[Shamaya]

Hey there guys. I've posted here only a couple of times recently, and I know this topic has been beaten to death, but I was hoping to be able to contribute.

As I recently I leveled warrior to 75 and wanted to maximize it for what I was intending to mainly use it on: merit level mobs or below, /nin. So I've been reading up on here, Allah', and elsewhere. I liked a lot of the content I was seeing, but it didn't seem entirely sufficient. Even moreso on Allah' than here, there seemed to be a big difference in opinion among veterans. Some adamantly supporting GA even if /nin. Others boasting swordchucks and even fewer, surprising to me, supporting Man/Ridill.

Many of the people reading this right now I'm sure, in light of recent threads, will realize I'm not touching on a debate about War/Sam vs War/Nin, nor is it a debate about what to use warrior for generally. It's about maximum damage in this specific situation (/nin, merit-level). Of course, there's much more in the game for warrior to do, so this will be of limited interest to some.


GA/Nin vs Man/Rid vs Swordchucks

There is the link to the project/breakdown. It's too long to copypasta here. Some things I can summarize though:

Shamaya said:

DPS =
(Dmg/Round * #Rounds(ws->100) + WSdmg)
/ (wsRate + wsPause)

Shamaya said:

Perdu Voulge: (178.9 * 5 + 842.4) / (20.37 + 3) = 1736.9 / 23.37 = 74.32
Ridill/Joyeuse ("Swordchucks"): (168.2 * 5 + 565.9) / (14.49 + 3) = 1406.9/17.49 = 80.4
Maneater/Ridill: (179.1 * 5 + 747.8) / (16.15 + 3) = 1643.3/19.15 = 85.8

These are DPS values. There is much explanation in the article for anyone who wants to read it.

I don't know how this will be received but I value any opinions. If anyone has suggestions or corrections to make, finds any mistakes, I'm particularly interested in that. It only provides opportunity to perfect on applied game math. d(^.^o)

[Rikkitikkitavi]

To my knowledge, it has always been accepted that for /nin merits Maneater/Ridill was the winning combination.

Rid/Joytoy is fun, but I never parsed as high, Vorpal Blade is a little un-dependable, but you swing lolfast.

But for most merit pties these days, I'd think a war with a polearm would win, especially if you have merits. I Haven't tried it because I don't have the gondo- 6% crit polearm, and I generally go SAM to merits.

Good info on the actual DPS numbers, although it may vary quite heavily based on the player.

[Faundrant]

After looking at the link posted, best I can tell you is that the results were setup to skew heavily to dual wield. The tester went to MMJ to test, and had the dual wield eat pizza, while the GAxe user was using Yellow Curry. While this in theory gives an advantage to the GAxe user due to higher attack rating, the accuracy calcs show it to be the lowest acc value of the three setups.

Also not sure where he is pulling his GAxe WS avg dmg of 842.2 from since his parses all show a consistant avg of over 1000+ dmg from GAxe WS. All in all Axe/Ridill will prolly win with correct support even at birds, though the testing was skewed heavily against the GAxe user.

[Deo2]

Ridill is only good for merits if you're the only one using it. Multiple Ridill in a party = more pecking flurry/feather tickle.

A rdm can keep up a normal pace with 1, but you shouldn't overtax the healers and support if you want to maintain a good exp chain.

I vote Man/Rid on power-alone, but I'm against a 2-4 ridill party on exp/hr basis. Have a GA or Polearm with you in case you run into another ridill war, I doubt he'll have the same adaptability or courtesy towards effeciency.

[Byrthnoth]

I don't have enough invested in this game anymore to work through the math, but how did you factor in the difference in pDIF curves for 2H vs. 1H into your DPS numbers? I've yet to see a good 2H pDIF curve post-patch, and (taken at face value) the thread in BG's Advanced section right now seems to indicate that the formulae on FFXIclopedia are incorrect for even the easiest pDIF point to test. At capped cRatio, there seems to be a ~15% difference between 1H and 2H pDIF.


