The Sealed Dagger: A Ninja Guide

言語: JP EN DE FR
2010-06-21
New Items
users online
フォーラム » FFXI » Jobs » Ninja » The Sealed Dagger: A Ninja Guide
The Sealed Dagger: A Ninja Guide
First Page 2 3 ... 222 223 224 ... 253 254 255
 Asura.Sechs
Offline
サーバ: Asura
Game: FFXI
user: Akumasama
Posts: 9876
By Asura.Sechs 2021-06-22 08:42:34  
I wasn't trying to attack you! It's just that there are 4 types of Daifuku and you just wrote "Daifuku" without which specific one.
I assume you were talking about Grape Daifuku but I wanted to make sure xD
 Lakshmi.Buukki
Offline
サーバ: Lakshmi
Game: FFXI
By Lakshmi.Buukki 2021-06-22 08:44:16  
Crawlers diremites marids and spiders resist chi, susceptible to teki and to

edit: diremites and marids might slightly resist all 3. Noticed normal damage at capped TP across all.

Flans heavily resist all 3
Soulflayers heavily resist all 3
Chigoes resist chi
Wamoura appear to slightly resist chi, susceptible to teki and to

Apkallu susceptible to chi
Rabbits susceptible to chi, probably all
Dahaks susceptible to all, especially teki
Manticores susceptible to hybrid but can't remember which. Might be teki
Skeletons susceptible to all (like a major weakness, they get destroyed)
Ghosts susceptible to all
Slimes susceptible to chi (yes, ninja can safely kill these)
aquans (jagils, crabs) are susceptible to chi, even with slashing penalty
Clusters heavily resist all 3

Wyrm and Behemoth both are susceptible
Hydra takes reduced damage
Dvergr heavily resists (blade ten destroys him anyhow)
[+]
Offline
Posts: 202
By malakef 2021-06-22 08:44:33  
Wait.. didnt i say grape? I did!
 Asura.Kitfoxtrot
Offline
サーバ: Asura
Game: FFXI
user: foxtrot
Posts: 55
By Asura.Kitfoxtrot 2021-06-22 08:51:30  
Ruaumoko said: »
Asura.Sechs said: »
I think it would be awesome if we could all work together to collect data on which targets in odyssey is susceptible to damage from any of NIN's hybrid WSs.
Did a run the other day. The following for sure are vulnerable.

Bats
Vultures
Colibri
Imps
Crabs
Wasps
Tigers
Rocs

Work on but take reduced damage.
Trolls
Mamool

Interesting you got decent #'s on big birds? Would've thought you'd basically be out of luck on any heavy slashing resistant mobs since you'd be taking a hit on phys dmg portion. Granted Hybrid set is sitting @ 1530 attck w/o food or buffs, so maybe able to still power through? Savage blade is obviously a different beast but its able to say screw mechanics to most of those mobs in there, due to naegl OP factor which includes likely being attack capped on most junk in there.

I haven't been able to mess around on nin in Sheol-C. Farming in A before segments/t4 set came out and most mobs were just about one shot with only trust buffs. But mobs are pretty weak in A comparatively.

Asura.Sechs said: »
Did you notice if all Hybrids were working well, or if one was working better than the other?
(without this aspect I'd be leaning to just use Blade: Chi)

Prob safe to say play into mobs ele weakness, if that's not an option (and not resistant to earth) > chi.
 Asura.Sechs
Offline
サーバ: Asura
Game: FFXI
user: Akumasama
Posts: 9876
By Asura.Sechs 2021-06-22 08:57:57  
Thanks Buukki.
Please note that Flans and Wamouras (the small ones, not the flying ones) alternate between two modes and take different damage from different sources according to their current mode.
 Lakshmi.Buukki
Offline
サーバ: Lakshmi
Game: FFXI
By Lakshmi.Buukki 2021-06-22 09:10:11  
I meant the Wamoura (flying), as the Wamouracampa (crawling PLDs) are a different monster I didn't list.

Flans could have been in smooth head mode, but I don't have time to determine the mode in Odyssey. The damage sucks the majority of the time xD
Offline
Posts: 298
By Izanami 2021-06-22 12:19:12  
I've been developing a code to automate finding the "best" gear set for each weapon skill given specific buffs and enemy stats. It works well for physical weapon skills, but I'm running into trouble modifying it to work with hybrids as well.

