Why are there fractional recasts on FFXIWiki + BGWiki?

[Kirschy]

I've been looking up the base recast time of certian spells lately, and I've noticed both Wiki have fractional recasts on certian spells. I was wondering if there was any proof or evidence that this is true...? The recast timer can only display a value of 0 seconds to 255 seconds, this would make it seem as though the recast is an 8-bit unsigned number.

[Suiram]

Unless I'm overlooking something, the most recent POLUtils can only find the Name and Description data for the spells as they're currently used. However it does read the old, unused files (apparently when SE moves to a new file for ingame data they'll often just leave the old one in there as a relic) and those do list some spells as having fractional recasts. For instance, the entries for what I'm pretty sure are the Blizzard lines of spells (the names and descriptions aren't listed and I didn't feel like looking it up to verify but the usable levels and MP costs seem right, and they're listed as ice element) are 11.5, 19.75, 27.75, and 36 second recasts for tiers 1~4 respectively.

[Kirschy]

Hmm, thats strange. Should be fairly easy to test if the game registers the fractional value, or if it's truncated. I wonder why the DAT files would even bother with storing the recast value if it's never used clientside.

I guess as a followup question, is anyone familiar with the data that's sent clientside for recast values? I imagine the recast plugin uses this data. I'm probably better off asking over at the Windower forums, but if anyone happens to know I'd be interested.

[Wizerd]

A better explanation is probably the fact that the official guide, when the game came out, listed fractional recasts for a lot of spells. I doubt they were ever fractional.

[Suiram]

Yeah but like I posted they're in the spell .dats as being fractional too ;x

[Wizerd]

Fair enough. But now this has brought up memories of all the shitty things the guide advised... And wrong information it had.. . etc. Good times.

[spooky]

Fair enough. But now this has brought up memories of all the shitty things the guide advised... And wrong information it had.. . etc. Good times.

RDM has Flash!

[Clash Perez]

Fair enough. But now this has brought up memories of all the shitty things the guide advised... And wrong information it had.. . etc. Good times.

RDM has Flash!

Once upon a time it did, didn't it? Back in the days of RDM's Cure V, and no Convert/Refresh?

[LinktheDeme]

Fair enough. But now this has brought up memories of all the shitty things the guide advised... And wrong information it had.. . etc. Good times.

RDM has Flash!

Once upon a time it did, didn't it? Back in the days of RDM's Cure V, and no Convert/Refresh?

I think RDM always had convert. I know they didn't always have refresh though

[Isiolia]

No, they got Convert in a patch. Same with basically everything you'd invite a RDM for today.

Cure V was never actually usable, it was on the scroll post-70 for RDM before the cap was raised.



Far as recasts, I dunno, I was under the impression it got truncated, but I've seen people swear they've tested it and the first second counts down faster or whatnot. 'course, they didn't provide evidence to support it, just said they tested it, and that I should do the same (fraps the timer, etc).

[Amele]

it's entirely possible that they're coded as float but get truncated in the process of calculating the actual recast - this may be a balance effect to slightly reduce the effectiveness of recast reduction on some spells. (i.e. a 3.5 and 3 will have the same recast without any haste, but it will take more haste to get the 3.5 to be a 2 than the 3 to be a 2).

anyone eyeballing anything on a game that uses the level of client-server trust this one does is nuts; I'd only believe frame by frame fraps at this point.

[Kirschy]

it's entirely possible that they're coded as float but get truncated in the process of calculating the actual recast - this may be a balance effect to slightly reduce the effectiveness of recast reduction on some spells. (i.e. a 3.5 and 3 will have the same recast without any haste, but it will take more haste to get the 3.5 to be a 2 than the 3 to be a 2).

anyone eyeballing anything on a game that uses the level of client-server trust this one does is nuts; I'd only believe frame by frame fraps at this point.