I don't know about you, but after looking at your parses I would certainly merit Great Axe. If a Great Axe user can sub NIN and parse that close to you while keeping the option to sub SAM and blow you out of the water (which he should do in almost every event in the game) open, it would really only make sense to merit for Great Axe. You may be able to play the job better than he did, and you have to wonder how much of the difference between you was made up by skill. The failure of every parse is that you can't parse against yourself. A difference of a few percent in a parse really can be a difference in playstyle.


Rescom's FFXIAH profile says he's at 4/8 GA and went 3/3/4 for merits with 5/5 Warrior's Charge. 4% Crit, 0 Stat merits. Nothing HQ, owned/owns Barone Legs for some reason. No Heca head. TPs in Iota ring. Seems like a solid player, not particularly devoted to WAR. Seems to like to merit, which is good. Your gear is better in some ways (Speed belt, Heca +1), and his is better in others (Aurum Sabatons, E.body). Your WS gear is miles better and your TP gear is pretty solid, but E.body and Aurums aren't anything to mess with and you weren't fully meritted at the time (even though he isn't either) so we could call it even. If you combined both of your gears and went 8/8 GA, 5/5 Berserk+Warrior's Charge+Double Attack+STR with a Perdu Voulge, then you'd probably outparse both of you.

Edit:
Basically, my opinion is that if you only have 8 merits you should obviously spend them on Great Axe. If you have 16 merits for WAR, then you should spend them on Polearm and Great Axe. If you have 20 merits for WAR, go 8 GA, 8 Polearm, 4 Sword. Ridill is good when you're /NIN, but Polearm is much more useful for a well-geared WAR. If we could reassign merits, then obviously GA, Polearm, Sword, and Axe would all have their place, as would 5/5 Bergressor and 5/5 Daserk. If you're only ever going to go WAR/NIN to Dynamis and Einherjar with poor support, then feel free to merit for Ridill. You'd have been better off spending your points on E.body though.

[Shamaya]

Hello again guys. It's the start of another day of work for me and I have nothing better to do than check forums for 6 more hours -.-

Quote

After looking at the link posted, best I can tell you is that the results were setup to skew heavily to dual wield.

Yes. This examination is only concerned with situations where Ridill/Nin is useful, because GA is more useful in so many other situations. Other than that it is very realistic. DW naturally has a better advantage with food since they get the full advantage out of marinara pizza. It is a goddess for my thief. Someone on Allah' suggested hedgehog pie for GA over yellow curry for this comparison, and I agree with that so I'll get around to changing it. If anything, the biggest shortcoming that could lead to skewing would be the lack of aggressor up/down sets, which could benefit the GA user more.

Quote

Also not sure where he is pulling his GAxe WS avg dmg of 842.2 from since his parses all show a consistant avg of over 1000+ dmg from GAxe WS.

While this is true, look at my vorpals as well. Both numbers are roughly 25% higher in parse than on paper (my vorpals reaching 800avg instead of 600; KJ reaching 1k avg instead of 850). As for the math I did it as accurately as possible. There are 2 reasons why the avg is higher. (1) This parse includes Puks, pets, and no lurkers, and (2) The GA user was using red curry. 4 more "could-be" reasons are that (1) the corsair was giving better corsair buffs over time than the ones I listed, and (2) the formulas are not 100% accurate. (3) My math doesn't account for warcry, savagery, or war charge. (4) KJ is rumored to have unexplainable, random "damage highs." . Looking at parses though I'm not so sure about these highs being anything more than normal pDif highs and multihit. If 850 is the avg here, a 5-hit high-pDif KJ with these settings would do 1500. With a 1k avg they'd high at 1800. Meh idk.

Quote

Ridill is only good for merits if you're the only one using it. Multiple Ridill in a party = more pecking flurry/feather tickle.

Actually this runs akin to war/sam arguments. Similar concept to "war/sam is better if you add more because mobs die faster, and you take less damage" the concept here would be "more ridill is less of an issue as mobs die faster, and you take less damage." Sometimes you'll be right, sometimes you won't. Either way I can't see it being too much of an issue, though I do see your point.

