double attack is +~50% damage, a single triple attack is +~100% damage, one double and one triple attack is +~150% damage, and two triple attacks is +~200% damage. Thus the damage distribution is incredibly multi-modal as we see in the above set distribution images I posted.
Edit: It's interesting and instructive to add a modified distribution image where we label each peak with the scenarios that contribute to them. Each weapon skill can multi-attack up to two times (if it has two native hits, or one native hit and an off-hand hit), so we just need to label the combination of two multi-attack procs that add up to the number of hits total for each peak.
[+]
Thanks and I have been using your thread plus a combo of the 2021 gearsets/logicals video to try and come up with best sets, as you mentioned I do not have R20 nyame gear yet.
If someone has a better AM3 set I'd still like to see that, too!
If someone has a better AM3 set I'd still like to see that, too!
Since nobody is commenting on TP sets, I've made some quick and messy modifications to my code to estimate TP sets based solely on the time it takes to reach a TP threshold and use a WS. I locked the weapons in place based on the off-hand weapons seen in my WS testing.
I should note that I needed to adjust the code to simultaneously swap up to three pieces in order for the final sets to become independent of the initial sets at low haste values due to the variety of dual wield options Ninja has.
See below for my preliminary results. I won't claim these are the best sets, or even necessarily correct sets, since there could be some simple mistake in my code. Additionally, a simple time-to-weapon skill may not be a robust enough method to find best TP sets, but I'm hesitant to use 'ideal DPS' as a measure of performance. I may end up using average DPS in the end anyway... At the very least, these sets are something to think about.
NIN/WAR, Heishi+Kunimitsu, 0% Haste, Low Accuracy (1133 eva):
NIN/WAR, Nagi+Gleti AM3, 0% Haste, Low Accuracy (1133 eva):
NIN/WAR, Heishi+Kunimitsu, 15% Haste, Low Accuracy (1133 eva):
NIN/WAR, Heishi+Kunimitsu, 30% Haste, Low Accuracy (1133 eva):
NIN/WAR, Heishi+Kunimitsu, Capped Haste, Low Accuracy (1133 eva), COR+BRD:
NIN/WAR, Heishi+Kunimitsu, Capped Haste, Mid Accuracy (1300 eva), COR+BRD:
NIN/WAR, Heishi+Kunimitsu, Capped haste, High Accuracy (1500 eva), COR+BRD:
NIN/WAR, Heishi+Kunimitsu, Capped haste, High Accuracy+Madrigal (1500 eva), COR+BRD:
NIN/WAR, Nagi+Gleti AM3, Capped haste, High Accuracy+Madrigal (1500 eva), COR+BRD:
I should note that I needed to adjust the code to simultaneously swap up to three pieces in order for the final sets to become independent of the initial sets at low haste values due to the variety of dual wield options Ninja has.
See below for my preliminary results. I won't claim these are the best sets, or even necessarily correct sets, since there could be some simple mistake in my code. Additionally, a simple time-to-weapon skill may not be a robust enough method to find best TP sets, but I'm hesitant to use 'ideal DPS' as a measure of performance. I may end up using average DPS in the end anyway... At the very least, these sets are something to think about.
NIN/WAR, Heishi+Kunimitsu, 0% Haste, Low Accuracy (1133 eva):
NIN/WAR, Nagi+Gleti AM3, 0% Haste, Low Accuracy (1133 eva):
NIN/WAR, Heishi+Kunimitsu, 15% Haste, Low Accuracy (1133 eva):
NIN/WAR, Heishi+Kunimitsu, 30% Haste, Low Accuracy (1133 eva):
NIN/WAR, Heishi+Kunimitsu, Capped Haste, Low Accuracy (1133 eva), COR+BRD:
NIN/WAR, Heishi+Kunimitsu, Capped Haste, Mid Accuracy (1300 eva), COR+BRD:
NIN/WAR, Heishi+Kunimitsu, Capped haste, High Accuracy (1500 eva), COR+BRD:
NIN/WAR, Heishi+Kunimitsu, Capped haste, High Accuracy+Madrigal (1500 eva), COR+BRD:
NIN/WAR, Nagi+Gleti AM3, Capped haste, High Accuracy+Madrigal (1500 eva), COR+BRD:
[+]
Thanks for your comited work. I improved a lot of my ninja WS sets with your work, and looking forward testting those TP sets (even if there's further modifications).