Do we know the exact method/equation used to calculate Hybrid WS damage? Let's use Blade: Chi as an example since it is a 2(+1)-hit weaponskill.

Specifically:

  • Ignoring Ternion, which hits benefit from the WSD stat? If it's only the first hit, then does the first off-hand hit also benefit?

  • Is the damage from each hit calculated as a (physical_part) + (magical_part), and if so, how do I calculate the (magical_part) and do I multiply the (physical_part) by WSD and the (magical_part) by WSD separately (essentially double-dipping on WSD)?

  • Does FTP transfer across all hits? BG does not list the Hybrids as FTP replicating, but I can make my simulated damage distribution almost match the observed distribution if I turn on FTP replicating and play with WSD elsewhere.



Using the Hybrid WS damage equation on BG, my simulated damage distribution was off by about a factor of 3 from the observed distribution so I clearly don't understand the equation on BG.

I then turned to SimonSes' Hybrid WS damage spreadsheet, which helped move my simulated distribution's median to near the observed median value. However, the spreadsheet uses replicating FTP, WSD on all hits, and has Fotia gear providing both FTP+0.1 and weapon skill damage +10% per piece, which makes me doubt the accuracy the spreadsheet as a whole, despite it returning a relatively correct average value. (Edit: see the video and my next comment below) Since the spreadsheet is designed to return an average value, I can't easily verify its true accuracy across the spread of expected damage.

Right now I'm using a slightly modified version of the equations in SimonSes' spreadsheet. In short, the process I'm currently using is as follows: In a loop for each hit of the weapon skill individually, I first calculate the physical_part using standard methods (Weapon_Damage, fSTR, WSC, WSD, PDIF, and FTP). For the magical_part, I take the player's magic_damage_stat and add it to the physical_part and multiply that sum by the standard magic_multiplier consisting of things like elemental_affinity, day_weather, magic_attack_ratio, orpheus_bonus, enemy_mdt, and enemy_sdt. For the first hit of the weapon skill (main-hit and sub-hit) I multiply the physical_part and magical_part by the WSD stat. I then simply add the physical_part and magical_part to get the total damage dealt by a single hit of the weapon skill. Finally, before moving onto the next hit of the weapon skill, I check for multi-attack procs and repeat the process for each multi-hit proc. This process is repeated for each hit (three times for Blade: Chi). I do not transfer FTP, and my multi-attacks and hits after the first main- and sub-hits do not gain the benefit of WSD.

The figure below shows the output distribution using this method overplotted onto an observed distribution (N=672). No buffs or debuffs were used for the observed distribution and the simulated distribution takes into account the enemy's stats, including magic and earth resistance (but not magic evasion since I could not find a source that publishes enemy magic evasion).The resulting simulated distribution matches observed distribution (N=672) decently well, but the peaks corresponding to multi-attacks in the simulated distribution are at higher damage values than what is seen in the observed data. Additionally, the simulated distribution's main peak is much narrower than the observed distribution's.

However, if I turn ON FTP replication and turn OFF WSD on sub-hit (for physical and magical), then I can get the assumed multi-attack peaks to align better, but the main peak is still too narrow and I'm not seeing the peaks for missed hits at all in the observed data.
Based on the code working well for physical weapon skills, I'm more inclined to trust the first situation without FTP replication. Considering that the multi-attack peaks don't really become prominent until over 100,000 simulated WSs, in both cases above, the small sample size (N=672) for observed weapon skills can explain the multi-attack peaks being slightly off, or non-existent, but it doesn't explain why the observed main peak is much wider than the simulated peak. There's a chance that standard magical resists from enemy magic evasion are the reason the main peak is so broad in the observed data and that the sub-peaks don't seem to line up exactly, but I'm not sure what a standard magic evasion value would be if I were to include it. Could anyone help me out here?
[+]
Offline
Posts: 37
By Malaketth 2021-06-22 13:05:47  
Start with this
YouTube Video Placeholder



Which should address all of that except for how it treats additional attacks. Own personal experience is each attack seems to be treated that way and is easily observed when you get an extra attack on Chi but that’s admittedly just an eyeball check.
[+]
Offline
Posts: 298
By Izanami 2021-06-22 13:26:35  
Malaketth said: »
Start with this

Which should address all of that except for how it treats additional attacks. Own personal experience is each attack seems to be treated that way and is easily observed when you get an extra attack on Chi but that’s admittedly just an eyeball check.

