Random Question Thread (FFXI Related)

Well. It's an underground lake that connects to the bastore sea. Saltwater lakes are a thing.

An underwater cave in the elshimo island that opens into the grotto.

Fantasy.

/shrug

Rostam Cors:

What Resin augment do you use on your capes, since it seems that PDT-10 is pointless due to the absolutely fat stacks of DT rostam(s) give us.

Even with just a single rostam, it looks like all I need is my normal "low DT" set to already cap my DT-, with 4 rolls on me.

Counter+10 seems nice on cape as a way to further mitigate stuff when Im soloing content.

Meva+15 however I could see being better for when I'm not soloing and need to deal with mitigating AoE nonsense.

Or is there a better meleeing TPing setup where you keep 10 PDT on cape and theres better options than Malig Chapeau + Empy Legs + Carmine Feet?

Cuz all 3 of those seem to be best anyways and their DT is just an extra bonus.

Which already puts us at 21 DT, and then Null Loop puts us at 26. At which point a single offhand rostam already caps our DT with 4x rolls (whether its a party situation or soloing with qultada)

So, it sorta just seems like PDT on the cape is not necessary?

Or do you keep the 10 PDT on it anyways as a buffer for if rolls ever get removed mid fight or something (IE aoe dispels and such)

DT as you wont always be wearing the rostam unless you're making a cape just for when you use a rostam, in which case, regen or something. Meva is already more than covered with null cape, could do counter i guess

Saw a cute little video the other day that did a little example of multi-attack with a breakdown for Quad, Triple, and Double attack %.

They showed the % for multi attack chance ticking down, raising the change for the later checks.

I was under the impression all checks happened at once, and the highest priority success is what you get.

Is that wrong?

The video provided an example of:
3% Quad, 5% Triple, and 17% datk.

Claiming if Quad check fails, Triple rate rises to 3.1% because it's rolling 3% triple out of 97% multi-attack chance remaining.

Then datk % chance is actually 18% because of it's rolling 17/94...

I'm pretty sure that's wrong, but I've got 5am brain, and don't really wanna rag on the video, cause it's a nice cute little basics on what to put in your TP sets. Just curious, because it would implay WARs don't have to gear to 100% datk if using Niqmaddu's and such.

Cute video was straight up misinfo, and blatantly wrong. I would ignore everything they said because they don't understand how ffxi works.

That doesn't sound right.

Assuming single wield to keep it simpler, a quad attack proc will mean no other form of MA can proc on that swing. Idk and it is not possible to know if the checks happen all at the same time or not. But we do know that QA takes priority and means no TA/DA on that swing, and if no QA then TA takes priority over DA and so forth.

That doesn't mean your DA rate is any less, just that a double attack proc and a quad attack proc cannot happen at the same time on a single swing.

If you remove a hypothetical quad attack proc from a series of 100 swings and you come up to 96, which is also wrong, your DA rate is still 18% out of 96. In the remainder 96 attacks you may get 17 DA procs. May. Or may not. Because pseudo-random. DA percentage is out of 100 not out of 96. Your DA rate is still 18%.

But you're not only swinging 100 times, you're just swinging. The more you swing, the more your DA rate approaches but never reaches exactly 18% because you have other forms of MA that take priority.

Statistically if you roll the dice/swing 1b+ times your DA rate will be very close to 18%, your QA rate very close to 3 and your TA rate very close to 5. Not exactly because that's not how pseudo-randomness works.

If you swing 1000x with 20% TA and 20% DA it is normal belief that it results in:
200 TAs
160 DAs (800*0.2)
640 single swings

No?

20% QA 20% TA 20% DA would give:
200 QA
160 TA
128 DA
512 single swings

I understood it the same way K123 put it. Each rolls in sequence, so higher multiattacks mask lower multiattacks mask OAT. You can model it with binomial probability.

Suppose it wouldn't be that hard to test; throw on a low damage weapon with an action packet logger and known amount of multihit against the test dummy. Shouldn't take that long to get a conclusive result.

That does seem likely.

The only reason I have doubts about it is due to all this having to happen within a swing, and the server has to calculate them and feed that back to the client within that swing/single packet for that swing.

If I were developing this I would pre-roll everything then prioritise based on rules. This in order to reduce overhead.