[+]
balder earring is a great example of why an arithmetic mean is not an ideal way to gear
you also have 99% hit rate when ranged caps at 95%
you also have 99% hit rate when ranged caps at 95%
[+]
You mean wearing something that only works 1 out of every 100 melee swings looks exaggerated on paper and isn't actually the right way to go. Blasphemy. How dare you.
[+]
Ramuh.Austar said: »
balder earring is a great example of why an arithmetic mean is not an ideal way to gear
The Balder Earring +1 itself is not contributing enough over the next-best option to make such a claim. If I rerun the "Low Accuracy, Capped Haste, BRD+COR" set without Balder Earring +1 being an option, then the code returns the exact same set, but swapping Balder Earring +1 for Telos Earring and increasing time to WS from 1.738681s to 1.738414s; ~0.2ms or ~0.015% longer.
Similarly, if we look at the Blade: Kamu sets I posted the other day: Balder Earring +1 provided either 10 or 24 more average WS damage over the next-best option (Brutal Earring in that case). This was only between 0.08% to 0.2% increase in the average damage, nowhere near enough to claim that stats similar to Balder Earring +1 discredit the method of using averages for large N.
I do agree that the average is not a good way to compare these sets. I struggled for a while to find a good way to compare WS sets with each other before posting my thread a while back, considering there is no accepted, simple, summary statistic for multi-peaked distributions (see Blade: Kamu/Shun and Evisceration distributions), unlike Gaussians (ish - see Savage Blade and Blade: Ten). I ended up giving up and settling with the average since it allows users to compare with spreadsheets more easily, while providing the distribution plots lets users see the details. I may end up similarly giving up on TP set comparison and simply use DPS for the same reason. DPS will likely be required for Kikoku and Kannagi sets anyway, in order to compare the contribution from their damage boosts with Heishi and Nagi.
I'm curious to what TP sets you're finding for NIN using your full DPS simulator. Posting them would give me something to work towards and compare with later.
Ramuh.Austar said: »
you also have 99% hit rate when ranged caps at 95%
I reran the 30% haste set using a 95% cap and note that the final set does not change, but the time to WS increases from 2.71s to 2.74s.
Izanami said: »
I'm curious to what TP sets you're finding for NIN using your full DPS simulator. Posting them would give me something to work towards and compare with later.
I think its more of a approach difference. AFAIK Austar idea of improving is to keep similar median, but make distribution (I think he mainly looks at 80% of simulations and cut down 10% from lower distribution and 10% from upper distribution or something like that) as narrow as possible. That's why he has problem with Balder's Earring +1. It's totally against his concept of prioritizing stable and repeatable damage.
EDIT: btw I'm just trying to explain his reasoning. I'm super far away of judging it.
[+]
Izanami said: »
Similarly, if we look at the Blade: Kamu sets I posted the other day: Balder Earring +1 provided either 10 or 24 more average WS damage over the next-best option (Brutal Earring in that case). This was only between 0.08% to 0.2% increase in the average damage, nowhere near enough to claim that stats similar to Balder Earring +1 discredit the method of using averages for large N.
TP set ideas are similar, but your current goal only gets TP in the fastest time, ignoring everything else so not really going to go into detail about that for now
SimonSes said: »
That's why he has problem with Balder's Earring +1. It's totally against his concept of prioritizing stable and repeatable damage.
are af+3 daken hands not worth it?
There are many easier and better to get TP hands like adhemar, herculean, and such. Daken isnt really a sought after stat, more of a bonus.
[+]
Ramuh.Austar said: »
If you saved the results as a list and checked the IQRs for 25% and 75%, you'd see that you'd probably be getting higher numbers in both of those with brutal instead.
I'm not sure why decreasing the maximum value would increase the upper limit. I'd expect both limits to decrease, narrowing your distribution, which is what SimonSes had posted.
I ran 1,000,000 simulations for each set (Brutal Earring vs Balder Earring +1) and printed the 50th, 75th, and 25th percentiles for each distribution.