Thanks, I had completely forgotten about that video. I remember watching it when it was new too.

I can't rerun my code for a few hours to check the new simulated distributions, but I do know now that I was not explicitly applying FTP to the magical_part and that I was applying the wrong FTP values to the physical_part. I'll change the code to use FTP=1.0 for the physical_part and use the BG-listed FTP values to the magical part as described in the video. The one thing the video did not make clear was if the FTP for the magical_part (the BG-listed values) are FTP replicating. But that's something I can change in the code and compare with the observed data relatively easily.

The video also explains why SimonSes uses FTP+0.1 and weapon skill damage +10% (for the physical_part) for fotia gear in his spreadsheet.

Any other insight from other players is still welcome for other parts of the process.
Offline
Posts: 8843
By SimonSes 2021-06-22 13:56:27  
Izanami said: »
if the FTP for the magical_part (the BG-listed values) are FTP replicating.

It cant be replicating, because its a single magical hit. The magical part cant multiattack and doesnt have offhand hit.
 Lakshmi.Buukki
Offline
サーバ: Lakshmi
Game: FFXI
By Lakshmi.Buukki 2021-06-22 14:30:27  
Malaketth said: »
And if you just absolutely must switch weapon type NIN is on Karambit don’t forget.

I don't have karambit, but I have used chi on both undead (skeletons and ghouls) as well as aquans (fish and crabs) and I can say that hybrid WS outperforms even piercing options On slashing resist monsters. If I save tp to around 2k, I can still do about 45k~ damage to these monsters, even though they resist slashing by as high as 75%. Kind of crazy how vulnerable they are to hybrid.

I would say ninja never needs to pull out a blunt weapon at all for Odyssey, and can default to Tauret if it absolutely needs a secondary option. Lamia and Flans are going to be your worst monster grouping for dealing damage.

I didn't mention qutrub because they obviously take extra damage, but they take near capped damage even at 1k to blade chi. One ninja can take out this group in like sixty seconds
Offline
Posts: 202
By malakef 2021-06-22 14:49:32  
I was specifically thinking of Lamia there but to be honest when we got those beastmen I just let the others kill them while I mop up the regular adds. And yeah any DD should be able to hit max/or near max damage on qutrubs so just use whatever.
[+]
 Asura.Kitfoxtrot
Offline
サーバ: Asura
Game: FFXI
user: foxtrot
Posts: 55
By Asura.Kitfoxtrot 2021-06-22 16:15:16  
Lakshmi.Buukki said: »
I don't have karambit, but I have used chi on both undead (skeletons and ghouls) as well as aquans (fish and crabs) and I can say that hybrid WS outperforms even piercing options On slashing resist monsters. If I save tp to around 2k, I can still do about 45k~ damage to these monsters, even though they resist slashing by as high as 75%. Kind of crazy how vulnerable they are to hybrid.

This on Sheol-C and past floor 1? Banishga? 45k is a good bit more than I'd expect for slash resistant. On sam few back was kagero > koki to kill skeles I think it was on floor 3 of C, but both ws (elements skele are weak to) were only doing maybe mid 20k's due to phys dmg reduction.
 Lakshmi.Buukki
Offline
サーバ: Lakshmi
Game: FFXI
By Lakshmi.Buukki 2021-06-22 18:25:16  
Last night we got Skeletons on a C4 Halo, and I was able to hit in the 40ks a few times. Like I said, I hold my TP to 2k to make the best of it, just in case it doesn't one-shot, I can follow-up with Teki or To, is the strat I have been rolling with. Some strange behavior for sure.