If you roll once for QA, wait for the result then roll again for TA wait for result and repeat that for how ever many MA rolls you need that's adding a ton of overhead to the single packet generation you need to tell the client how many swings that was.

That's not going to scale well as you add characters and max haste swings.

So I think it's more likely the rolls are already known, there is a priority order, and the code just exits out if it gets a proc on an MA based on that priority order.

It's been done, that's how it's been found to work.

The post/test is already here somewhere.

Pretty sure you code, so I'm not really sure where you're getting the idea that multiple rolls would be a problem or delay. Obviously implementations can vary, but standard Mersenne Twister implementation is not a particularly expensive operation. What result do you think they're waiting for?

I'm saying it doesn't matter what implementation they use (they have their own random number gen as seen from the old cases bug they had).

It's simply a complexity issue. Get rolls for all things then decide what you need, or roll once per possible MA.

Do X operations all at one time, then decide.

Or do X operations one at a time as you need them.

Doing more things at once is more efficient than one at a time.

Even things like multiple function calls will add overhead. Not a lot sure, but still. Depends how many times you're calling a function whether the overhead matters to you or not.

Whats the scenario that exists where:

1. "I care about 10 PDT on my cape"
but also
2. "I don't want to offhand a Rostam"

Intersect? What else would you be offhanding instead?

I guess yeah, the better question would be "Would I rather use ambu cape, or would I rather use Null Cape instead"

What Dodik is saying makes sense but you'd have sufficient data within 2000 swings to know. Just set up a THF with equal amounts of TA and DA and hit a urchin outside Adoulin with D1 daggers

I don't think I've offhanded a Rostam in a very long time. It's an extremely bad OH.

I offhand Gleti's Knife most of the time. Crep Knife is another good option, Tauret if you want MAB. Blurred Knife. Kustawi for ranged attack/accuracy.

Rostam is godawful for an offhand, it offers very little.

How does the failure mechanic work in oddessy boss mode?
Not the rp mechanic.
The retry mechanic where the boss does not gain HP and you can try again if you wanna burn another ki

Is it limited to job cycling?
Party leader?

More a case of either:

* Main handing rostam
* Off-handing Gleti's or something other than rostam

There are no reasons you would want to offhand a rostam. The augments don't work in OH.

If you're shooting you probably want a rostam with storeTP augments, or you're using a different main hand weapon. Tauret or naegling are better options than OH rostam if you're shooting.

Also Cor changes weapons very frequently. You should build your sets to work with multiple weapon sets or you'll end up with 25 ambu capes not 10.

I have no idea what you're talking about. Can you clarify what mechanic you're talking about?

Maybe it would be better to call it a retry mechanic

I saw on a video where a group re entered arebati and it's health did not go back to 100%

So I was curious if it's set at thresholds and what the conditions are for keeping it from resetting

Every 5% HP you take off from a boss (inclusive, so really it's 6%), that HP does not regen if you re-enter.

If you take Arebati to 94% and lose, you enter second KI and he starts at 95%.

If you take Arebati to 69%, lose, and he regains to 100% HP because the add is out and you were dead for 5min, he still starts at 70% if you re-enter on another KI.

Same rules apply for staying in Gaol. Cannot use same jobs again on subsequent KIs.

This is basically what all multi-KI strategies do. Every 5% you take off is permanently taken off that guy (in your set of 3 Moglophone IIs). Note you do need to reach the next lowest %, so basically 5%+1 HP.

90-94.99999% HP? It will appear at 95% in your next KI.
85-89.99999% HP? It will appear at 90% in your next KI.

On and on forever, down to 5%.

You have up to 3 attempts to kill it, but job restrictions mean this is rarely relevant, unless you pre-planned your jobs and 1hrs to make it possible.

Worth mentioning: Once you've attempted an NM at a vengeance level, you cannot change the vengeance; you can only attempt it AGAIN at the same vengeance level, unless you exit and reset the whole thing.

