Sam Gear Lvl 99

Sorry, I was curious in seeing some actual responses instead of 'lol, troll' so I tried finding numbers myself, but I guess it didn't turn out too well. <_<

the ones in your item set. theyre not a true 5 hit. they must be updated occasionally if the shoha set mentions a perle hauberk that's been replaced by avant+1. would say its pretty safe to assume theyre your used sets.

That one was what I used, it works fine enough (1.40 + 0.25 TP multiplier), but to get it nice number from hasso zanshin (bumping from 39.9TP to 40.1TP), I've made some changes :
Ace's Mufflers->Brego Gloves
Misuuchi Kappa->Atheling Mantle
for a total 1STP, which gives me the number 20.2 TP from normal hit and 40.1 TP (2 hits worth of TP) from hasso zanshin.

So you get 19.2tp from your WS set?

Can anybody math SAM/WHM hasting all my friends vs SAM/WAR ?

lim (u->me) ʃe^x = f(u)^n

because math seems to get guys all hot n bothered here.

To answer Vrytreya's issues:


That's not how it works. If you have an average of 2.00 rounds per weaponskill, it means (simplified example): you have a 10% chance of needing 1 round, a 10% chance of needing 3 rounds, and an 80% chance of needing 2 rounds. Out of 100 cycles, you should manage 10 weaponskills after 1 round, 80 after 2, and another 10 after 3 rounds, leading to an overall average of 2.00 rounds per weaponskill.


The number of rounds needed per weaponskill are calculated discretely (for -all- weapon types in all spreadsheets, not just 2-handers), after factoring in all possible WS TP returns (with their probability of occuring), Save TP and, if applicable (since it doesn't work in conjunction with Save TP), Conserve TP (with all possible discrete values that can result from that factored in to the final result). If you get 19.2 TP back from the weaponskill, and get exactly 20.0 TP per melee hit, you need 5 hits to reach your next weaponskill, not 4.04.

The number of rounds it takes to generate those 5 hits is probabilistically determined based on accuracy, DA, TA, QA, OAx and Zanshin. To illustrate: If you have 95% accuracy and no DA/TA/Zanshin/etc., and need to land 1 hit, it will take, on average, 1/0.95 = 1.0526 rounds to land that hit. That is, 95% of the time you will get it in one round; 95% of the time you missed you'll get it in the round after that; 95% of the time you miss that second round you'll get it on the third round; etc. Add that infinite series up and you average about 1.0526 rounds.

The math for dealing with multiple hits per round and the various multi-attack rates and Zanshin and such will of course be a fair bit more complicated than that, but that explains the basic principal.


Llewelyn:

Suggestion for how to get the average non-sphere effect time: Build set 1 with the extra 3% DA (you don't even need the specially modified piece of gear; just do the normal setup and then add 3% DA to the blue (set bonus) line for that set), and set 2 without. Take Nightfyre's suggested 90 delay and put '90' as the time split value for set 2 (on the Setup page); then take the total cycle time from the sam spreadsheet, subtract 90 from that, and put that number as the time split for set 1. Finally, take the overall average generated for total DPS.

Just because it's a silly question... I'm only going to make rough approximations, though.

You and your friends can vary between, say, mnk with 25% haste and no Hasso, Sam with 20% gear + 10% Hasso, to war with 25% haste and 10% hasso. So, 25% to 35% intrinsic haste.

You cast Haste on each person. This provides an improvement between +25% damage (mnk) to +30% damage (war).

The primary issue is that you have to recover 40 MP every 180 seconds for every person you haste. However, we have a solution:

Twilight Mail: 2 MP/tick
Ogier set (4/5): 2 MP/tick
Fields of Valor refresh: 1 MP/tick

You thus need to wear that gear for 15 seconds out of every 180 in order to haste 2 people, 45 seconds out of every 180 in order to haste 3 people, and 75 seconds out of every 180 in order to haste 4 people. Assuming 2 other people besides yourself being hasted, that means you're at lower-than-average damage output 25% of the time.

