Random Facts Thread: Now with easier searches.
Random Facts Thread: Now with easier searches.
May need more tests, but this at least rules out the old formula:
- Fun background story on how I found this value
- The expected values via math
- The test results
- Overall point: Thunder Breath's formula is 27/40HP. The existing (HP/1.4823) value is incorrect.
Background and development of the formula: Boring for most people. I'm just someone interested in how people come to create, reject, or test ideas, so if anyone shares that interest, this would be for you. Otherwise, feel free to skip this spoiler.
Spoiler: show
The math comparison: Proposed (27/40HP) v Existing (1/1.4823)
- You get a difference in values at 1550HP and 1560HP
- Proposed formula: (27/40)*(1550) = 1046.25 = 1046 DMG expected
- Existing formula: (1550/1.4823) = 1045.672266 = 1045 DMG expected
- Actual damage: 1046 damage
- Proposed formula: (27/40)*(1560) = 1053 DMG expected
- Expected formula: (1560/1.4823) = 1052.418539 = 1052 DMG expected
- Actual values: 1053 damage
Pictures
http://imageshack.us/photo/my-images...226071855.png/
http://imageshack.us/photo/my-images...226073435.png/
In short: Thunder Breath's formula is (27/40)*(HP)
I think this belongs here, especially since it doesn't function how you'd expect (least not imo)
Absorb Attribute overwrites BRD songs if you steal a song from a mob. Had a stolen Minuet overwrite a Troub'ed March in Dynamis Xarc the other day which was pretty disappointing as I was expecting it to just add a 3rd song.
Would it be possible to check what happens if you have two BRDs?
I'm at work at the moment but I'll check after LS events tonight. I suspect it will overwrite whatever song on you has the lowest duration remaining.
The only possible reason I can think of is it's a safeguard for Empy Harp coding in some way and having a third song on you would allow you to keep 3 songs up permanently like how you only need to sing the original songs with Empy Harp and can then keep them up with whatever instrument.
I got inspired by Slycer's pretty, new MB page and tested some MB damage stuff last night/today.
Apparently there's a +40% limit on MB damage from equipment/atmas/merits. Job Trait surpasses this cap and they all sum into the same term, but the other three sources combine to cap at +40% MB damage. In conclusion, I really don't understand SE. This is like when they capped Waltz potency at 30%. Why would you bother?
Also, Chelona Trousers and AF2+2 hands are pretty great during MB. That is all!
Someone speculated that Spirit damage and Breath damage were the same, but:
Spoiler: show
Used 1590 HP. Looks like a 45% reduction for breath damage. That could be 1/2 resist and 1.1 bonus (Depends on how order of operations works), but looking at Zantetsuken damage, that should be 1.3 multiplier. I couldn't find reports of the damage bonus atonement receives.
Edit: Could also mean Atonement is the odd one out since atonement is atonement is weird as hell.
Also, need this explained:
Spoiler: show
So I did asome stun tests on the TEST server. I was level 50 DRK / 25 SMN and spammed stun 2 nights in a row on Om'hpende and I actually got 1 "resist" message.
casted 711 stun, successful 710, resist 1 (parser)
I had to actually go to the parser's log and confirm the "resist" message.
Going by the logic that stun follows normal resistance mechanics (partial resists, caps)
with no partial resist, the chance to "resist" are 0.05= 1/20
with 1 partial resist, the chance to "resist" are 0.05^2= 1/400
with 2 partial resist, the chance to "resist" are 0.05^3= 1/8000
So it's safe to say that stun has 2 resist states.
You can say it's not 1/20, but you need more data to say it's not 1/8000.
Sentinel's Scherzo seems to only work on single-hit monster TP moves. We've been using Scherzo in Legion instead of PD lately, but I've noticed myself taking >75% damage with it up from the Naraka's 4-sword "attack", Damnation, and Ig-Alima's 4-punch combo. Even though these things show up in the log as a single damage number, Scherzo is individually applied to each hit. That's probably always true, considering the testing for it relied on single-hit monster TP moves.
This unfortunately limits the utility of Scherzo. I was hoping to rely on it to circumvent SE's broken damage mechanics, but no such luck.
While the delay between spell casts is known, I'm not sure it's ever been quantified (or I haven't seen it if it was). Was doing some testing regarding spellcasting for the purposes of analyzing the effect on magic procs in dynamis. Found that the minimum time between the ending of one spell cast and the start of the next is roughly 3 seconds. This has a significant effect on the maximum rate of casting spells.
