Asura.Chiaia said: ยป
Old Min: 108 (480 + 60) * .2
New Min: 123.6 (480 + 138) * .2
This is the part I'm questioning because I have yet to see one shred of information that actually proves floor is determined after +delay, it just seems to be assumed.
Based on your first response to my response to Cherrywine I got that you were questioning this. I would recommend going to test it yourself and let us know your findings! The wiki always tries to link to sources if we can but some of that has been lost to time or no one helping out and linking stuff when new info comes out.
Your original sentence about just adding 41 more MA to go down to 96. Just acted like this was defiantly a thing. Not some new unproven idea. I just hate some newbie comes along sees this and doesn't get the full context and then we have everyone thinking this. Probably won't happen because were talking "lol mnk" here so no one is bandwagoning it but still
The wiki always tries to link to sources if we can but some of that has been lost to time or no one helping out and linking stuff when new info comes out.
I can't even find the "old info" related to this. Just one day someone edited a wiki with an entry and suddenly everyone just agreed that it must be true because the wiki says so.
It was the exact same for Sinker Drill, everyone just assumed it was DEX mod cause the wiki says so, and the guy who put it there said the JP wiki said it was DEX and so forth. Turns out it was STR / VIT. The way Hybrid WS's work was another, a popular poster who frequently posts correct info said "it must be this way because it's always been understood to be this way". Turns out it's always been wrong and they work in a fundamentally different way then previously assumed. And if something was "so old as to be lost in time", that means any testing was sketchy at best and prior to the combat system being modified several times.
As for H2H floor determination, until there is something more concrete then "some guy told me so" or "well I heard it was this way a decade ago" then it's currently unknown. All current forms of delay modification happen after the floor is calculated, H2H being a special thing opens the possibility of it being different but doesn't affirm that possibility.
Sorry if this sounds pedantic but I absolutely hate ***being assumed to be true just because some guy some where said it was on an internet forum.
For Martial Arts, which is not a percentage-based decrease, you must use your base delay (480 for H2H) plus weapon delay in the calculations instead of calculating it independent of delay as shown above.
Looked through all references, and they were just to typical fractional haste values and utsusemi recast values. The line just appeared in the wiki without the usual linkage that goes at the bottom. Byrth is normally very solid on info, and he almost always puts a link to the page where someone demonstrated something, unless it's something to blatantly obvious that no link is needed.
i wouldn't really put that much stock in delay tests conducted with frames(essentially every test prior to 2013..), especially now that we know how wide packet intervals are
really need #/rounds over period of time, maybe fight one of those blunt immune mobs in romaeve until it depops and measure total rounds
i wouldn't really put that much stock in delay tests conducted with frames(essentially every test prior to 2013..), especially now that we know how wide packet intervals are
Frame testing is acceptable but only in VERY large sample sizes, like 10K+ continuous attacks. Any lag would be ironed out in the average.
AFAIK the only frame testing that was done was on HF and no attempt at determining delay floor was used. It's how they know HF is a separate 75% increase in attack speed.
This type of test would need two separate runs, first being with capped magic / gear haste with no MA gear. Then add ~40 MA gear and do it again. Both assuming a ~80 delay weapon. Compare the results and see if adding the MA made a statistically significant difference.
Plugin logged an attack count and timestamped each round, recording from incoming 0x28 packets and discarding those with identical sync bytes. I engaged with plugin loaded, looked at total number of rounds -1 and time period from first round to last.
First attack round 7:10:06.384
Last attack round 7:13:06.116
Total attack rounds: 87
179732ms / 86 rounds = 2089.9 ms/round
Same test, replacing earrings with 2x Mache Earring+1.
First attack round 7:18:43.056
Last Attack round 7:21:42.463
Total Attack rounds: 90
179407ms / 89 rounds = 2015.8 ms/round
Same test, keeping the Mache Earrings+1 and adding haste samba from DNC main with 5/5 merits. DNC main engaged first to apply samba before test began, reducing sample size slightly. Samba never dropped.
First attack round 7:28:23.044
Last attack round 7:31:18.211
Total rounds: 88
175167ms / 87 = 2013.4 ms/round
Did the test I suggested in Ro'maeve, can decide what you want from results. Full methodology in spoiler.
Spharai, capped gear haste, capped magic haste, job master, no martial arts gear: 2089.9ms per round
Spharai, capped gear haste, capped magic haste, job master, 26 martial arts from gear: 2015.8ms per round
Spharai, capped gear haste, capped magic haste, dnc main samba w/ merits, job master, 26 martial arts from gear: 2013.4ms per round
Spoiler is broken for some reason, would be good to get a sample size for posterity, and is there a weapon you can use with a lower +delay then Spharai, makes seeing any caps much easier. We want to try to reach near 308 delay. As it stands those results are too close to each other at 3.6% difference.