I also wrote up a solution based on resting and hMP gear (required 1 minute of resting every 15 minutes if /rdm), but this is cleaner. However I'll note that using the resting method, the sam ended up at 60%-65% of their full sam/war potential due to time spent casting and resting, so I'll assume the same 60% damage output before Haste is applied.

However you lose all your gear haste to get refresh, which drops you to about 47%. Adding Haste puts you at 56% during the refresh cycle, and 76% during the non-refresh period. Net average of 71% of your original potential. That means you need to improve the damage output of others by enough to cover that 29% loss.

As we saw above, though, as long as the other DDs were at least roughly on par with your original damage, a single one being hasted comes close to making up for the damage you lost, and two additional DD's being hasted (especially if one of them is /dnc, or even better, dnc main, due to Haste Samba) puts you at a net positive.

So there you go: sam/whm in a party of 3 or 4 DDs of comparable strength to what he'd be as sam/war gives you a net increase in total damage done. Doesn't even count the benefit of using Dia II if you can work that in.

Or you could just bring a whm. Or rdm. Or brd. Or sch. Or pld/whm. Or...

Thats awesome, so when I'm not hasting I should wear Mekria body, unless they need the stp from Radiant Mail.

18.6+1.5 from 2-hit Shoha.

Thanks for the explanation, Motenten.

But I guess I still objected with general public look of taking the end product "average" as absolute.

I mean if you have a fair coin that theoretically has 50/50 chance to roll head or tail and not-so-fair coin that theoretically has 50.5%/49.5% chance to roll head or tail, you will pick a not-so-fair coin on a game whose rule is "who got the most head wins". But in practice, how can you guarantee your win? Given limited number of rolls(so it's relatively finite, not infinite) can you really be secured whatever number of flip you decide you will always an upper hand as not-so-fair coin tosser?

I forsee this quickly descending into accounting for situations like "brb, goldfish caught fire".

There appear to be a whopping 3 people maximum within recent discussion that appear to have even the slightest interest in getting a straight answer out of this and given the way this discussion has progressed, I doubt there's really even any interest there now.

Yes. You can rephrase that question "does probability exist?"

Nightfyre : an answer to what question?
Sylow : and people seems to overshadow or forget about that when they are talking about average. All of sudden the probabilities taken into calculation gone and only the outcome of a single number exist.

That's the whole point of probability theory. Quantitative analysis of the event spaces of random phenomena.

You don't consider the limited number of rolls on such case.
Sure you get a number converged to the calculated number as your roll attempt increases, but I think we are from that on most of the practical cases.

Vrytreya -- Each post you make seems to make less and less sense. I think you need to step back and do a serious re-think of the opinion you're trying to express. You do not seem to have a very solid apprehension of the math involved.

Then please tell me at which point of the discussions my words becoming less sensible, maybe we (I?) can rewind back to that point again and make it more clearly (hopefully) to everyone?

My brain dumb is more now?

How are you justifying low sample sizes as an excuse to dismiss averages in math, yet arguing that something with a 3% chance of occurrence for other players is better for dps?

Maybe it's the 3am talking. Maybe I burned my brain trying to make sense out of the last two pages.

Point out where I claimed that is better. I provided an alternative where, if there's any difference from the suggested optimal setup, it shouldn't be so disagreeable; Which was actually apparent from the other side, as far as I my tone comprehension goes.

?etirtiniodima

It shouldn't be necessary for me to go through your past posts in order for you to be able to form a complete, coherent, and well-thought-out reply. Nevertheless..


This shows that you don't understand the basic fundamentals of probability math.


You then make a sort of hand-wave dismissal of the math, and claim that things work 'differently', but don't provide any explanation of what you mean, how you think the math 'should' work, or why it's necessarily different.


While the sentence is not particularly sensical, it describes what it feels like trying to understand your assertions about the math.


A self-contradictory paragraph. The entire definition of "which one is better?" with respect to doing damage is the exactly the same as "which one is likely to do more damage in this battle". You're saying exactly the same thing, with different words, and claiming to be saying something entirely different. You essentially have a case where you define B to be equal to A, and then say that A is false while B is true.