Are we certain that Hybrid WS damage for the physical portion is always slashing? Or is it perhaps some form of neutral damage, or otherwise not subject to monster weapon resistances?
 Asura.Sesshmaru
Offline
サーバ: Asura
Game: FFXI
user: Sesshmaru
By Asura.Sesshmaru 2021-06-23 02:38:53  


Could not resist not to play NIN after looking at this thread :D
[+]
 Asura.Sechs
Offline
サーバ: Asura
Game: FFXI
user: Akumasama
Posts: 9876
By Asura.Sechs 2021-06-23 03:07:04  
@Langly
Do you plan to update the Spreadsheet in the near future, with all the new Odyssey gear?


Edit:
If you trust me I can do it myself. I'm about to do it locally for my own copy of your spreadsheet, would be a shame for all that work to go to waste.
It's a lot of gear and it takes a while to add everything in its various aug versions.
[+]
 Phoenix.Logical
Offline
サーバ: Phoenix
Game: FFXI
Posts: 510
By Phoenix.Logical 2021-06-23 07:33:39  
Asura.Sechs said: »
@Langly
Do you plan to update the Spreadsheet in the near future, with all the new Odyssey gear?


Edit:
If you trust me I can do it myself. I'm about to do it locally for my own copy of your spreadsheet, would be a shame for all that work to go to waste.
It's a lot of gear and it takes a while to add everything in its various aug versions.

I already updated the spreadsheet with the new gears and various tiers a few months ago for Langly. We spoke when I did it and from what I recall he found no issues with my changes. You can find them here. https://github.com/NextGames2000/Spreadsheets

Just go with the Ninja spreadsheet as the rest are for specific setups.
[+]
 Asura.Sechs
Offline
サーバ: Asura
Game: FFXI
user: Akumasama
Posts: 9876
By Asura.Sechs 2021-06-23 09:19:08  
Can I update the ffxi spreadsheet repository with that link?
 Phoenix.Logical
Offline
サーバ: Phoenix
Game: FFXI
Posts: 510
By Phoenix.Logical 2021-06-23 09:22:12  
Asura.Sechs said: »
Can I update the ffxi spreadsheet repository with that link?

Absolutely. I intend to keep adding any gear that comes out to it so it should remain updated.
 Quetzalcoatl.Langly
Offline
サーバ: Quetzalcoatl
Game: FFXI
user: Langly
Posts: 683
By Quetzalcoatl.Langly 2021-06-23 09:25:14  
I don't plan to at the moment. I would happily welcome any collaborative efforts to keep it updated.

I'll add Logical's repo to the front page so if someone goes looking they might find it.
 Bismarck.Gippali
Offline
サーバ: Bismarck
Game: FFXI
user: Gippali
Posts: 590
By Bismarck.Gippali 2021-06-23 09:40:40  
I still don't see that Nagi, Langly. >.>

Lakshmi.Buukki said: »
Crawlers diremites marids and spiders resist chi, susceptible to teki and to

edit: diremites and marids might slightly resist all 3. Noticed normal damage at capped TP across all.

Flans heavily resist all 3
Soulflayers heavily resist all 3
Chigoes resist chi
Wamoura appear to slightly resist chi, susceptible to teki and to

Apkallu susceptible to chi
Rabbits susceptible to chi, probably all
Dahaks susceptible to all, especially teki
Manticores susceptible to hybrid but can't remember which. Might be teki
Skeletons susceptible to all (like a major weakness, they get destroyed)
Ghosts susceptible to all
Slimes susceptible to chi (yes, ninja can safely kill these)
aquans (jagils, crabs) are susceptible to chi, even with slashing penalty
Clusters heavily resist all 3

Wyrm and Behemoth both are susceptible
Hydra takes reduced damage
Dvergr heavily resists (blade ten destroys him anyhow)

Lynx are also on the list for Chi/Teki susceptible, didn't check To though.

Edit: Saplings seem to be the same as the Lynx.
[+]
Offline
Posts: 8843
By SimonSes 2021-06-23 09:42:35  
Izanami said: »
I've been developing a code to automate finding the "best" gear set for each weapon skill given specific buffs and enemy stats. It works well for physical weapon skills, but I'm running into trouble modifying it to work with hybrids as well.

Do we know the exact method/equation used to calculate Hybrid WS damage? Let's use Blade: Chi as an example since it is a 2(+1)-hit weaponskill.