Balder Earring +1
50th: 12614.0
75th: 15594.0 (+2980.0)
25th: 9565.0 (-3049.0)
Brutal Earring
50th: 12615.0
75th: 15591.0 (+2976.0)
25th: 9573.0 (-3042.0)
As expected, the absolute limits become narrower, but the median values are statistically identical as far as we can tell from using 1,000,000 simulations.
I may not understand enough about your method or about statistics in general, but as far as I'm aware: this isn't a single-peaked distribution so this method of comparison is already flawed. We would probably need to look at the median and 1-sigma ranges of each of the individual peaks and maybe include something about the peak-to-peak spacing weighted by normalized peak height. If the peaks overlap, then we'd need to fit each peak to estimate the contribution from one peak to the wing of an overlapping peak, kind of like a 1-D K-means. Summarizing multi-peaked distributions is not easy and I believe there is no accepted standard way to do this, not that I have any idea where to find such a method outside of Wikipedia. At some point it just isn't worth the effort for marginal gains in damage, so I've stuck with using the average to keep things simple and easy to understand. I may swap to median later since the average value is not always an observed value, even if the average is close to the median for large N.
At the end of the day, I've seen improvements in my damage output on NIN by using the results from my code, so my method can't be terrible considering how simple it is. I think it's good that we're using different methods. I'm excited to see if our different methods result in significantly different sets.
Ramuh.Austar said: »
TP set ideas are similar, but your current goal only gets TP in the fastest time, ignoring everything else so not really going to go into detail about that for now
Correct. This is my first step to determining the true best sets, and it likely isn't the best way. I explicitly said that the sets I posted may not even be correct, let alone the best. As I mentioned, I'll have to include some form of normalized damage to compare TP sets properly. For now, I just want players to be aware of specific combinations of gear to use for TPing in since we haven't had any up-to-date TP set discussion in a long time. I'll post the "best" sets later in my Python3 thread as well as the method used to determine them.
Izanami said: »
I'm not sure why decreasing your maximum value would increase your upper limit
Izanami said: »
this isn't a single-peaked distribution so this method of comparison is already flawed
Izanami said: »
I may not understand enough about your method
Not the best example for NIN but here are two results from my MNK one. It is on Apex mobs with just solo Trust buffs and is doing TK > TK > TK making a skillchain, so TP is not held unless the SC window is not open for some reason.
ItemSet 382703
In this example I will compare only the waist slot.
Moonbow +1
Code
4562.3 16514.1 12499.71 0.32 0.46 0.22 4410.21 4711.64 15885.36 17197.48
Fotia
Code
4684.71 16060.47 12275.74 0.3 0.47 0.23 4523.97 4838.06 15520.02 16605.65
The results aren't labeled since my goal is the have them print to the GUI but they are
Code
Average DPS, WS Average, Skillchain Average Auto Attack %, WS %, Skillchain % DPS 25 Q, DPS 75 Q WS 25 Q, WS 75 Q
If just comparing the WS results themselves, it's pretty clear that Moonbow +1 has a higher average and a higher skill chain average. However it's DPS 25 and 75 are bother lower, and the average DPS is also lower, even though it's WS only values are higher. Since I'm not specifically focused on an average WS damage and am more concerned with having a reliable WS result for a skillchain, I would use fotia belt in this case.
Did anyone here have the balls to grab Nyame C for their Ninja to test it out? Very interested to see if so. Thanks!
Asura.Pusheen said: »
Did anyone here have the balls to grab Nyame C for their Ninja to test it out? Very interested to see if so. Thanks!
I almost did it. It was tempting, but I did B in the end XD
Being myself tho, I needed to at least calculate it :D
It's pretty raw Google sheet. I did it for myself with no plans to share it. I guess I could at least add some labels and color to it.
Ok I actually spent some time just now to make this sheet more readable and useful. I made special Tab "Variables", where in green cells I put everything that should be edited by non advance user. Ofc like always, to edit it, you need to make your own copy ("File" -> "Make a copy"). Gear is set in stone for now (well not really, you can easily edit it if you know what you are doing, but there is no selection of gear). Maybe I will have time one day to add option of choosing gear.
[+]
YouTube Video Placeholder
2:58 - Math Breakdown
4:21 - Tanking Footage
About two months ago a subscriber, Lepton, reached out via Discord and told me he had won the Yagyu Darkblade from Mog Bonanza. Unfortunately he did not have the time as things stood IRL to login. In a remarkable act of faith he gave me his account details and pre-loaded it with crysta so I could experiment with it. I think he said that me listing Ninja as my #1 personal favourite job in the game in a recent video did it.