Quote

I don't have enough invested in this game anymore to work through the math, but how did you factor in the difference in pDIF curves for 2H vs. 1H into your DPS numbers? I've yet to see a good 2H pDIF curve post-patch,

You did answer your own question here. As a fellow math user I'd love up-to-date values as well. When they introduced the 2h update they adjusted pDif maxes, and then readjusted them. That shouldn't be an issue here. But as far as the curve otherwise, I think you're right that it is quite possible they're changed. It could be so for 1handers too though. One thing I've been concerned about since the 2h update and not looked enough into is,

cRatio

I could swear when the notes for the 2h update came out, they said something like "dmg for 2h will be adjusted for high level mobs." I'd always wondered if they'd given 2h a cRatio advantage since then, similar to Rng's. If so, this could be the difference people are noticing as well. However someone would have probably noticed by now too...what with /checks and all :/

As far as a 15% difference in 1h vs 2h pDif, I could actually very well use this parse and this math as data for that investigation. Those parses were an hour each and use the same settings as here, only the GA used red curry. I can look at how much higher my avg melee hits are on the parse and compare it to the ga avg hit, and account for how much higher each are here in practice than the pDif shown in the math. If the ratio I get is the same between the GA and DW, it is more likely than not that the pDif curves are equivalent (at least in this atk/def range). If not, then we can further substantiate a disparity in the pDif curves. I'd do the math now but my parse images are blocked here at work

Quote

I don't know about you, but after looking at your parses I would certainly merit Great Axe.

I do like GA. I see what you mean and I look forward to having fun with GA/Sam. I agree GA war is impressive even /nin. I admit GA/nin potential is higher than I expected--but I still do not think I would merit it (can't say 100% yet). You have some very good points examining those parses, you took a very close look, haha. Let me also point out a few differences though.

I could have sworn he had 8/8 GA (btw yes, for others that were concerned, the math comparison does include 8GA merits). You might be right though, I'll have to ask him. Also looks like merits otherwise are not geared towards GA. I can tell you this though, his merits might be very outdated. My gear and merits are outdated on ffxiah because I just don't update them. I think Rescom was boasting 300+ merits and was telling me he was very proud of his GA war and put so much work into it, and wholly felt that GA/Nin was stronger and it was one of the reasons he did not wishlist a ridill.

It may be so that our gear balances out. An a.berk and aurum feet would do wonders to my DPS. But I could be underestimating full heca+1. Aside from that though, I should note that in both of those parses he was using red curry. I'm not going to use red curry more than once in a blue moon. Also I think our merit disparity is greater. I really need to finish these bergressor merits.

As for playstyle, we were both equally aggressive/skilled. I was on vent with him at the time as well. He thought I was "playing for the parse," but in reality I almost never do. So in the first parse I came out a significant lead ahead, despite these disparities. In the second parse, he asked that I voke and said that he wouldn't, and that he was going to play for the parse as much as possible. The other warrior and I voked. I did cast ichi. Rescom did not voke and said he casted ichi 0 times. So there's a noticeable difference in DoT moving from one style to the other, even if it is less effective for the party/chain overall. In that second parse, I was 1% behind almost the entire time, but at the end I 'gave up' and started talking to a chick until the end of the party, at which point I fell to 2% below.



Not bad discussion so far but I want to press that against stressing merits-only mentality, and similarly not stress g.colibri-think. Also when I started this I wasn't so concerned about GA vs Ridill. I was more concerned with Swordchucks vs Man/Rid, since a lot of people on forums seemed to be endorsing them. I'm using the merit party settings as a bar, as they are often a good one. There are plenty of other situations with settings similar to a merit party. The most notable of which is probably low-man melee-style limbus. Essentially we'd just bring a 5-6 man 'merit party' to the zone. My last LS also used merit-style theory to clear dynamis zones. Basically we'd have 2 "merit parties" and 1 party of backup/misc. Each party would /assist their tank, which was a DPS war/nin. This was our strategy for full-clearing zones with 18. I love my HNM shell but I really miss that LS

[Kaparu]

I don't have it in me to discuss this nonsense anymore, but I'll never tire over saying that its idiotic for any job proficient in a two-handed weapon to sub ninja in a merit party(Kraken Club-paladins not withstanding).