As stated, OH rostam is pretty poor, best guess is it makes a very high macc offhand though i cant see a large case for that beyond light/dark shot. You use it in MH so you can assume 6-12% DT as you're unlikely to be roll-less, so in your armor slots(maligx4 and emp legs) you're at 34 DT or 40/46 with 1 rostam. so a single resin aug or a d.ring will do it for ya.

this just doesn't make sense to me, because any modern compiler will exit statements early(short-circuiting).. you don't need to generate any attacks before deciding how many you will have, you can still do all the attacks at once

pseudocode:

can you provide pseudocode of how your 'one roll' method would look? i'm just not getting how it would be any simpler or cleaner, every implementation i visualize is sloppy and hard to mantain, just to potentially save a couple random number generations

I was thinking something more along the lines of:


RandomNumberGen takes a set of numbers and a set of probabilities, eg [0.03, 0.05, 0.10] for 3% QA, 5% TA and 10% DA respectively, and generates a random number from the list based on that probability on each call to '.next()'.

Meaning there is a 3% chance it will return 1, a 5% chance it will return 2 and a 10% chance it will return 3 in the above example.

This means you can reuse the same lookup table (RandomNumberGen) for the same probabilities, and also possibly pre-generate it for all possible probabilities if you want close to realtime performance. Lookup for an already generated set of numbers is obviously going to be O(1).

Probabilities don't change for a character unless they change gear.

Ideally you cache the RandomNumberGen per set of probabilities, per character gear set, or pre-compute all of them, something appropriate.

The goal is to reduce overhead on subsequent calls to calculate_attacks() for the same probabilities (MA rates), because this will be used very often and needs to scale.

See Alias Method for more details.

The key here is the first half of the scenariom "I care about PDT on my cape"

Which inherently means you care about damage taken.

Which inherently means you are taking damage, enough to matter.

Which is typically one of:

A: You are soloing stuff and thus tanking everything
or
B: You are in a party but theres a buncha AoEs so wearing DT gear means you either avoid dying outright, or at the very least make the healers job a lot easier.

IN these scenarios, I'm like 99% confident Gletis knife is a bad offhand.

The reason being that even just getting 12DT off Rostam frees up a tonne of DT on your other slots

For all intents and purposes, Rostam totally changes your idle set up, because now you can use swap your body/gloves/rings/back/belt around to better options.

1. The +12~+24DT means you can swap malig gear out for adhemar
2. The 6DW on adhemar body means you can use DA cape, Gerdr Belt (I believe this comes out ahead of using Yotai)
3. You can use Chirich+1 and Epona's on rings

You effectively gain a dump truck of TA and STP, because your slots have been freed up a fair bit.

The downside is you lose 85 attack on your WS, which I'm not 100% sure is worth this trade. You effectively gain a substantial amount of faster TP rate so you can WS more often, but the 85 att on gleti's knife is nothing to scoff at.

I have to sit and actually run the simulator on it, but atm my gut says that the far faster TP rate (And thus more frequent weaponskills) should compensate for the 85 attack loss.

Especially since we probably already have a ***tonne of attack as a corsair, especially if either using Sylvie solo, or being in a party with a brd+geo

So Im not certain the 85 attack loss is a huge deal.

Stupid question, but is there a minimum land rate for spells cast by monsters on players?

Kinda like physical accuracy can never go below 20% rate, does the land rate of spells cast by monsters on players also have a minimum threshold under which it's impossible to go no matter level difference, meva or elemental resistance?

Possibly, but I think your premise is off.

If you're doing dps on Cor you are doing one of two things:

* Shooting, ranged WS
* Meleeing, either ranged or melee WS

If you're shooting you probably want a rostam MH, yes.

If you're meleeing you could use a path B rostam MH. But that's not great for melee WS (savage). Great if you're doing leaden though.

Melee WS needs naegling in main hand. Rostam OH with naegling main hand doesn't give you anything. Gleti's/crep OH with naegling MH does.

Yes, an OH rostam would save you DT in other slots. But there is not anything an OH rostam is useful for in circumstances where you are doing DPS on cor.

So start with the premise of, I am either shooting or meleeing for TP and go from there.

In the case you are shooting and happen to OH a rostam because no better OH options then sure, can swap some stuff around. But that's pretty niche.

I wouldn't use adhemar body unless it was casual content
I also change weapon combos based on the weapon skill I am focused on
Gleti and savage blade are great together even unaugmented
Crep is an ok runner up the additional effect is entertaining but I've come to need the 2 haste and the ta

But if you are talking building around ranged weapon skills
Yeah a rostam main hand is great having one off hand for the price is crazy not to mention what you are giving up.
A/B fantastic tp gain tools the DT is just a nice little bonus