While I do think the frame by frame counting method is very accurate, I don't think it's accurate enough to give us enough proof on fractional recasts. As shown with Delay and Casting, the game appears to have 'tiers', and thus a degree of inaccuracy in regards to activating animations/displaying serverside info.

I think knowing what kind of data is sent, when the client recieves recast data, will help a bit. As long as we can prove that the game truncates before calculating changes to recast we can model it successfully for almost every situation, which is what most of us are looking to do.

I think youre right on when you mention testing if a higher haste value is needed to bring a fractional recast down by 1 second. I designed 2 tests which will I'll run later today when I can find the time:

Burst - Base Recast: 43.25
Flood - Base Recast: 43.75

Victory March @ 499-505 skill = 113/1024
Nashira Turban(2%) = 20/1024
Total Haste: 133/1024

[(1024-113)/1024] * 43.25 = 37.632 = 37 Sec Recast
[(1024-113)/1024] * 43.75 = 38.068 = 38 Sec Recast

If the numbers used in the serverside calculation were fractional, there would be a change at certian values of haste. If the game truncates or uses 0-255 value before doing changes, there should be no haste value which gives two different recasts.

A follow up test would be:

Victory March @ 546-552 = 120/1024

[(1024-120)/1024] * 43.0 = 37.961 = 37 Sec Recast
[(1024-120)/1024] * 43.25 = 38.182 = 38 Sec Recast
[(1024-120)/1024] * 43.75 = 38.623 = 38 Sec Recast

Other tests can be designed, these were just the ones I came up with when thinking about it.

[Amele]

yeah, I agree generally that reading the numbers should be sufficient; the comment about frame by frame fraps was directed more at people who would claim 'the first second runs fast!'

that's a good looking test - anxious to see the results once we can find a blackmage and a bard willing to help (I'll start looking myself.)

[Kirschy]

I just ran some testing with the following results:

75BLM/37WHM:

No Haste~
Burst: 43 Second Recast
Flood: 43 Second Recast

502 Skill(113/1024) + Nashira Turban(20/1024) = 133/1024 Haste~
Burst: 37 Second Recast
Flood: 38 Second Recast

Two recasts with the same base recast at 0 Haste, have different recasts at 133/1024 haste. It follows that the recast value IS stored as a fractional value. Worth noting is that according to the math, Flood would have been a 38.068 recast. This should mean that it should only stay at 38 second recast for a less than 1/10th of a second, but it was clearly staying as a 38 second recast for the full second. This leads me to the following assumption:

The fractional recast value is stored serverside during all calculations, and then truncated/floored at the end when the information is sent to the client and displayed in the recast value.

(Note: I have no intentions of trying to figure out if a fractional difference between two spells allows one to be recasted again a fraction of a second faster. Between lag and the obvious slight inaccuracies in animation displays, it doesn't seem worth persuing. My main goal is understanding how the system for works for creating an accurate model.)

I also ran the second test I designed. With 550 skill and Victory March, Burst and Flood both had 38 second recasts. This confirms that both have fractional values. It also means my fastcast data is flawed =/.

I have POL Utilities, but I can't find spell recast data. Does anyone happen to have a full list of spell recasts, so I can see which ones have fractional recasts =x.

[Amele]

good to see that the test worked out - sad to hear fast cast needs looked into again

if I come across a list of recasts I'll pm it to you - otherwise I'd have to dig the dats probably (if they're still even displayed there?) It'd be funny if this is the one thing that brady guide got 100% correct.

[Suiram]

If you go to String Tables > Common there are the old files I referred to in an earlier post. To my knowledge POUtils doesn't read the most current file, and I'm not sure how old the are, what they're missing and such.

[Kirschy]

Thanks Suiram, exactly what I was looking for. Some of these values differ from the wiki values. Both wiki's have Water @ 8.0, while the DATs have 7.75. Both have Water IV @ 36.0, and the DATs have it at 32.0. I think I need to play around and see how accurate the old DAT files are.