Here's the writeup of my little experiment (was in reply to a poster on OF):
Did some testing.
First, basic testing agrees with Yinnyth. I was able to consistently get 12 threnodies per minute once I got the recast timing down.
So next I tried to figure out why.
FRAPS analysis of casting time.
I don't have a perfect brd gear set; missing 6% in cast time, and don't have Minstrel's Ring. End up at 40% total.
Changed subjob to /rdm for the extra fast cast to see how it would affect things. That puts me at 55% fast cast.
After adjusting to the slightly different timing, I was able to get 14 spells cast per minute.
14 spells per minute is 4.28 seconds per spell. Can figure the 0.28 is human error, in that I can't cast the next spell at the perfect recast moment, and it's a 4 second recast on a 1 second spellcasting time (after fast cast is applied). That's a 3 second casting delay.
To test this effect further, I retried using Minuets. Nominal cast time is 8 seconds.
With /rdm and fast cast gear, time between start of Min4 and Min5 is 7.233 seconds. Actual cast time should have been 3.6 seconds. Difference is 3.6 seconds; with allowance for human error, we still get the 3 second delay.
Tried again with /whm and no fast cast. Time between start of Min4 and Min5 is 11.367 seconds. Subtracting the nominal cast time leaves a 3.367 second window, for 3 seconds plus human error.
Overall conclusion: Minimum delay after the completion of one spell before you can cast another is 3 seconds. Using the fastest casting spells available, plus accounting for human error, lowest realistic time per spell cast is 4 seconds, allowing for 15 spells cast per minute.
Just as a test of principle, did you try seeing how many songs you could get off during Nightingale? If it has 0 casting time, you'd predict ~19 songs including JA delay, yeah?
Also, we don't know whether the casting time sources are additive or multiplicative. If they were multiplicative it would just subtract a little from your margins that you attribute to human error.
I've noticed sometimes I melee between spell casts even if I'm essentially spamming the spells. What about this?:
1) The first JA Delay tests spammed JAs while meleeing and used the melee hits as a measure of the delay added by the JAs. Furthermore, they showed that the delay for melee hits following a JA usage (2 seconds, soft delay) was different from the delay for further JAs following a JA usage (1 second, hard delay).
2) Your test Casts spells and takes the next time when you can cast a spell as a measure of the delay added by spells. It doesn't use meleeing or JA usage to measure this.
3) It's possible that casting spells gives the same set of delays as using a JA, but there is a third category of delay specific to spell casting that wasn't tested in the original tests.
Quite possible, yes. I didn't do the melee testing because I wasn't trying to see the effect of the delay on melee, I was trying to see the effect of the delay on the next spell; essentially similar to trying to find the hard delay of 1 second on consecutive JA uses. Finding the effect on swing delay would be useful for blu/rdm/etc melee, though.
Hadn't tested with Nightingale. Will try that...
I currently have 5/5 Nightingale merits, which means instant cast on all spells for the duration. Disabled Spellcast, and just kept as much song recast delay and haste gear on as possible.
First attempt: 16 songs with several mistakes.
Second attempt: 17 songs with a couple small errors, but possibly sloppy timing.
Third attempt: 17 songs with a couple large errors. Should have easily hit 18.
Third attempt was FRAPSed. Analyzed to find minimum delay between casts. When performing the test, I found I could start casting the next spell at the moment just at/before the point where the character starts shifting from the 'ready' stance back to the 'idle' stance. ('ready': spell finished casting, arms held out to the sides away from the body, body angled relative to the target; 'idle': standing, arms straight down, facing target straight on; this is on a Taru character)
If I started from one of the 'mistake' casts, where I was fully idle before starting the next song, the point where the character starts returning to idle stance after the cast (and thus when I could start the next spell following) is right at 3 seconds (up to maybe 3.13 seconds) after the initial cast.
Done perfectly, I could maybe see it hitting 19. Not sure about 20, but I would definitely consider it the limit. Inter-spell delay definitely seems to be 3 seconds.
Cool, and now that I think about it I don't believe what I proposed earlier is possible. I know I've done JA -> Spell a lot quicker than 3 seconds apart. If anything, I feel like there might not be a limit there at all. In the past I know I've had JA usage show up in my chat log after my spell casting starts (with the JA).
http://www.bluegartr.com/threads/104...=1#post5043156
For the record, likely be additive. I cannot understand why SE would make elemental cast reduction additive and not songs. Then again, SE does weird stuff.