Remember 96 is the lowest no matter what, so want to get as close to that as possible via MA + Gear + Magic Haste, again just to clearly see the separation so it can be a wiki link.
Matamata stuns with it's plain attacks, pretty sure. Also has a slow TP move. When silenced, the urns do nothing. Couldn't think of much else that would survive for an extended amount of time without being logistically troublesome or moving. Obviously, if a better idea comes along I have no problem loading up the plugin again and punching something for a longer interval.
Didn't really select gear for anything other than it being in my inventory/wardrobes and not having martial arts. Think I was on BLU beforehand.
Quote:
Spoiler is broken for some reason, would be good to get a sample size for posterity, and is there a weapon you can use with a lower +delay then Spharai, makes seeing any caps much easier. We want to try to reach near 308 delay. As it stands those results are too close to each other at 3.6% difference.
Remember 96 is the lowest no matter what, so want to get as close to that as possible via MA + Gear + Magic Haste, again just to clearly see the separation so it can be a wiki link.
Removed spoiler tags. If you can think of a mob that can be reasonably held still, won't use debuffs that could interrupt sample, and will last 10-15 minutes I'm more than happy to repeat the test with whatever easily accessible weapons you want tomorrow evening.
I'll agree that my sample is insufficient for claiming an exact formula, but it sufficiently rules out 480+weapon delay * .2 as a cap. The numbers aren't even close.
I'll agree that my sample is insufficient for claiming an exact formula, but it sufficiently rules out 480+weapon delay * .2 as a cap. The numbers aren't even close.
It's more sample size vs variance. You can have a smaller sample size if the variance is really large, but with only 3.6% measured and the predicted also being close, anything small could cause a misreading.
Isn't there a monster somewhere that's immune to all damage except magic? I think in Zitah or Ro'Mave can't remember.
Not really about an approved test as something that can be linked in wiki and have a concrete answer instead of "someone told someone who told me and I believe them". I have cherry blocked so rarely see anything of his outside of quotes or an occasional manual check.
Well, to add the 40 MA that Saevel wants to see tested would take:
Mache Earring +1: 13
Mache Earring +1: 13
Two pieces of Rao +1: 8
Bhikku Cyclas +1: 6
In another post he wanted 41 MA added... So:
Mache Earring +1: 13
Mache Earring +1: 13
Shaolin Belt: 10
Count's Cuffs: 5
I'll agree that my sample is insufficient for claiming an exact formula, but it sufficiently rules out 480+weapon delay * .2 as a cap. The numbers aren't even close.
Isn't 2013 minimum obtained kind of close to calculated 1987? Certainly a lot closer than first cases calculated and actual
It's more sample size vs variance. You can have a smaller sample size if the variance is really large, but with only 3.6% measured and the predicted also being close, anything small could cause a misreading.
Isn't there a monster somewhere that's immune to all damage except magic? I think in Zitah or Ro'Mave can't remember.
Those mobs depop 3 minutes after engage, they are what I used for initial sample and why it is so small. Given the data was logged by a program, the only margin of error should be packet jitter(which is no more than ~165ms total distributed across 90 attacks). We aren't measuring a random sample, we're looking at data that's updated in 330ms intervals.
Just wondering what you would consider sufficient. Metalloid Amoeba seems to be a good test mob and would cap at about 29 minutes of consecutive punching. Is that a good enough sample?
Assuming I run the 29 minute test twice, what weapon, which buffs, and how much martial arts would you like me to wear? I do not have access to a non-master MNK or any gear that requires quest lines.
(You're the skeptic here and it's not any additional work for me to do it under your preferred parameters, so I would like to know exactly what you'd like me to use when testing.)
I'll agree that my sample is insufficient for claiming an exact formula, but it sufficiently rules out 480+weapon delay * .2 as a cap. The numbers aren't even close.
Isn't 2013 minimum obtained kind of close to calculated 1987? Certainly a lot closer than first cases calculated and actual
I think you meant 1875*... But, regardless, from the current understanding of H2H delay, as posted on the wiki, Spharai would benefit from 4 Martial Arts in gear.
Quote:
Spharai
So the minimum delay we can experience with Spharai is:
(480 + 116) * .2 = 119.2
When Haste-capped and Mastered, Spharai delay is:
(480 + 116 - 210) * (1024 - 256 - 448) / 1024 = 120.625
You could add up to 4 Martial Arts gear and not exceed the delay cap when wearing Spharai.
**
So, anywhere from 1-4 MA in gear should lower the delay between attack rounds, assuming the current formulas are accurate.
Comeatmebro's results don't confirm that the formulas are 100% accurate. Since he added Haste Samba rather than more MA, I don't think we can definitely say the results crush Saevel's theory that a delay of 96 could be achieved while wearing a H2H weapon. But it does confirm a cap was reached somewhere.