You then seem to completely ignore the concept of random distributions. I can have 25% Double Attack, and go an entire fight without ever getting a double attack proc. Does that mean I shouldn't have worn the DA gear? No; because it was impossible to say beforehand whether a DA would proc. The only thing I can do is estimate how much benefit that gear 'should' give me, *on average*, over a long period of time.

It's nonsensical to say that any single battle is any different from any other (for any particular target), and saying that because I might not get the benefit of a piece of gear in one single battle that I should ignore its average effect is just....


Of course. That's the entire *point*. A single battle, or even a single parse, will never give you the exact results that correspond to a piece of gear. In testing the crit rate formula, there were parses of 30,000+ hits that were still a good quarter of a percent off of what was determined to be the 'true' value. Again, random, unpredictable effects means you have to rely on the math to determine which is actually better in the long term.


Then you aren't looking at the entire picture, or the people advising you aren't. There's far more than just one number. There's the number you get when you aren't hasted; there's the number you get when you're buffed to zerg, with Soul Voice'd Marches and such; there's the number you get when the mob dispels all your buffs; there's the number you get when you have food/Berserk/Red Curry Bun/Stalwart's Drink/Chaos Roll; etc, etc.

There are some gear comparisons where one particular piece of gear is better than another under all possible circumstances. There are also comparisons where which one is better depends on the conditions you're fighting under. You need to take all of that into account in order to determine which gear to use for any given situation, but it doesn't change the fact that *for* a given situation, one piece of gear will be better than another, and that the way to determine that is to see the long-term average damage potential that you seem so ready to dismiss.


You object without giving any basis for your objection.


Go to a casino and ask how they can guarantee that they win money on such thin margins (I believe Roulette has close to that level of odds between player and bank). The answer is simple: look at the long-term average. Keep playing long enough, with enough people operating under the same conditions, and the bank will eventually pull a net win. Any *individual player* may end up with a large net loss or net gain in a single night of play, but that's meaningless to the overall assessment.


This is entirely what probability deals with. If you don't understand the fallacy of the assertion you're making, you won't see that pretty much every conclusion you draw from it is wrong.


All of the probabilities together add up to that single number. That's the entire point. Saying that the average is some magical value completely divorced from the probabilities indicates that you don't understand the underlying math.


This makes almost no sense whatsoever, but if I were to try to guess what you mean, you're falling victim to a variant of the gambler's fallacy. You're apparently trying to say that a small sample taken from a larger distribution will behave differently from the larger distribution, and in a predictable manner. That is intrinsically self-contradictory.


Then you should have stuck to that argument: "This is what I prefer; it's slightly less than optimal (under conditions a, b and c), but not enough that I think it makes much difference overall, and I like the shiny." Instead you tried to attack the math behind the opposing argument without really even understanding it.



Also...


What the heck is 'DT' supposed to stand for?

Damage Taken.

VZX arguments about Discrete and Continuous Data:
I've always learned it as Discontinuous/Continuous data. The concept is that when you make a hit, the damage is discontinuous. So if the delay of your weapon is 3 seconds and you'd deal 100 damage, you can not claim that after 1.5 seconds, you'd deal 50 (it would be 0). I could imagine this would cause problems when calculating DPS because if you come up with a fractional number even if you had calculated the probability of hit builds discretely, can you really linearly correlate a 1.1 attack rally to damage and if you do, over what type of sample size will you need to get it to this point with reasonable confidence? --If you have an explanation to this, I'll be interested in knowing by PM as this is off-topic.

Noise to data:
I think the both sides are both correct if the calculations are correct but they just have different opinions on this subject.

It's true that DPS can maybe give you estimated "What is deemed better" answer, but it is also true with VZX that no parse has ever demonstrated that number. How long must a fight be before you can confirm that the 3% or even 10% of pure "DPS" is relevant? We all realize that sample size is important to converge at that average-- but how much do we have to fight in order to see this-- and will it even matter over such a long sample size? Can you be confident that after a 3 min fight with a VWNM that you will always be 3% ahead?

