Cast and Recast

[Prothescar]

My results are not matching yours. My system predicts 15s for Protectra IV and 25s for Stoneskin, which is exactly what I'm getting with the gear that you're using. Idk how you're getting 24s...

Stoneskin: 1800 * 0.841796875 = 1515 = 25.25. result: 25
Protectra IV: 1080 * 0.841796875 = 909 = 15.15, result: 15

With Goliard Saio, Blessed Mitts, Blessed Feet, and Goading Belt, in-game results match this perfectly for me. Are you sure you aren't missing something somewhere?

[Motenten]

Ack, my fault -- my description wasn't complete. It's 162/1024 haste and 1% Fast Cast recast (Loquacious) for those results.

[Prothescar]

I'm still not sure how you're getting 24 for SS.

1800 * 0.99 = 1782. 1782 * 0.841796875 = 1500 = 25, result: 25
1080 * 0.99 = 1069. 1069 * 0.841796875 = 899 = 14.98, result: 14


The only way I can replicate your results is by doing fast cast second instead of first, which would be incorrect.

[Motenten]

Hm. That's actually the issue, that earlier testing indicated that Haste was calculated first. If that's not the case... I'll have to set up some more tests to verify that.

[Prothescar]

have gone through various tests used in this thread and have yet to find something that disproves it. honestly it all lines up so perfectly it's hard to imagine it being inaccurate; server runs at 60 cycles per second, we know this, and everything lines up perfectly with fast cast coming before haste when basing the model on that. having a second timer separate from the master one just for recasts would be redundant and overall fairly worthless.

[Motenten]

Tests for Haste first vs Fast Cast first, using the integer /60 calculations.

Protectra III
50/1024 haste, 10/1000 FC (W.Turban + Loquacious)
16 second recast = Haste first
15 second recast = FC first
Result: 15 seconds
Using 7-bit flooring, predicted results are 15 seconds regardless of order.


Curaga IV
51/1024 haste, 20/1000 FC (Goading, Loquacious, Incantor)
9 second recast = Haste first
10 second recast = FC first
Result: 10 seconds
Using 7-bit flooring, predicted results are 9 seconds regardless of order. (Wrong)


Shellra V
190/1024 haste, 30/1000 FC (Haste, Goliard Saio, Loq, Incantor, Prolix)
15 second recast = Haste first
14 second recast = FC first
Result: 14 seconds
Using 7-bit flooring, predicted results are 14 seconds regardless of order.


Protectra IV
70/1024 haste, 40/1000 FC (W.Turban, Heiros hands, Orvail legs, Swith cape)
Using 7-bit flooring:
16 second recast = Haste first
15 second recast = FC first
Result: 16 seconds
Int 60 prediction: 16 seconds in both cases



So Int 60 method results in FC first in all tests. The 7-bit flooring predicts haste first, but is wrong regardless of order in one of the tests.

More testing will be needed (including to confirm N=60 as the correct denominator), but for now it seems to be a very solid model.

[Byrthnoth]

Gendewitha Gages are somehow weird in terms of cast/recast. They might have a tiny amount of Haste on them that was left out of the description.

Weakened Reraise casts:
* Gages alone - 116 seconds
* Gages + Rostrums - 113 seconds - should be impossible
* 28% Haste - 105 seconds
* 28% Haste + Gages - 101 seconds
* 28% Haste + Gages + Rostrums - 99 seconds
* 24% Haste + Gages + Rostrums - 100 seconds

Unweakened Reraise casts:
* Gages alone - 58 seconds
* Gages + Rostrums - 56 seconds (should be 57s if they were 10% together)
* Orvail + Loquacious + Rostrums - 57 seconds (which is proper for 10% Fast Cast)

[Foldypaws]

Gendewitha Gages are somehow weird in terms of cast/recast. They might have a tiny amount of Haste on them that was left out of the description.

Weakened Reraise casts:
* 28% Haste + Gages - 101 seconds
* 24% Haste + Gages - 100 seconds

Wait, what? You *remove* haste and recast goes *down* ?

[Byrthnoth]

Sorry, the second case had Rostrums too. The goal for that test was 99 seconds and I only hit 100.

[CDF]

Not too familiar with recast so I would appreciate feedback on this:

Revisiting Ionis Fast Cast and Haste bonuses
http://www.bluegartr.com/threads/108...=1#post5808918

Motivation for this post because I've seen "2% gear haste" on FFXIAH three times without source and I am inclined to believe it's not true as there is no mention of moderate Haste increase anywhere from the Ionis NPC.

Considering the date of the previous post, wasn't Minor Fast Cast effect (Scouts' Emblem edification) active as well? If Fast Cast +1, Haste values 7/1024, ... 12/1024 are possible as well. However the Victory March recast doesn't permit any Fast Cast so I'm not sure.. See edit.

Note that Couriers' Bay Roof edification provides moderate Critical Hit rate increase (in-game description), not moderate Haste increase.

Today (all Ionis buffs in effect) casting Klimaform I get 176 second recast (using lua to print directly into chat) with Ionis active, which is consistent with Fast Cast +2, Haste +12/1024 . 176-second recast rules out Fast Cast+3 4 holding haste at 12/1024. It also rules out haste 20/1024 with non-zero fast cast Fast Cast +2.

Excel formula in case you were wondering whether I got the order right (multiply base recast by 60, apply fast cast, truncate, apply haste, truncate, divide by 60, truncate to get reported recast in-game). Below is example for Fast Cast +2, Haste +10/1024.

