Random Question Thread VI: I'm No Mathologist, But...

[DAKPluto]

Nashira/Marduk feet for SMN BP rage vs. SMN AF2

@311 Skill (not including skill from feet, ignore merits)

AF2 all the way. Hands down. At 311 skill and using AF2 hands/AF legs you should be capped on accuracy on pretty much everything.

[Appie]

Does this look like a decent Guard+ skill-up set?

Cavalier's Mantle
Tiger Mask
Black Adargas
Temple Hose +1
Melee Gaiters
Black Belt
Ochimusha Kote
Jelly Ring
Tiphia Sting
Woodsman Ring
Arhat's Gi +1
Guarding Torque
Coral Earring
Guarding Earring

Anything that should be replaced?

Boxer Mantle, yes it has eva skill, but that doesn't mean shit when you're tanking 3-5 mandies. Equipping -eva stuff just because it has -eva isn't very useful either like your hands, ring, ammo and earring (or other reason for those?) Just grab another mandy if you evade too much. Arhat's Jinpaichi +1 instead of Tiger Mask, the -6% dmg will help you more than 6 guard skill. Second ring, I'd do that Campaign -dmg ring, also Denali hands. I had neither back then, so I just cruised a Phalanx Ring I think in the end and don't remember hands. Pants, I had Temple Hose +1 too, but if you got the Shantotto pants with -dmg it might help too.

If you're dual boxing a RDM for your skill ups with Phalanx 2 you could drop a bit on the -dmg and go with more guard skill and hp. I just had a WHM to dual box. While I skilled, my gear gradually changed from +guard and -eva (much like yours in the quote) to -dmg>guard skill>hp/def. If you do this without something to dual box, then I dunno how the setup would go, I'd assume similar and with a /dnc sub obviously.

[CDF]

Yes "all the time" is not necessarely a clever way to explain things, "on average" is more relevant. What it means is, for example :

without rajas : 13 hits: x1% 14 hits : x2% 15 hits x3% 16 hits x4%
with rajas : 13 hits: y1% 14 hits : y2% 15 hits y3% 16 hits y4%

and x1*13+x2*14+x3*15+x4*16 +1 =~ y1*13+ y2*14+y3*15+y4*16

And if you want to account for rounds it's much more complicated and can't really expalain what happens without doing martingales and such.

The original error (not yours) was thinking that H2H and dual wield only get savings from Rajas "half the time," as though you don't get any benefit if the required number of hits to reach 100 TP is initially even. Your explanation is mostly correct aside from saying you will save ~1/2 a round half the time because you can save only an integer number of rounds some of the time, but on average you will save ~1/2 a round ("all the time" being redundant and uninformative) after considering all the possible number of rounds you might save (considering double attack, non-100% hit rate, etc.).

As for accounting for rounds, you wouldn't use martingales (I don't see where they apply in this case, anyway), but you can use conditional expectation to calculate exactly the expected number of rounds (average) and expected number of hits (average) saved from using Rajas.

[pchan]

As for accounting for rounds, you wouldn't use martingales (I don't see where they apply in this case, anyway), but you can use conditional expectation to calculate exactly the expected number of rounds (average) and expected number of hits (average) saved from using Rajas.

Nope, because you have no way to know if the last hit to reach 100 TP is the first, second, 3rd, 4rth or 5th in the last round. For instance if you only had 2 swings per round, no DA/KA, you would need to know the probability of of the first hit to exceed 100 TP to be odd or even conditionnally to the initial TP return which is not a simple problem due to accuray not being capped, and if you had DA/KA it's even worse. You cannot list all possibilities.

[CDF]

Nope, because you have no way to know if the last hit to reach 100 TP is the first, second, 3rd, 4rth or 5th in the last round. For instance if you only had 2 swings per round, no DA/KA, you would need to know the probability of of the first hit to exceed 100 TP to be odd or even conditionnally to the initial TP return which is not a simple problem due to accuray not being capped, and if you had DA/KA it's even worse. You cannot list all possibilities.

You don't need to know whether the last hit to reach 100 TP is the first, second, third, fourth, or fifth in the last round. It is much easier to consider whether the last attack round gets you to 100 TP at all, regardless of dual wield, double attack, etc.