[Shamaya]

Oh Kaparu, <3

You're actually the #1 incentive I would have to go into FortAxe math for this comparison as well, since you swear by that. It might actually improve the accuracy of the methodology if I had enough reason to account for TP-surplus.

Please though let's keep this thread clean. The /sam v /nin debate is concluded so far as that there are situations where /nin is useful. The only debate is when and how often. That is for other threads though, as the topic I set in the OP is only /nin.

[Byrthnoth]

Well, in the past I endorsed Ridill/Joyeuse in situations where Acc wouldn't be an issue. Otherwise I estimated GA would be about as good as any Ridill combination based off a rough feeling I got and approximately no math, and I predicted somewhere between the two was a point where Man/Ridill won due to the extra main hand Acc. To be fair, your GA user should not have parsed 85% Acc at MMJSP, especially when he obviously has so much Acc gear available to him. An extra 5% hit rate would have pushed him up a point or two and made the parse even closer.


As far as the level correction function goes, I have no proof but I don't think it was adjusted for 2Hers. My basic reasoning is that I don't remember the 2H damage boost dramatically increase the upper end of HNM non-SAWS damage for one-hit WSs after they toned things back a little and not counting the SA-from-the-front patch. I think what they did was simply extend the pDIF formula to reward having extra Attack. Like, I think they extended the cRatio->pDIF curve past 2.0 for 2H weapons to reward the extra attack given through the stat bonus. It not super hard to hit cRatio of 2.0 on WAR though. 831 Atk caps it for 83 Blusterers, for instance. With Pizza +1, Perdu Voulge, and Double Minuets, that's a Berserk and Warcry away. With Lucky Chaos Roll (with DRK), Double Minuets, Red Curry, Berserk, and Warcry you could get 1368 Atk on WAR in respectable TP gear. UUURRG FEEL THE BURN!!!!
As an interesting note, if you look at the testing on BG (which I haven't verified is done correctly or anything) it shows that 1H and 2H pDIF are both capped at 3.15ish, which indicates that they did slightly muck with the 1H pDIF function as it used to be capped at 3. Did you guys notice your SAWS damage jump up after those patches?



Btw. Every time I've done the Fort Axe math it has come out unfavorably, as did my own experience with it. I love it for Campaign though. The main problem is that you don't actually increase your WS frequency as much as you could. With ~22% DA rate and Perdu Voulge, you average about 4 attack rounds per WS (One DAs). With Fort Axe, you still average in the low 3s simply because there's no way to get 100TP off the first attack round. If you're actually going to do the math for it though, you have to assume the best WS gear possible as its low base damage during WS can be nicely offset by a good WS set and 50% STR mod on KJ.
My numbers are a little off, but the last time I ran them I got:
2nd - 15% / 3rd - 51% / 4th - 27% / 5th - 10% as the 100TP distribution for Fort Axe with 22% DA, 93% Acc, and 55% Fort Axe proc rate.
If it has half the damage and less than double the WS frequency, how can it win?

[Shamaya]

That 100tp distribution is helpful to know. Ideally that's something I should have also incorporated with Swordchucks and Man/Rid, but I didn't. To get a more accurate comparison, I'd have to increase complexity another notch and account for this as well. I'm not sure if I'll be able to find the time to do it..

Don't forget about puk-flash messing up the acc at south camp, though you're right that his acc is still low regardless (this was one of the more irritating things about not playing Thf at the time; Thf tears through that flash with sa/ta or sa/ta+sb/ms/ms).

I never noticed real WS increases after the 2h buff. I did, however, notice pDif crit caps above 3.0 when doing evisceration testing on bunnies with and without x's knife. I didn't really ponder it enough though, and never really finished that testing (same as others on Allah').