I've spent the last month going hard in getting the character to endgame level. Job master, getting Nyame to R20, getting the Heishi and Fudo Masamune; the works.
I know that Nitenichi was the first to report on the Yagyu Darkblade but, as far as I'm aware, this is the first footage of it being used. Here I give an honest opinion on it, on it's pros and cons.
I still think this will become available along with the other Nomad Bonanza weapons as some things balance it out.
[+]
All this video is going to do is piss me tf off but Im going to watch it anyways.
[+]
You wanna be pissed off? Try being me and Logical when our LS mate won the Darkblade for his NIN that doesn't even have Utsusemi: San unlocked. I can't stand to look at him when he's on NIN lol.
I'm gonna paste my comment from YT, since it's more suited for forum anyway.
Nice video. I don't believe those weapons will ever be released outside of bonanza tho. I think there was some interview with devs saying that those weapons were specifically created for Bonanza. I don't think SE even realize this weapon has such strong implication for NIN's Enmity generation. I bet they have no idea about it.
Edit: This is from interview with devs
"The new unique rank 1 prizes are attractive.
Matsui: This cost us dev resources, so hopefully they are attractive. We think it's more likely to get rank 1 this time around too.
Fujito: We thought they can't be made too good as they're only receivable via 'chance'. But, still, we also introduced other content in the past that gives comparable items (Estuceon for Kupo Shield).
Matsui: We made equipment that doesn't make it so the job cannot be played without it, just gives a boost."
"Just give a boost" "can't be made too good" I think it clearly shows they don't know true potency of this katana and implication of hate generation with 6 Utsusemi targets. It also say they don't plan to let us get those weapons other way. If anything they will make events "that gives comparable items".
Nice video. I don't believe those weapons will ever be released outside of bonanza tho. I think there was some interview with devs saying that those weapons were specifically created for Bonanza. I don't think SE even realize this weapon has such strong implication for NIN's Enmity generation. I bet they have no idea about it.
Edit: This is from interview with devs
"The new unique rank 1 prizes are attractive.
Matsui: This cost us dev resources, so hopefully they are attractive. We think it's more likely to get rank 1 this time around too.
Fujito: We thought they can't be made too good as they're only receivable via 'chance'. But, still, we also introduced other content in the past that gives comparable items (Estuceon for Kupo Shield).
Matsui: We made equipment that doesn't make it so the job cannot be played without it, just gives a boost."
"Just give a boost" "can't be made too good" I think it clearly shows they don't know true potency of this katana and implication of hate generation with 6 Utsusemi targets. It also say they don't plan to let us get those weapons other way. If anything they will make events "that gives comparable items".
I'm with Simon on this.
Altough... granted it's pretty clear to me this has been the situation so far, it doesn't mean that they won't change their mind in the future. It's not written in stone.
Altough... granted it's pretty clear to me this has been the situation so far, it doesn't mean that they won't change their mind in the future. It's not written in stone.
SimonSes quoted Devs saying said:
introduced other content in the past that gives comparable items (Estuceon for Kupo Shield)
So we won't be getting bonanza weapons (Kupo Shield-level reward) but will be stuck with a grindy (likely heavily botted) process that will get us an equivalent (Estuceon-level reward) weapons with the properties of the Bonanza weapons.
What I read from it.
Another grindy heavily-botted item. The shields with botters at the entrance of zones synthing on mains while their throw-away LV99 Sparks THFs bots farm up items. So disgusting, but oh well. SE doesn't seem to care about this game's cheaters so I just have to accept it. Expect something similar with Weapons. "Please look forward to it!"
(I want to be a glass-half-full person, but I calls 'em like I sees 'em).
Phoenix.Capuchin said: »
You wanna be pissed off? Try being me and Logical when our LS mate won the Darkblade for his NIN that doesn't even have Utsusemi: San unlocked. I can't stand to look at him when he's on NIN lol.
Your LS mate... I haven't met him but I don't like him.
All FFXI content and images © 2002-2024 SQUARE ENIX CO., LTD. FINAL
FANTASY is a registered trademark of Square Enix Co., Ltd.