Consider first the expected number of rounds to reach 100 TP if you need only one (more) hit to reach 100 TP. You can then condition this expectation on the number of (landed) hits of the next attack round. With R denoting no. of rounds and H denoting no. of hits of the next attack round, you could write the expected value as

E[R] = E[R|H=0]P(H=0) + E[R|H>0]P(H>0)

Here, E[R|H=0] = E[R] + 1 because missing completely means you need another round to try to get 100 TP, and E[R|H>0] = 1 because if you landed at least one hit, it is enough to get to 100 TP. P(H=0) = 1 - P(H>0). Therefore, E[R] = 1/(1-P(H=0)), and P(H=0) is easy to calculate. You should see that swings per round, DA, TA, hit rate, etc. are all accounted for when you determine the probability distribution of the number of (landed) hits in any given attack round.

This is the starting point to calculate the expected number of rounds to reach 100 TP for any required number of hits to reach 100 TP.

[Secksi]

Can you find multiple Mog-Tablets without going back to the Moogle?

Find tablet > find tablet > go to moogle get both rewards(obviously using or tossing a kupo ring).

[pchan]

Consider first the expected number of rounds to reach 100 TP if you need only one (more) hit to reach 100 TP. You can then condition this expectation on the number of (landed) hits of the next attack round. With R denoting no. of rounds and H denoting no. of hits of the next attack round, you could write the expected value as

E[R] = E[R|H=0]P(H=0) + E[R|H>0]P(H>0)

Here, E[R|H=0] = E[R] + 1 because missing completely means you need another round to try to get 100 TP, and E[R|H>0] = 1 because if you landed at least one hit, it is enough to get to 100 TP. P(H=0) = 1 - P(H>0). Therefore, E[R] = 1/(1-P(H=0)), and P(H=0) is easy to calculate. You should see that swings per round, DA, TA, hit rate, etc. are all accounted for when you determine the probability distribution of the number of (landed) hits in any given attack round.

This is the starting point to calculate the expected number of rounds to reach 100 TP for any required number of hits to reach 100 TP.

I don't think I understood your notations. R is the number of rounds that lead you right under 100 TP and H is the number of hits of the next attack round ? If so H should always be >0, because if it is 0 it means you were not on the last round before reaching 100 TP.

[Martinius]

in regards to SSR's Citadel Chelonian:

i found a ton of great info on homingmissile's website, but searching these forums has been an exercise in futility. a couple quick questions:

1) can it be pulled with 6 ppl? we normally go mnk/sam/thf/whm/brd/rdm.

2) the first room has 6 single gears. they're not mentioned in the train info... are those killed first, or are they easy to sneak around?

3) as a mnk, should i be subbing war for voke, or is chi blast enough to aggro my particular rampart? along those lines, are flash/dia/bio/elegy sufficient?

4) what's his def/ws like? there's a big difference between genbu def and aspi def. and does it do just songs wipe, or full erase like aspi? admittedly, we'll probably start doing it instead of the boss, so we can just work it down.

thanks

[CDF]

Sorry I wasn't clear. R is the number of rounds to reach 100 TP, given that only one (more) hit is required to reach 100 TP. In other words, suppose you are at the point where you require only one more hit to attain 100 TP. What, then, is the average number of rounds you will need to get to 100 TP?

The next round (where you can get 100 TP if you land at least one hit) isn't necessarily the last round because you can miss completely (H = 0).

[Shalafi]

How many BLMs does it take to burn DM?

[pchan]

Sorry I wasn't clear. R is the number of rounds to reach 100 TP, given that only one (more) hit is required to reach 100 TP. In other words, suppose you are at the point where you require only one more hit to attain 100 TP. What, then, is the average number of rounds you will need to get to 100 TP?

The next round (where you can get 100 TP if you land at least one hit) isn't necessarily the last round because you can miss completely (H = 0).

Ok, well, correct me If I'm wrong. If R1 is the last round with no miss prior to reaching 100 TP, then R is the number of additionnal rounds after R1 in order to exceed 100 TP? For instance if I get 10 TP per hit and 2 swings per round, and if my sequence of hit(1)/miss(0) is
[1 1] [1 1] [1 1] [1 1] [0 0] [0 0] [1 1]
then R1=4 rounds, H=0 and R=3 ?

[Dowzer]

How many BLMs does it take to burn DM?

The one time I did it, 16 BLM, 1 RDM and I was on RDM60~.

All of them died on the first volley, the Galka was left standing and died shortly afterwards.