Interesting that this thread (pDif cap testing)just popped up, in the midst of all this.

[Byrthnoth]

I think that distribution is correct in spirit, but I'm not entirely sure how I got it as it adds up to over 100%. I think I used an approximation.

The single-round distribution should be:
First hit - 93%
DA on First hit - 20.46%
Sea Wep Proc - 51.15%
DA on Sea Wep Proc - 11.253%

Some people were questioning whether or not Fort Axe can get 4 hits recently, but I swear I remember it happening, I have it in my xls, and wiki backs me up for all it's worth.

Edit: Reworded it to make it clearer.

[solara]

I apologize for stepping into the realm of pointless arguments flaring up, I have to beat this dead horse.

Someone on another forum is attempting to use your post as an argument that Man/Rid War will consistently out perform Perdu War/Sam, despite the fact that you are clearly attempting to illustrate the difference between the two in situations where using /nin is appropriate.

Can you please clarify that the apple/oranges comparison was not the point of your test, and whether you believe (in general, not just from this specific test) that Dual wield Ridill is superior to Perdu War/Sam?

I honestly just want to be able quote you so I don't have to get into it myself.

[Shamaya]

Someone's referring to this? kewlll. Well some things need to get sorted out. I've neglected commenting in here and on allah' recently, but there are impending updates to the math. I will be updating some things such as fStr calc, "ws pause", ga food, agressor up/down sets (hopefully), and maybe another thing or two.

You can tell them that this is wholly insufficient as a basis of war/sam vs man/rid. If they want, they can try using my same math for themselves and come up with a comparison that will work, but that's on them. The best thing one could do for war/sam vs man/rid would be to start with the assumption of full-time seigan, and compare the numbers there. I think that just by doing, that, we'll see how close the numbers are.

I think a proper ridill/war with all the most amazing gear and merits is at least as strong (probably a little stronger--my guess) than a full-time-seigan war/sam. Just a guess tho. And when I say that, I'm saying that they'll be close and that it won't require anything near full-time hasso to push war/sam over that. Assuming non full-time seigan, obviously it's completely dependent on several factors. Will the war/sam be more likely to die? Maybe. What's the estimated increased % chance and how would that affect our comparison? Obviously there are two things that are very important; well, 3 things: (1)the amount of hasso being used, (2)the strength of the healer, (3)the strength of the party--how fast are they killing. You're right. This is obvious and dead-horse. I feel like most wars probably would. That if we're talking in terms of "any decent party" by today's standards, war/sam > ridill as far as DPS. I certainly wouldn't say war/nin is useless. It all comes down to when your party is better off having at least one voke with shadows. Anyone who says a voke with shadows will never help your party (Kaparu) is being unquestionably retarded. So utility issue: both subs are good. Dps issue: War/sam is generally stronger.

If there are any warriors with exceptionally big balls, the best thing to do would be a math writeup on full-time-seigan GA/sam vs proper ridill/nin.

[Flokk]

This isn't a dead horse at all. Everybody who's leveled warrior since the two hander buff seems to operate under the assumption that gaxe > everything, regardless of SJ or situation. So it's nice to see the status quo challenged. For my part I still man/rid and even swordchucks (depending on the situation and how many TP moves I can eat obv) from time to time with great parse results. Can't remember the last time I lost to another war/nin (or much of anything really), might have even been Rykoshet like two or three years ago. But I'm not a FFXI math guy so...will wait and see what this turns up.

[Gredival]

Shamaya, on 25 October 2009 - 11:09 PM, said:

Someone's referring to this? kewlll. Well some things need to get sorted out. I've neglected commenting in here and on allah' recently, but there are impending updates to the math. I will be updating some things such as fStr calc, "ws pause", ga food, agressor up/down sets (hopefully), and maybe another thing or two.