Specifically:

  • Ignoring Ternion, which hits benefit from the WSD stat? If it's only the first hit, then does the first off-hand hit also benefit?

  • Is the damage from each hit calculated as a (physical_part) + (magical_part), and if so, how do I calculate the (magical_part) and do I multiply the (physical_part) by WSD and the (magical_part) by WSD separately (essentially double-dipping on WSD)?

  • Does FTP transfer across all hits? BG does not list the Hybrids as FTP replicating, but I can make my simulated damage distribution almost match the observed distribution if I turn on FTP replicating and play with WSD elsewhere.



Using the Hybrid WS damage equation on BG, my simulated damage distribution was off by about a factor of 3 from the observed distribution so I clearly don't understand the equation on BG.

I then turned to SimonSes' Hybrid WS damage spreadsheet, which helped move my simulated distribution's median to near the observed median value. However, the spreadsheet uses replicating FTP, WSD on all hits, and has Fotia gear providing both FTP+0.1 and weapon skill damage +10% per piece, which makes me doubt the accuracy the spreadsheet as a whole, despite it returning a relatively correct average value. (Edit: see the video and my next comment below) Since the spreadsheet is designed to return an average value, I can't easily verify its true accuracy across the spread of expected damage.

Right now I'm using a slightly modified version of the equations in SimonSes' spreadsheet. In short, the process I'm currently using is as follows: In a loop for each hit of the weapon skill individually, I first calculate the physical_part using standard methods (Weapon_Damage, fSTR, WSC, WSD, PDIF, and FTP). For the magical_part, I take the player's magic_damage_stat and add it to the physical_part and multiply that sum by the standard magic_multiplier consisting of things like elemental_affinity, day_weather, magic_attack_ratio, orpheus_bonus, enemy_mdt, and enemy_sdt. For the first hit of the weapon skill (main-hit and sub-hit) I multiply the physical_part and magical_part by the WSD stat. I then simply add the physical_part and magical_part to get the total damage dealt by a single hit of the weapon skill. Finally, before moving onto the next hit of the weapon skill, I check for multi-attack procs and repeat the process for each multi-hit proc. This process is repeated for each hit (three times for Blade: Chi). I do not transfer FTP, and my multi-attacks and hits after the first main- and sub-hits do not gain the benefit of WSD.

The figure below shows the output distribution using this method overplotted onto an observed distribution (N=672). No buffs or debuffs were used for the observed distribution and the simulated distribution takes into account the enemy's stats, including magic and earth resistance (but not magic evasion since I could not find a source that publishes enemy magic evasion).The resulting simulated distribution matches observed distribution (N=672) decently well, but the peaks corresponding to multi-attacks in the simulated distribution are at higher damage values than what is seen in the observed data. Additionally, the simulated distribution's main peak is much narrower than the observed distribution's.

However, if I turn ON FTP replication and turn OFF WSD on sub-hit (for physical and magical), then I can get the assumed multi-attack peaks to align better, but the main peak is still too narrow and I'm not seeing the peaks for missed hits at all in the observed data.
Based on the code working well for physical weapon skills, I'm more inclined to trust the first situation without FTP replication. Considering that the multi-attack peaks don't really become prominent until over 100,000 simulated WSs, in both cases above, the small sample size (N=672) for observed weapon skills can explain the multi-attack peaks being slightly off, or non-existent, but it doesn't explain why the observed main peak is much wider than the simulated peak. There's a chance that standard magical resists from enemy magic evasion are the reason the main peak is so broad in the observed data and that the sub-peaks don't seem to line up exactly, but I'm not sure what a standard magic evasion value would be if I were to include it. Could anyone help me out here?

You confuse this too much by doing separate calculation for magic damage for each hit.

Physical part is where multiattack proc and each hit has 1.0 ftp
You should calculate this like normal physical multihit WS with normal wsc and wsd on first hit only and no ftp transferring etc.

Then when you get damage for that, you put that damage as base for magical part which is like magic WS. So can't multi attack, is one hit and has one ftp listed on bg wiki. It's like death penalty, but physical part is added to base damage.

So to be super clear about WSD. It works for only first hit of physical part, but it works for whole magic part, because magical part is single magic hit, not magic multihit.