If what is said is true, that it takes 30,000+ hits to be .25% off of the true crit value, what can be said about a 3 minute fight with a VWNM? How much variation will there be between a parse of the same person? How many fights must you do before you truly see the utility of a 3% advantage?

I think this is VZX's argument-- and I don't think that they are as shallow as the forum makes it out to be. While it may be nice to know what gear may allow you to perform your best mathematically, playstyle and noise is probably to cover any of those minute advantages you have on any content that is relevant in game at the moment.

I don't understand what is really going on your spreadsheet, I'll give you that.
Without getting my reply any longer than yours:
I guess this is where the point I really want to stress : long-period of time.

Most of the parses people want to show is from 1, or 4 battles from VW. I can't speak about Dynamis since I don't do it. How come one can be so damn certain on dictating which gear to use in order to give boost on this small sample/subset of data? "Well if you take long enough..." "No, you don't." In practice all these nice things breaks by the external factors you can't control. i.e. you miss 3~5 secs because your haster is busy with other stuff, no more temp so you have to switch to pdt mode, mob running around, and all. How one should be strictly concerned to the lose of 7 dps?

Really, that's my whole point.

Instead of getting a proper response like "why?". The response was a ridicule to my such position, even a false statement like "you are not getting 5-hit from it".

I don't intent to attack the math behind opposing argument. My hope was for him to explain step-by-step of how my option is suboptimal. He throwed away the number coming from your spreadsheet and say "Well your option isn't doing so well compared to the other option based on this". Now to response on such argument I must completely understand what the heck is really going on your spreadsheet, mustn't I? It's like putting a huge barrier of dismantling your stuff (when even he probably doesn't understand it completely). It's mind-boggling.

I'll concede on the average 2.00 round per cycle part. It wasn't a careful analysis coming from me.

I will stop here. I think you misunderstand what I understand and what not. I understand and agree with all those "on the long run" concept, but again, that isn't the problem.

The argument is flawed in the greater sense of what is good or isn't. We, or some, strive to be the best mathematically. We understand that doesn't always translate into winning a parse-but that's not the overarching goal. The goal is to kill your target faster, and the only way to ensure doing that is by looking at the equation and filling out all of the variables with the best options for the final product.

If you have Avant Mail +1, should you completely forego Ace's Mail? If you had Askar Body at 75cap, would you never try to get Usukane? Or Ares's? If my favorite job is Samurai and I own a Kikugosaku, do I quit there or do I attempt to obtain a Masamune, TPBonus, Amano, etc? If the final question isn't "what performs better" regardless of the modicum of increase then the question probably isn't even being asked - by all others who do, a 0.5% increase is still better than no increase.

To this specific argument, does 3%DA really help? ***'s situational, yadayada. If you're really trying to min/max, then 3%DA is probably not the greatest benefit your super well geared party members need. Fighter's Roll, Warrior's, Thief's, OAT users, Mythic wielders, Samurai's - all jobs that aren't benefiting hugely in TP or WS phase by 3%DA as much as you might be by using proper gear and WS'ing more often, harder, etc.

There are many and more significant way to achieve such goals (killing enemy faster). Ensuring your other PT/alliance member knows what gonna happen is the one of the most significant way to get stuff executed beautifully.
Caring about the 0.5% increase of my damage while unable to give detailed direction to other people how the fight will go is backward prioritizing. Even if they perfectly understand how it gonna go, caring improvement of such amount is like caring whether there's a slightest poisonous particle on the air I breathe wherever I go in order for me to stay healthy.

"We kill the target 10 secs faster this way"
"So *** what? That guy will take 10 minutes resupplying"

After sleeping off a nasty headache and waking up with the Caramelldansen song running through my head, it suddenly sprung to mind that 'DT' was referring to Damage Taken (I have no idea how Caramelldansen is related to damage taken, so I'll put it off to being able to think more clearly without the headache). I'd completely forgotten about that trait on the Mekira body.