TRUNC(TRUNC(TRUNC(180*60*(1-0.02/2))*(1-10/1024))/60)

Edit: found out how odd values of FC are handled (JP wiki, this post: http://de.ffxiah.com/forum/topic/311...nes/6/#1939009) so no need to revisit Ionis observations from the previously cited post

Edit 2: as a check on whether I'm doing this right, I used Natirah Hat on SCH/WAR and WHM/WAR to get following recasts:

Klimaform: 160 seconds (using Anhur Robe [FC+10], Arbuda Grip, Hasty Pinion, Bokwus Gloves [Haste+5%] gives 162 seconds)
Reraise: 53 seconds
weakened Reraise: 110 seconds
weakened Reraise + Swith Cape: 109 seconds ********* need to check this*******
Reraise + Rostrum Pumps: 53 seconds
Reraise + Swith Cape + Rostrum Pumps: 51 seconds

12 FC and 50/1024 haste seem to match the above results except the one that I need to check. Not sure what's up. I definitely cannot replicate these results: http://www.bluegartr.com/threads/115...=1#post5726864 (haste was zeroed out?)

[Byrthnoth]

Testing Atrophy Chapeau with Reraise while weakened (and 25% gear Haste):
84 seconds with hat + 25% Haste (80% recast
81 seconds with hat + Swith + 25% Haste (78% recast)

I believe this indicates it is 11% Fast Cast.

[Motenten]

Reviewing CDF's post here


Examining the 7% fast cast issue.

Testing Spell: Klimaform
No haste, no arts, nothing else aside from fast cast being tested.

0.5% recast resolution (eg: 3.0% vs 3.5%, even though we consider the value to be floored)

Predicted recast for given recast percents:
0.0% - 180
0.5% - 179
1.0% - 178
1.5% - 177
2.0% - 176
2.5% - 175
3.0% - 174
3.5% - 173
4.0% - 172
4.5% - 171
5.0% - 171
5.5% - 170
6.0% - 169
6.5% - 168
7.0% - 167

To distinguish between 4.5% and 5.0%, use Protect II (5 seconds at 4.5%, 4 seconds at 5.0%)


Loquacious: 178 (1.0%)
Prolix: 178 (1.0%)
Swith: 178 (1.0%)
Orvail: 176 (2.0%)

Loquacious + Prolix: 176 (2.0%)
Loquacious + Swith: 176 (2.0%)
Loquacious + Orvail: 174 (3.0%)

Prolix + Swith: 176 (2.0%)
Prolix + Orvail: 174 (3.0%)

Swith + Orvail: 173 (3.5%) [Expected 4%]

Loquacious + Prolix + Swith: 174 (3.0%)
Loquacious + Prolix + Orvail: 173 (3.5%) [Expected 4%]
Loquacious + Swith + Orvail: 171 (4.5%-5.0%) [4 second Protect II == 5.0%]
Prolix + Swith + Orvail: 171 (4.5%-5.0%) [4 second Protect II == 5.0%]

Loquacious + Prolix + Swith + Orvail: 169 (6.0%)

Incantor: 178 (1.0%)
Incantor + Prolix + Orvail: 173 (3.5%) [Expected 4%]
Loquacious + Prolix + Swith + Orvail + Incantor: 167 (7.0%)

Loquacious + Prolix + Swith + Incantor: 172 (4.0%)

All anomalous results involved the Orvail pants. For the 5% recast results, Orvail must by 5% fast cast (to add with 3% Swith and 2% ring or earring). Swith+Orvail, however, doesn't give a 4% recast result, as one would expect from the 8% total fast cast; instead it gives a 3.5% result.

And Orvail plus two 2% items (Loq+Prolix, Incantor+Prolix, Incantor+Loq) all give 3.5% recast results, where the fast cast total of 5+2+2 = 9% should give a 4% result if floored.

On the other hand, the other 9% total (Loquacious + Prolix + Swith + Incantor) -does- give a 4.0% recast result.

And when all items are used together, for 5+3+2+2+2 = 14% fast cast, it gives the expected 7% recast result.

Confirmed 5% FC on Orvail pants by checking Klimaform as /rdm. On its own, can confirm it as 9.5% or 10%; adding Loquacious with Regen V confirms 11%, not 10.5%, so total basic fast cast of Orvail + /rdm has to be 20%.


Honestly, I have no idea what's going on. It seems any time the Orvail pants are used, if the result 'should' be a 4% recast, it instead uses a 3.5% recast. Beyond that it appears to behave normally.

[noodles355]

Hidden haste+# or something messing it up?

[Motenten]

Only if the haste is a negative value, since the results are lower than predicted rather than higher.

[noodles355]

Hidden slow+# or something messing it up?

[Prothescar]

Slow +0.2% is the preliminary conclusion that I've made so far based on Motenten's data and what tests I've done in the last hour or so. Whether it's a hidden stat or an error remains to be seen, but values line up perfectly with +0.2% Slow and no other piece that I've tested has the issue of errant decimal Fast Cast values.

So for now I'd say that orvail pants have a hidden -2/1024 haste, it's a perfect solution and really I can't find anything else that it could be other than a convention-breaking new type of variable that exists on no other piece of gear.

[Fwahm]

Wouldn't that be the same result as if the pants gave 4/1024 less fast cast than 5%, with the other pieces possibly giving ever so slightly more than what they are thought to give?

[Prothescar]

No, Fast Cast is not calculated in fractions of /1024. Also, since they're separate terms that are multiplied separately, it'd be noticeable if it wasn't haste/slow.

[dasva]

Wouldn't a possible hidden slow be easy to test?

[Prothescar]

i was planning on proving it concretely, but not tonight. it involves more than a single test. for now it's the only reasonable answer, so I'll be expanding on it tomorrow