Yeah, JA->Spell can be done near-instantly. A quick bit of testing indicates that the 1 second hard cap on JAs applies, but that's fast enough to almost never be an issue.
Spell->JA has a notable delay, though. Did another Nightingale test for this. Used a trigger to have Spellcast execute the command sequence:
input /ma "Valor Minuet V" motenten; wait 1.0; input /ja pianissimo motenten
Then tested at waits of: 1.0, 2.0, 2.5 and 3.0. Pianissimo failed in all cases except for the 3.0. So that's another indication of the 3 second inter-spell delay, and that the delay also applies to JAs.
As a followup on the effectiveness of magic procs in Dynamis, ran a test on it. May not be the best thread for it, but meh..
Dynamis magic proc test.
Setup:
Thf/dnc - main DD
Rdm/whm - magic proc'ing (me) (so no benefit or penalty for Light/Dark Arts from /sch on any of the spells), and only used Composure for recasting buffs, cancelling it before trying to cast any enfeebles (so no penalty on casting time).
Thf/dnc - secondary DD who joined us for a short time. No effect on test numbers.
Most of the test was as a duo.
Area: Dynamis - Buburimu. Fought: dhalmels, ravens, mandys, rabbits, scorpions, ravens, mandys; in that order.
Spells for procs:
Gravity, Gravity II, Blind, Slow, Paralyze, Silence, Poison, Bind, Blind II, Slow II, Paralyze II, Addle, Break; in roughly that order of priority.
Also: Sleep, Sleep II, but need special considerations there since I also used them outside magic proc time.
Only used Dia III on proc'd mobs, and did not attempt to use Dia as a proc spell at all.
Also used nukes on a few mobs at the end. Of 3 mobs, 2 were proc'd immediately with Bliz 4 and Bliz 3; used Bliz 4 on the third, but it did not proc; ran out of time.
I proc'd 46 mobs. Excluding the 2 proc'd with nukes at the end, 44.
Total number of successful enfeebles: 453
Total number of successful sleeps: 118
Estimated sleeps outside of magic proc times: 10-15. Will use 107 as number for sleeps.
Total proc spells cast: 560 (matches total result if I exclude mobs fought after time rollovers)
Proc rate: 44/560 = 7.9%
95% confidence interval: 5.89% - 10.4%. Best overall estimate: 8.01% (Adjusted Wald calculations).
Tentative conclusion: magic enfeebles have an 8% proc rate.
Other stuff:
Total farming time: 1 hr, 45 minutes
Overall average enfeeble cast rate: 5.33 per minute.
Spent most of my time trying to cast enfeebles, with occasional breaks to rebuff. Light rate of mis-casts due to inter-spell delays.
Estimated non-proc casts: 275
Overall average casts per minute: 7.95
Total number of mobs killed was about 75% as many as when we go mnk+thf. Since his thf is a good bit weaker than mine, might have come close to matching the number of mobs killed had he been rdm while I went thf. Main limiter for this setup is still the proc rate, though.
Re: Nukes -- 2 out of 3 procs give a confidence interval on nuke proc rates of between 20% and 94%. Obviously not practical as a real estimate, however the lower end is well above the upper limit of the confidence interval for enfeebles. Would be worth doing another test solely with nukes.
Avatar BPs also proc magic weaknesses and the proc rate felt substantially higher than 8%. Anyway, though, good to know!
I did some Aspir testing and think the formula for 300+ is likely (Dark Magic Skill)*0.4
That makes Aspir II (Dark Magic Skill)*0.6, although I suspect it's actually Aspir potency (the formula above) floored and *1.5.
Next up is Drain, which is somewhat annoying. I would like to give it a prettier formula, but the prettier formulas don't fit as well so I'm just going to use this for now.
You probably shouldn't use the 303 skill data in construction of the formula:
Drain (303 skill) | 60 | 170 | 321 | 247.5
That data represents a significant departure from the min = 0.5 max formula - there's no way to tell from that whether the max is something higher (up to 340?), min is lower (down to 160?) or some combination of a higher max/lower min. The other possibility is that some range of dark magic has a different function for determining minimum potency, but that's highly unlikely. Granted something in the range of 321 makes sense as it's 320 at 300 skill, but I'm just suggesting that it's probably off slightly (wouldn't be surprised if it was actually 1 or 2 higher).