You can tell them that this is wholly insufficient as a basis of war/sam vs man/rid. If they want, they can try using my same math for themselves and come up with a comparison that will work, but that's on them. The best thing one could do for war/sam vs man/rid would be to start with the assumption of full-time seigan, and compare the numbers there. I think that just by doing, that, we'll see how close the numbers are.

I think a proper ridill/war with all the most amazing gear and merits is at least as strong (probably a little stronger--my guess) than a full-time-seigan war/sam. Just a guess tho. And when I say that, I'm saying that they'll be close and that it won't require anything near full-time hasso to push war/sam over that. Assuming non full-time seigan, obviously it's completely dependent on several factors. Will the war/sam be more likely to die? Maybe. What's the estimated increased % chance and how would that affect our comparison? Obviously there are two things that are very important; well, 3 things: (1)the amount of hasso being used, (2)the strength of the healer, (3)the strength of the party--how fast are they killing. You're right. This is obvious and dead-horse. I feel like most wars probably would. That if we're talking in terms of "any decent party" by today's standards, war/sam > ridill as far as DPS. I certainly wouldn't say war/nin is useless. It all comes down to when your party is better off having at least one voke with shadows. Anyone who says a voke with shadows will never help your party (Kaparu) is being unquestionably retarded. So utility issue: both subs are good. Dps issue: War/sam is generally stronger.

If there are any warriors with exceptionally big balls, the best thing to do would be a math writeup on full-time-seigan GA/sam vs proper ridill/nin.

I'm actually the genesis of this. It's a running thing with Maverick to squabble about how worthy Fafnir was as an event and Ridill always comes up because the assumption is Ridill is useless post August 07 crutch.

Anyways this is my input on the issue (mostly stuff left unaddressed on the THF Picture Thread debate that never got resolved, which I think is also what inspired the round of testing anyway)

I think it's important most of all to note that the data assumes Mamools, which IMO is favorable for the Great Axe.

The argument has been made that using Pizza for the Ridill but Curry for the Great Axe results in the GAxe actually having the lowest acc value. But I'd wager the defense of the mobs in general will hurt Ridill more because Great Axe is driven by WS and Ridill is much more reliant on DBC (damage between cycles). It has always been the case that the TP formulas mean dual wielding takes longer for a 100 TP cycle than Great weapons do, but the number of hits skew DoT in favor of the dual weapons. The major change of the 2H patch was to make 2H weaponskills great enough to outstrip DBC because Rampage was no longer doing just as much as Raging Rush would. The defense of the Mamools will affect Ridill's DBC which is very underestimated as a damage source.

I'd like to see how the numbers come out if the GAxe used Pizza as well for curiosity's sake but I think everyone ends up playing with birds almost all the time. And I'm fairly sure birds will favor Ridill quite a bit more by letting Rampage spike many more high WS's in addition to a much higher DBC, so if we re-ran these numbers I think it'd be more informative to see them based on Birds.

As far as switching the subs.

1) I am inclined to think that a full Hasso can outdo a Ridill, but that's about as valuable as saying that a MNK with /WAR and full timing Berserk and tanking with counter stance with an outside healer toon beats both possible builds for WAR. I'm of the opinion that we should calculate detriments along with benefits so a model needs to take into account any dead time. All bullshit aside about how a mob should never stay on you too long, Mamools at least can still wreck someone very fast and it is much likelier to happen when you don't have shadows and you're not using Seigan.

2) I'm almost positive a Seigan-only will lose. /SAM without Hasso only adds 15 STP which is basically the freedom to take off STP bodies and Rose Strap and still land a 5-hit+WS. My inkling is that the bonus of switching these slots won't overcome the DoT damage from Ridill. Moreover a Ridill Warrior still beats the Great Axe to the Weaponskill punch by 1.53 seconds in the worst case scenario. (The way I crunched the data is really biased to the Great Axe because I punish the Ridill a full round of delay in its cycle where in reality a large portion of the time the ability to double attack or miss in only one hand will offset the .43 extra rounds of TP the Ridill needs... if you really want to get into the nitty gritty we have to find a way to calculate the estimated times shaved off from both users by double attacks. But that would mapping out tons of possible ways double attack can proc on a Dual Wield Ridill build. I'm not sure if Sham actually did that but I'm not going to try reverse engineer the formula from the data). So assuming Sham's data still holds, it comes down to -30 dmg on WS+DoT vs. 5% more cycles.