So, that little bit of stupidity aside..


Vry's use of the terms is correct. Discrete means that it's impossible to get "half of a hit" worth of TP, for example; you either get a hit or you don't. While discrete data will tend to be discontinuous when graphed, discontinuous data is not necessarily discrete.

Discontinuous is an entire different matter. The graph for tan(x) [tangent of x] is discontinuous at x=0 because tan(x) approaches positive infinity from the right side and negative infinity from the left side. When you actually hit 0, the line is at an entirely different location depending on which side you approached from.


You're falling into the same fallacy as Vrytreyan here. A single fight will never match predicted DPS because a single fight can generally be considered just one sample point (depending on the approach you're taking on the analysis; you could also see it as as many sample points as there were weaponskills performed, based on the fact that spreadsheets calculate things in terms of weaponskill cycles), and a single (or any number of, really) sample point is subject to margins of error. Even if you consider it multiple sample points, a few dozen weaponskills still has a massive margin for error.

That sample point is only required to fall within certain very finite bounds. For example, given a certain attack/defense/mob level/etc (ie: cRatio), you know that a hit *must* fall within a certain range (unless absorbed, or a physical shield was up, etc). On the other hand there is no absolute requirement that you must even be able to hit the mob. Since there is always a chance that you can miss a hit (5%), there is literally nothing that guarantees that any of your hits will land (barring Sneak Attack).

Therefore all you can say is that "these values must fall within these bounds" and "these other values will be distributed along this probability curve", and then combine them and sum up all the probability values to give you a final 'average' that tells you the overall result that you will tend to get as your total number of samples increases.


Actually, 30,000 hits would give something like a 0.38% margin of error. Regardless, you're looking at it backwards. That sample size is necessary to pin down the true underlying formula. However once you have that formula, sample size is irrelevant. You now know what the distribution will be shaped like, so it's merely a matter of aknowleging that each following sample point is within that distribution curve.

The entire point of randomness is that you can't predict what the next value is going to be. If you could, it wouldn't be random. However you can say what it's likely to be, and you can say what the average result will be given the full possible range of random values that can occur.

Saying you might do better or worse than average in this particular fight is meaningless nonsense. Again, if you could predict the results, it wouldn't be random, and most of the underlying mechanics of the game rely on the random number generator.

So, how many fights do you have to parse to see the results of a 3% advantage? All of them and none of them. All of them because any one fight will never show you that exact 3% advantage (even leaving aside the changing dynamis of the fight). None of them because you've already calculated the probabilistic result, and that advantage will tend to show up over the long term, whether you look for it or not.


Perhaps not shallow, but definitely flawed. They're the sorts of questions any professor of a Probability & Statistics class will have to answer over and over, as each new set of students fails to understand the basic principles.


This is an entirely separate issue. It means you're not accounting for the preconditions properly. Yes, Haste might drop while the mages are busy, so you need to determine the effect of this gear change with and without haste. Yes, Dia II/III might not be active on the mob (such as when Bio II/III procs need testing), so you need to consider the effect of this gear change at different effective cRatios. Etc.

You can certainly argue that people are not comprehensive in their analysis of these issues, but that's different from saying that the 'average' is a useless value for any given set of preconditions.


Will get to Vrytreya's points in the next post.

Because those are two entirely separate points. A parse is an example of what happened, and is only loosely illustrative of the underlying math. Those who are currently giving out more certain recommendations are doing so based on the math, not on the parse.

In the early years of FFXI, parses were the definitive way of showing what was 'better'. This was partly because people who actually understood probabilistic effects hadn't put their hand to helping work things out, and partly because many of the basic formulas were poorly understood. This led to common flaws in recommendations, such as the infamous "big number syndrome".

However we now have a very solid understanding of the underlying math. There are still small uncertainties in some areas, but you haven't even approached that in your arguments.


More straw man arguments. The issue is not that the recommendations are wrong, but that recommendations are not complete.