And super clear to ftp transfer. There is no ftp transfer on physical part and each hit is default 1.0 fTP . fTP of magical part varies with tp and don't have ftp transfer, because there is no 2nd magical hit.
[+]
 Quetzalcoatl.Langly
Offline
サーバ: Quetzalcoatl
Game: FFXI
user: Langly
Posts: 683
By Quetzalcoatl.Langly 2021-06-23 10:04:31  
Quote:
I still don't see that Nagi, Langly. >.>

Jokes on everyone, I said I was doing a Nagi, but in reality after I finished the death penalty I turned in for a Yagrush. WEEEEEEEEEEEEEEEEE! O_O
[+]
Offline
Posts: 298
By Izanami 2021-06-23 11:34:10  
SimonSes said: »
You confuse this too much by doing separate calculation for magic damage for each hit.

Physical part is where multiattack proc and each hit has 1.0 ftp
You should calculate this like normal physical multihit WS with normal wsc and wsd on first hit only and no ftp transferring etc.

Then when you get damage for that, you put that damage as base for magical part which is like magic WS. So can't multi attack, is one hit and has one ftp listed on bg wiki. It's like death penalty, but physical part is added to base damage.

So to be super clear about WSD. It works for only first hit of physical part, but it works for whole magic part, because magical part is single magic hit, not magic multihit.

Thanks. I was thinking that was the case based on your spreadsheet and your previous comment as well as old hybrid testing on physical immune monsters.

I've adjusted my code to use 1.0, non-replicating, FTP on the physical part and use the BG-listed FTP on the physical portion of the magical part as shown in the video linked earlier. The magical part is calculated after all of the physical hits have finished and thus cannot multi-attack, but will still do damage if all the physical hits miss (due to the additive magic_damage_stat). I apply WSD only to the first main and sub hits of the physical part as well as the entire magical part.

I increased my observed sample to N=1745 and reran the code with the previously mentioned changes. The new simulated distribution is now almost a perfect match to the observed distribution.
There are still two features that worry me in the observed distribution:

  • There is a large, narrow, peak at about 10800 damage that is not seen in the simulated distribution. I'm not sure what caused this. It might still be the relatively small sample size, but even if I draw from the simulated distribution N=len(observed_data) times, I rarely see such a peak.

  • There are no multi-attack peaks in the simulated distribution, even though you could argue that they are weakly seen in the observed distribution. My test gearset was multi-attack starved (+13% DA, +8% TA), though. Even if I artificially inflate triple-attack to 50%, only the physical portion shows a multi-attack peak. The total damage may simply be smeared due to the magic damage multiplier and hybrid FTP applied to the physical portion of the magical damage.




Finally, if I use standard hybrid buffs and leave my multi-attack the same as provided by my stats (+13% DA, +8% TA) the simulated distributions show almost exactly what I expect, although with a surprisingly low median damage value considering the buffs I used. I was expecting the peaks to be further separated and at larger damage values based on experience of Blade: Chi damage in Dynamis Jeuno. But Dynamis Jeuno has a considerable damage bonus based on statue color that I have not applied here in addition to the variety of enemies that would also cause the damage to vary wildly. I was also applying an 0.75*0.85 magic multiplier to account for the Apex Toad target I used for trials earlier. So damage is lower than expected due to this as well.

There may still be minor problems with the code, but after making the suggested changes, I can at least somewhat trust the results it outputs.
[+]
Offline
Posts: 8843
By SimonSes 2021-06-23 12:59:28  
Izanami said: »
I apply WSD only to the first main and sub hits

Only main hit. WSD doesn't work for offhand.

Dynamis jeuno gobs has like 50% bonus that boost both physical and magical part, which makes whole WS to do 2.25 more damage.

At what TP you try to simulate that?
[+]
Offline
Posts: 298
By Izanami 2021-06-23 13:40:08  
SimonSes said: »
Izanami said: »
I apply WSD only to the first main and sub hits
Only main hit. WSD doesn't work for offhand.