3) My personal experience has been, consistently, the exact opposite of every Great Axe defender. Everyone talks about how they outparse Ridill users and I almost never party with a Great Axe user -- even fully Great Axe "invested" WARs whose builds are optimized for the GA -- who end up outdoing the Ridill users. Even with /SAMs and what I assume to be Hasso from the egregiously high MP consumption. Again, this is a generalization, but I never seen anything to back the "consensus" that Great Axe is stronger. I've even seen fresh Ridills beating Great Axe while the fresh Ridill had no Suppa or Sword merits (at Bird camp)

[Tikki]

I might as well throw in. I just like when whoever is playing their job with their chosen combination plays it to its maximum effectiveness or is working towards it. That tends to make the largest difference in damage over someone who noobs it or half-asses it or just happens to ride the bandwagon without actually understanding why.

[solara]

I still disagree with your original numbers on high vs low delay weapon's speed to ws Gred, I even got a second opinion of them back when we were discussing it. Never got around to bringing it up since I had quit for awhile, but I figure it may mean something to whoever does the testing on this.

Your original numbers:
[quote]
Now as an artifact of how FFXI calculates math and when you look at it from the marginal perspective, Haste returns vs. Delay turns out to be a slightly non linear equation which tends to benefit lower delays.

Here's some numbers. The first delay is the delay of a popular set of Thief daggers (chosen to emphasize speed comparative to both Ridill and Gaxe), Ridill is the second delay, and Great Axe is the third delay.

252.5 at 23% Haste = 194.4
252.5 at 25% Haste = 189.3

2% Haste actually gives 2.62% Marginal Haste

412.8 at 23% Haste = 317.8
412.8 at 25% Haste = 309.6

2% Haste actually gives 2.58% Marginal Haste.

504 at 23% Haste = 388.0
504 at 25% Haste = 378.0

2% Haste here actually gives 2.57% Marginal Haste[/qoute]

Kirshy's take:

"To start off, I'll say that the reason there's a different "Marginal Haste" value in the above examples is that there was rounding done. Without the rounding, they have the exact same "Marginal Haste" values. (If you do the above math without rounding they all match.)

That being said, it's very hard to test the actual attack speed value. Using frame by frame testing, I can get very close, but nothing with fractional accuracy. There are multiple animations for an attack swing, and even using a modded attack swing, it varies for each swing. My tests always involve ~10 swings and taking an average (which gets very close.) This opens it up for some different possibilities. One is that Attack Delay is floored at some point. Possibly even by 10s. (So a 504 Weapon Delay is realli 500 delay.) It's also possible it could go up to three decimal places. (504.322 delay.) In general, ~60 delay = 1 second. FFXI runs at ~30 fps. ~2 Delay = ~1 Frame. Even if the frame by frame testing of attack delay didn't have variations, detecting a fractional delay value would be impossible.

In summary, testing shows that Attack Delay is: Delay * Haste. This equation means the same amount of haste is the same improvement regardless of inital Delay. Attack Delay may have rounding that occurs somewhere, but is untestable. This rounding wouldn't affect daggers differently than great axes."

[Gredival]

solara, on 03 November 2009 - 12:55 PM, said:

"To start off, I'll say that the reason there's a different "Marginal Haste" value in the above examples is that there was rounding done. Without the rounding, they have the exact same "Marginal Haste" values. (If you do the above math without rounding they all match.)

Quote

testing shows that Attack Delay is: Delay * Haste. This equation means the same amount of haste is the same improvement regardless of inital Delay. Attack Delay may have rounding that occurs somewhere, but is untestable. This rounding wouldn't affect daggers differently than great axes."