[Kirschy]

in regards to SSR's Citadel Chelonian:

i found a ton of great info on homingmissile's website, but searching these forums has been an exercise in futility. a couple quick questions:

1) can it be pulled with 6 ppl? we normally go mnk/sam/thf/whm/brd/rdm.

2) the first room has 6 single gears. they're not mentioned in the train info... are those killed first, or are they easy to sneak around?

3) as a mnk, should i be subbing war for voke, or is chi blast enough to aggro my particular rampart? along those lines, are flash/dia/bio/elegy sufficient?

4) what's his def/ws like? there's a big difference between genbu def and aspi def. and does it do just songs wipe, or full erase like aspi? admittedly, we'll probably start doing it instead of the boss, so we can just work it down.

thanks

1)Can be done with 6, your only limiting factor is time. But as you get better with the strategy, you'll get more efficient.

2)There are a few ways of dealing with the gear train. While you can avoid killing them, they will come back, and you'll need to continue avoiding them. You can find a good deal of info from others, I personalli kill the gear train.

3) You shouldn't be using /war on MNK unless you're doing it for greater damage and your support can handle it. If you need to pull with something other than chi blast, I suggest using a ranged item, or the Ichi level Ninjutsu.

4) CC is similar to ada, tortoise song dispels songs/rolls only. Fairly easy to kill.

[requim]

Rajas + tor's is 5str/dex 5stp up and 2-3 acc 3-4 atk down

The hit rate is going to be below a 2% increase in expected damage, the attack might be like 0.5% since it's only about a 1% increase in attack for many jobs, which is like a 0.02 increase in pdif.

5str is giving you a guaranteed fstr for a one handed job who's damage is going to be around 30-60, giving a minimum increase of 60=>61 ~1.67% or a maximum 30=>32 ~6.67%. Not sure what the latest tests on dex show, was it changed back to a constant number or tiered? I'ld eyeball that 5 dex for jobs with pretty high dex would produce the same increase as 3-4 attack, both being negligible.

Rajas ring will on average give you a 2.5% increase in WS frequency as well since it drops a hit for most jobs (certain jobs already use stp gear so would have slightly varied results), but due to the nature of two attacks per round, half the time it will be over.

L/K have an advantage of 1.5~3% for the accuracy and attack differences.
Rajas/Tors has a 1.5~7% increase from fstr and a 2.5% increase in WS frequency, worst case still beating L/K.

Rajas/Ulthalam's is pretty similar except L/K can break about even since with sushi it can get 6-8 difference in accuracy, which at the lower end of acceptable accuracy can push a 5% improvement from hitrate.

Rajas' min 1fstr and 2.5% increase in WS frequency will be giving you around 4% increase in damage before accounting any of the other stats/ring, so in basically the worst case scenario it breaks even with L/K, best case it stomps it. Obviously in some situations you might care about the def/resists, but those are irrelevant for damage comparisons.

[CDF]

Rajas ring will on average give you a 2.5% increase in WS frequency as well since it drops a hit for most jobs (certain jobs already use stp gear so would have slightly varied results), but due to the nature of two attacks per round, half the time it will be over.

I'd like to see the reasoning for this "on average" assertion. No, "half the time Store TP +5 won't do anything" is not one.

[pchan]

R
Rajas ring will on average give you a 2.5% increase in WS frequency as well since it drops a hit for most jobs (certain jobs already use stp gear so would have slightly varied results), but due to the nature of two attacks per round, half the time it will be over.

No, rajas will give you 5% in WS frequency, not 2.5%. Due to uncapped accuracy(95%), the last round will be achieved about equally on the first hit or the second hit (at least it's intuitive). In the cases you get 100 TP on the first hit, rajas saves one full round - not 1/2 which is your mistake. In the case you get 100 TP on second round, rajas has no effect. So on "average" it save 1/2 round. If saving a hit is equal to -5% hits, then saving 1/2 round is equal to -5% rounds since 1 round~2 hits.

[Bakthi]

1. Leviathan's Couse
2. Engetsuto/+1
3. Gondo-Shizunori

I've seen varying opinions on whether Engetsuto or Gondo is better, and I'd like to understand some of the reasoning. I don't see the Store TP saving a hit (unless I'm doing it wrong), even if you're counting on connecting 3 extra hits each time. So it's 2 acc versus 6% crit. The crit isn't helping on Penta, but is the acc really better overall? Unless you're saying "It's so difficult to tell that it's not worth breaking Gondo when Engetsuto is so cheap", which would be fine too.