Are you sure about WSD not applying to first off-hand hit? Unfortunately I just got to my office so I can't change the code and rerun it for a few more hours, but I believe that when I did try that earlier (before I fixed the math in the code), the damage did not line up at all. If I remember correctly, without the first sub-hit gaining the full WSD bonus, the simulated distribution was (by-eye) about 10%-15% lower than the observed distribution.

The only evidence I have on-hand to imply that the first sub hit also gains the full WSD is that we do know for sure that the first sub hit gains full TP if it successfully lands, while all multi-attacks and hits after the first (for multi-hit weapon skills) only gain TP=10*(1+sTP/100) TP per hit.

Additionally, I was unable to match the observed distribution (before I fixed the math) until I applied the hidden bonus ~+100 accuracy to the first main hit AND the first sub hit. This hidden stat is similar to the magic accuracy boost that Magic Bursts get. Because I was doing this with only accuracy buffs from KuyinHathdenna and defense buffs from Brygid, my weapon skill accuracy was very low on Apex Toads and not having this bonus ~+100 accuracy would have a very noticeable affect on the simulated distribution. As a side note: I was properly accounting for the extra ~5 stat that KuyinHathdenna and Brygid were providing. I made sure that my code correctly calculated and printed the exact same primary and auxiliary accuracy and attack (from the /checkparam <me> command in game) as well as all stats before I began collecting observed data.

I'll have to rerun the code again and turn off sub hit WSD and sub hit accuracy bonus to confirm whether or not they are required for simulations to match observed, but I'm pretty confident that the first sub-hit does gain full WSD bonus. I'd be happy to be wrong if it meant I was closer to the true formula for this code.


SimonSes said: »
At what TP you try to simulate that?

The simulation used the TP range [1000, 1300] (edit: 1300** plus Heishi and Moonshade TP bonuses) uniformly sampled. When I was performing the observed 1745 Blade: Chi, I wrote a lua to automatically use Blade: Chi above 1000 TP and print the TP at which it used the weapon skill. They were almost all between 1000 and 1300 TP, so I wasn't trying to imply that 1000-1300 TP is the best range to use Blade: Chi at.

For the gear-set calculating part of the code, I plan to run it at various TP ranges and various buff states on Apex Toad level stats (without the MDT or SDT) and let it modify all gear including main and sub weapons to find the best set.
[+]
 Quetzalcoatl.Jakey
Offline
サーバ: Quetzalcoatl
Game: FFXI
user: Jakey
Posts: 300
By Quetzalcoatl.Jakey 2021-06-23 14:28:44  
Izanami said: »

Are you sure about WSD not applying to first off-hand hit? Unfortunately I just got to my office so I can't change the code and rerun it for a few more hours, but I believe that when I did try that earlier (before I fixed the math in the code), the damage did not line up at all. If I remember correctly, without the first sub-hit gaining the full WSD bonus, the simulated distribution was (by-eye) about 10%-15% lower than the observed distribution.

The only evidence I have on-hand to imply that the first sub hit also gains the full WSD is that we do know for sure that the first sub hit gains full TP if it successfully lands, while all multi-attacks and hits after the first (for multi-hit weapon skills) only gain TP=10*(1+sTP/100) TP per hit.

Additionally, I was unable to match the observed distribution (before I fixed the math) until I applied the hidden bonus ~+100 accuracy to the first main hit AND the first sub hit. This hidden stat is similar to the magic accuracy boost that Magic Bursts get. Because I was doing this with only accuracy buffs from KuyinHathdenna and defense buffs from Brygid, my weapon skill accuracy was very low on Apex Toads and not having this bonus ~+100 accuracy would have a very noticeable affect on the simulated distribution. As a side note: I was properly accounting for the extra ~5 stat that KuyinHathdenna and Brygid were providing. I made sure that my code correctly calculated and printed the exact same primary and auxiliary accuracy and attack (from the /checkparam <me> command in game) as well as all stats before I began collecting observed data.

I'll have to rerun the code again and turn off sub hit WSD and sub hit accuracy bonus to confirm whether or not they are required for simulations to match observed, but I'm pretty confident that the first sub-hit does gain full WSD bonus. I'd be happy to be wrong if it meant I was closer to the true formula for this code.

Are you taking into account that offhand hit rate caps at only 95% not 99% like mainhand?