1) Doesn't FFXI conduct math by dropping after the first decimal place, regardless of whether it would be a round up or a round down? I only executed as the formulas demand. Perhaps it's the case then that it does not always work out this way, but it does seem Maneater/Ridill does gain more haste simply from the way the engine truncates numbers.

2) Years ago (like when CoP was just finished) it was commonly accepted that it was more important for already "fast" DDs like Ninjas and Monks to emphasize Haste than other DD's because they got more benefit out by starting at lower delays. It seems like a lot of people who weren't very concerned with gear optimization back then never heard about this, but I swear it was something people always threw around when talking about gear optimization. This was also back when Haste wasn't that incredibly popular, mostly because it was impossible to get as much haste as we can get now. If its not explained with differences "marginal" haste I'd like to know if anyone remembers how it was explained. I got into a debate on BG on this very issue and there were some who agreed with me but none who explained it. I know there was some logic behind this but I forgot how it was explained.

Quote

That being said, it's very hard to test the actual attack speed value. Using frame by frame testing, I can get very close, but nothing with fractional accuracy. There are multiple animations for an attack swing, and even using a modded attack swing, it varies for each swing. My tests always involve ~10 swings and taking an average (which gets very close.) This opens it up for some different possibilities. One is that Attack Delay is floored at some point. Possibly even by 10s. (So a 504 Weapon Delay is realli 500 delay.) It's also possible it could go up to three decimal places. (504.322 delay.) In general, ~60 delay = 1 second. FFXI runs at ~30 fps. ~2 Delay = ~1 Frame. Even if the frame by frame testing of attack delay didn't have variations, detecting a fractional delay value would be impossible.

I don't see why a floor is necessary / we can't simply use 60 delay = 1 second. Damage and delay should be happening on the server irrelevant of display lag on client side software, that should transcend any limitations imposed by the framerate. That's why your damage dealt to a mob as you draw your weapons gets logged even though your sprite gets "Cure Locked."

I'm just calculating the seconds gap between the two weapons by doing [(Delay) x (Cycles Necessary to hit 100 TP)] / 60.

[Byrthnoth]

Gredival, on 03 November 2009 - 05:01 PM, said:

1) Doesn't FFXI conduct math by dropping after the first decimal place, regardless of whether it would be a round up or a round down?

2) Years ago (like when CoP was just finished) it was commonly accepted that it was more important for already "fast" DDs like Ninjas and Monks to emphasize Haste than other DD's because they got more benefit out by starting at lower delays.

I don't see why a floor is necessary / we can't simply use 60 delay = 1 second. Damage and delay should be happening on the server irrelevant of display lag on client side software, that should transcend any limitations imposed by the framerate. That's why your damage dealt to a mob as you draw your weapons gets logged even though your sprite gets "Cure Locked."

I'm just calculating the seconds gap between the two weapons by doing [(Delay) x (Cycles Necessary to hit 100 TP)] / 60.

1 - For some reason I'm under the impression that FFXI tends to round up and round down against the players. So it would round up for weapon delay but down for damage/etc. Unless someone has done the test, I doubt anyone knows, but rounding up for weapon delay would hurt 1H users more than 2H users. Given the absence of proof either way, it'd be more reasonable to just use un-floored or ceiling-ed delay.

2 - For a while there was this theory that 2H users didn't get as much benefit out of Haste because it didn't impact their number of hits per monster unless it was present in higher quantities (not normal in the CoP era). For instance, for having 10% Haste to make any difference to a 504 delay GA user, they'd have to be fighting the same monster for 75.6 seconds. For dual wielders, thanks to DW and lower delay weapons to start with, this amount was lower (10% Haste helped in 61.4 seconds for Man/Ridill). With high Haste totals now though, the argument just doesn't work at all because the time it takes to see a benefit from Haste is lower than the amount of time a monster lives with things like double Marches, Haste, W-turbans, Dusk, Haidates . . .

The rest - Kirschy was just talking about how she measures Haste, I think. Basically (I think), she's saying that can't measure more accurately than 2 delay.