(I can't actually seem to build a reasonable "true" 6-hit 480 delay polearm set for WAR, but maybe I'm missing something obvious.)

[Darlantan]

I've seen varying opinions on whether Engetsuto or Gondo is better, and I'd like to understand some of the reasoning. I don't see the Store TP saving a hit (unless I'm doing it wrong), even if you're counting on connecting 3 extra hits each time. So it's 2 acc versus 6% crit. The crit isn't helping on Penta, but is the acc really better overall? Unless you're saying "It's so difficult to tell that it's not worth breaking Gondo when Engetsuto is so cheap", which would be fine too.

(I can't actually seem to build a reasonable "true" 6-hit 480 delay polearm set for WAR, but maybe I'm missing something obvious.)

I think the reasoning is something along these lines:

The 2 acc on Engetsuto gives around 1% hit rate. 1% hit rate is benefiting both the tp and ws phase, as well as the ws frecuency thus being about 1% extra damage.

On the other hand, Gondo 6% crit. is only benefiting the tp phase, not adding anything to the ws frequency and having no effect on Penta damage. If the WS/TP is 50% each for instance (which generally /sam will always be leading more to the WS side), I guess you could say the 6% is only working half the time (3%). Furthermore, polearm on warrior is mostly used when fighting relatively weak targets, so the critical hits are not doing nearly as much damage as they would on harder mobs. If a normal hit strikes a colibri for 100 damage and the critical hit for 150%, its only making it better by a 50% so its adding like 1.5% in the end. Factoring a WS/TP split more common of 60/40 and such, and they will come out about the same, and Engetsuto doesnt need to be unlocked, not to mention the +1stp could be useful depending on missed hits from penta and such, if you are trying for a 6-hit anyway (/sam, stp body and strap, rajas, brutal) because if you are just sticking with a 7-hit you are already overshooting stp and it wont really matter (but if you are going with a 7-hit in most scenarios might as well just stick to your G. Axe).

I am no mathologist, but posting in case it is at least minimally accurate and helpful, if its not, please disregard it.

[Mazmaz]

Would you say Rajas/Ulthalam is better than Kusha/Lava for WS like Asuran Fists as well?

[Bakthi]

1) can it be pulled with 6 ppl? we normally go mnk/sam/thf/whm/brd/rdm.

Definitely. This is how we do it:

- Everyone immediately runs to the rampart/gear room upon zoning up.
- Each person is assigned a rampart as you are running; everyone targets theirs and reports that they have done so.
- The THF uses PD+Flee and rounds up the room
- Each person claims their rampart and starts moving back to the beginning of the level. I would have the MNK, SAM, RDM, and BRD do it. I can tell you that it's easy as BRD. You shouldn't lose them if you use distance to not get more than 20' away when they pause.
- The THF dies just inside the gear/rampart room, and gets back up when the gears are gone. They sit in the safe corner in the hallway so we know where the train is. But then we usually have another DD, so you may want them to come back with you.
- Sleep 3 of the ramparts once you get them back to the start room. /as someone. Have the BRD 2hour. Optionally, have the SAM 2hour the 2nd rampart and the MNK the 3rd (or whatever). Kill adds if your sleepers start getting overwhelmed, but try to kill off the ramparts first so the CC rampart will pop sooner.
- Assuming the train isn't in your way, have the THF grab CC and ever so slowly pull it back to the start.

This is the safe way. It costs slightly more time and THF exp than other, more complicated ways, or ways requiring more skill or super gear. There are cooler things you can do if you want, but you might want to try this first. I have never been a fan of killing the train. There are groups good enough to kill all 4 ramparts in the safe corner in like 5 seconds or whatever, but we are not and that's fine. Now that we're used to CC, if we have 7+ people and want LAC, we will have a RDM pull the chariot at the same time we pull the ramparts, kill the chariot first, then kill the ramparts (all in the start room). This can get chaotic, and I don't recommend it until you're used to CC.

SV isn't that awesome on the boss (since it is likely to get at least half-wasted by BJ), so I don't mind using it for the ramparts instead. The timing often works out so that CC is ready to pull when the train has recently headed south past him. Time-wise, our better groups have often been able to do Hammerblow+Powderkeg+GG3+CC+LAC with a couple of minutes left. A lot of it is getting used to stuff, but still, you need a good team to do that.