Most recommendations do not try to account for every single precondition because it's a lot of work, and you're doing that work for someone else who isn't willing to learn how figure this out for themselves. Plus, anyone who isn't doing these comparisons for themselves probably don't have the wherewithal to regularly adjust their gear to properly meet these changing conditions anyway.

As such, you check a few basic 'common' conditions and make the recommendation based on that. The larger the difference, the less likely that different sets of preconditions will yield different results.

If someone shows interest in a wider set of preconditions, you can expand on the recommended comparison. If someone who has a decent understanding of these comparisons still asks "Why?", you can get into the underlying math that supports one or the other, and what about each piece of gear leads to the final conclusion.


This is a common response on most forums, and while I do not advocate it, it is very easy to bypass. Simply explain your reasoning up front.

If you make your statement along with the reasoning behind it, people will respond by addressing your arguments. However you did not do that. You made some offhanded comment and dismissed any counterargument. If you behave in that manner, people are going to be contemptuous right back at you.

They are under no obligation to ask you "why?" when you make an assertion contrary to popular convention. The burden of proof is on your shoulders, not theirs. If you make no effort to provide that proof, they are under no obligation to treat it seriously.


This is an issue of trust. He apparently trusts that the work I do in creating and maintaining these spreadsheets will yield a reliable answer, and is likely influenced by others whom he also trusts. If he trusts A, B and C, and they all trust D, it's natural to expect that he would also trust D, unless otherwise given reason not to. Your assertion, however, is not supplemented by that same degree of trust, and is further harried by your flawed logic and confrontational approach.

I do my best to ensure that that trust is deserved, and regularly work to improve and fix the various spreadsheets I maintain. If I had simply tossed the first few out a year ago and ignored them since, I believe it would be much harder to trust that they were reliable sources of information.

If you do not trust the results of the spreadsheets, then you can either ignore them, spend time learning how they actually work (including understanding where they can be wrong and why), or provide some sort of alternative to them. If you're not willing to do one of the latter two, then it's understandable that people are not as likely to believe your own unsupported assertions.

Now, I can understand the reluctance to trust something you don't understand when you actually do want to understand it. Unfortunately there's no easy answer to that. You need to put in the time, effort and energy to learn it; there's no free ride.



Those are two utterly unrelated activities; another straw man arguement.


Another straw man argument. Someone taking 10 minutes to resupply is a leadership and organization issue, and has nothing to do with the fight or what gear you're wearing.



I'm not entirely sure which gear sets are being compared, but decided it's likely Vry's "Sam - TP, Defensive setup" and "Shoha - Current set", as the others are for a 4-hit Kiku.


Decided to put together a more varied comparison.

Set 1: Mekira head/body, Brego hands
Set 2: Unkai head/body, Ace's hands

Other gear is the same between the sets:
Thew, Unkai neck, Brutal, Unkai ear, Rajas, Hoard, Misuuchi, Goading, Unkai legs, Ace's feet

Target: Pil (lvl 110, 560 def)



This shows how the different sets vary with different preconditions. There are aspects that close the gap in terms of percentages, while at the same time the higher the buff level, the more in absolute DPS needs to be made up by the sphere effect.

So then you need to know how much DPS is gained due to 3% DA by someone else under similar conditions.

Ran through several variants using my own gear, and had mnk improving by 2 to 4 DPS depending on buffs/debuffs. (was annoying to do the comparison, so not filling out a chart here)

Given that you're sacrificing 5 to 12 DPS of your own, you'd need to be buffing 3 other DDs to break even overall.

Since that is not an impractical requirement, and you receive other benefits from using the Mekira body (PDT), it's perfectly reasonable to say that it's a competitive option under some conditions. This is even more the case if you're doing more proc duty while the others are focusing on damage, since boosting their damage is more valuable than the lesser reduction in your own damage (and vice versa if they are focusing on procs while you focus on damage).

So, properly define the conditions and circumstances you are working under, and it makes it much easier to explain when and why one piece of gear is an improvement over another.