Also no other types of ws have WSD apply to offhand even though they also have tp return from offhand. The easiest way to test it for hybrids specifically is use on low hp mobs to see damage from killing before magic portion has a chance to go off (magic damage happens last and its very easy to see when it happens). With a large amount of WSD it should be easy to see if both hands do similar damage (using same weapon) or if its skewed towards the mainhand using the tp returns to know if it died from one hit or two.
Offline
Posts: 298
By Izanami 2021-06-23 18:27:08  
Quetzalcoatl.Jakey said: »
Are you taking into account that offhand hit rate caps at only 95% not 99% like mainhand?

I am using the 99% main-hand accuracy cap and 95% off-hand accuracy cap.

If I remove the +100 accuracy bonus to first off-hand hit, then the distribution looks wildly different (see the Figure below). I think this alone is enough to confirm that the first off-hand gets the accuracy boost. Whether or not that boost is the +100 I'm applying (based on BG-wiki) is a different problem.

If I remove the WSD applied to the first hit in the off-hand, then the distribution is shifted to significantly lower damage (see the Figure below). This implies that either my algorithm is missing a significant portion of damage, or that the WSD does apply to the off-hand's first hit.This is (by-eye) 10% lower than the observed distribution, so the first thing I checked was if the published INT, MDT, and SDT values are correct for Apex Toads such that I am using the correct magic multiplier, as me simply removing the 0.85 factor from monster SDT seems to fix the issue by itself.

To test this, I changed to BLM and headed back to the Apex Toads with locked nuking gear. Because Apex Toads are published to have no Wind damage penalty, I started by casting a few Aero, Aero II, Aero III, Aero IV, and Aero V spells on a single Apex Toad. I used the spell damage coefficients from the table on BG-wiki and the table of Apex enemy stats from one of the Japanese wiki webpages to calculate expected damage dealt with each nuke. My observed damage matched exactly with what is expected for an Apex Toad with 299 INT, -25% MDT, and no Wind damage penalty for all spells casted. This confirmed that the Apex Toads at least have -25% magic damage taken.

I then tried casting Stone, Stone II, Stone III, Stone IV, and Stone V on the same Apex Toad as I used for Aero testing. The damage for each spell matched exactly what was expected for an Apex Toad with 299 INT, -25% MDT, and no Earth damage penalty for all spells casted. This is surprising considering that they are published to take 85% damage from Earth elemental sources in addition to their -25% MDT. It is possible that the author did not test this and simply used values from another source based on other toads.

I finally updated the code to remove WSD applying to the first hit of the off-hand and corrected the 0.85 SDT factor to 1.0. The resulting distributions line up quite well again, but still don't account for that somewhat worrying deviant peak at about 10800 damage nor the potential multi-attack peaks weakly present in the observed distribution.

However, at this point I think I've been provided with enough information to put me on the right track to finishing this code without much more trouble. I should be okay to continue working on this without bothering everyone at every step. Thanks for the help SimonSes and Malaketth. Especially SimonSes for making the spreadsheet that provided most of the hints I needed.
[+]
 Sylph.Reain
Offline
サーバ: Sylph
Game: FFXI
user: dmregm
Posts: 396
By Sylph.Reain 2021-06-24 01:17:18  
The 85% earth isn't a problem. If it's 50% or under you're guaranteed a resist.

So on apex toads, To and Chi should be fine, Blade: Teki will be resisted.


Compare, to this month's ambuscade.



Water 70% is okay. All other elements are guaranteed to resist unless you can raise them to 60% (You can't I believe). This month's Ambuscade has abnormally low magic defense. This makes Blade: Teki so strong this month. The boss has 60 and the adds have 50. Monster magic defense doesn't go below 50. So malaise would only lower the boss by 10, and not do anything to the adds.

See this page: https://www.bg-wiki.com/ffxi/Resist#Elemental_SDT

Rayke and skillchain windows can change these percentages which makes Rayke incredibly strong.
[+]
Offline
Posts: 8843
By SimonSes 2021-06-24 02:11:24  
Izanami said: »
This implies that either my algorithm is missing a significant portion of damage

5%WSD from NIN gifts maybe?
First Page 2 3 ... 222 223 224 ... 253 254 255
Log in to post.