I'm all for testing things, even things we have long assumed to be settled. I definitely don't agree with discarding a standard just because the foundation of said standard has been lost, so therefore, surely any untested opinion should carry just as much weight. I also look down on creating an echo chamber for oneself to ensure their opinions are questioned as rarely as possible.
**Edit #1, I guess it should be said Spharai could benefit from up to 5 in MA gear, but going from 4 to 5 will reach and exceed the cap and begin to cut into TP gain.
*Edit #2, I reread and see where you are getting the 1987. Derp. Given the range of error of 165ms, then, yeah, it all appears to fall very closely inline with the current understanding of H2H delay on the wiki.
Can use 2 pieces of rao, hiza head +1, and 2 earrings for a total of 44 martial arts. Just need to know what weapon will be suitable, since apparently Spharai has too much delay.
Seems to me it'd be much easier to see if delay is a factor when using a higher delay weapon.
Yea, I don't understand what saevel is trying to say honestly.
going to try lynx baghnaks, lowest delay i can get, if a cap is applied post-weapon delay you're looking at 510 * .2 = 102delay = 1700ms
if it's applied pre-weapon delay, you're looking at 480 * .2 = 96 delay = 1600ms
that alone makes the margin of error pretty small, while trying to measure for something like spharai is much bigger.. but i will do a larger test with lynx baghnakhs and 44 martial arts, capped magic haste since that seems to be what was requested
Pretty easy to see from the first log how it's not really random and it will quickly normalize, for those who couldn't understand the explanation. Since the server is calculating your attacks constantly, the time the packets may come in will vary, but you're guaranteed that they even out significantly faster than a true random distribution because they're representing a fixed speed. Did that one as long as it was to illustrate it, the later logs are shorter samples so that I'm not up all night working on this.
Second soap gone, so done for tonight. If any specific requests, will run them tomorrow. The numbers are consistant enough for me to be confident they're correct, but they don't seem to match our delay models.
Just wondering what you would consider sufficient. Metalloid Amoeba seems to be a good test mob and would cap at about 29 minutes of consecutive punching. Is that a good enough sample?
Like 200~300 attack rounds would be good enough.
I believed you guys were probably correct at the start of this page but probably correct isn't good enough. I was hoping someone could dig up some post made long ago and we could link that into the wiki for any future reference. The fact that nobody could point to something and just assumed it was true cause other people assumed it was really irks me.
Yea, I don't understand what saevel is trying to say honestly.
Trying to outline the delay floor, which is easier to get under with with lower delay weapons because MA is a static value instead of a percentage. Trying to do that with the least number of assumptions. 96 delay should be fine since you can do a run where your over, then do another run with MA where you should be well under and compare the differences looking for variance.
Given the data was logged by a program, the only margin of error should be packet jitter(which is no more than ~165ms total distributed across 90 attacks).
When the packet jitter is larger then the measured differences we need to average it out with sample size, sucks for the 3min depop time cause it would need to live for ~7 min to be sufficient. The important thing isn't so much time as number of attack rounds, since that's what is being tested.
Could you format that in a page or make a summary in an edit and when I get home tonight I'll link it into the wiki if someone else doesn't beat me to it?
Pretty easy to see from the first log how it's not really random and it will quickly normalize, for those who couldn't understand the explanation. Since the server is calculating your attacks constantly, the time the packets may come in will vary, but you're guaranteed that they even out significantly faster than a true random distribution because they're representing a fixed speed. Did that one as long as it was to illustrate it, the later logs are shorter samples so that I'm not up all night working on this.
Second soap gone, so done for tonight. If any specific requests, will run them tomorrow. The numbers are consistant enough for me to be confident they're correct, but they don't seem to match our delay models.
Only one way to respond to all this: And SE doesn't care about monk clearly. Hell they probably don't even play the job. Some Japanese player probably complained. They said well let's do something. They did something it was awful and we all moved on ! There is no way in hell SE is going to spend the time to do the testing you guys do. They just don't care that much.
I appreciate all the work and care you take with monk. If SE cared half as much as we did. Do you think they monk would be in it's current state ?
Think it may be worth reevaluating delay models in general, with the better tools we have now. Those numbers were accurate to within 1ms very quickly, but they don't actually match what we expect to see(1760ms instead of 1700 on the lynx baghnakhs for example).
Mathy question, I'm currently reviewing the MNK spreadsheet
BG-wiki lists VS as adding 10% crit at 1000 tp, 25% crit at 2000 tp and 45% crit at 3000 tps.
So let's say you have 1500 tp, you are adding 17.5% crit on top of other sources of crit.
However that is not the way the spreadsheet is counting it. The spreadsheet, for 1500 tp, is counting
10% +0.5*0.25=22.5% crit
Who is correct ? It's important for godhands, and if it's a mistake on the SS, godhands is overestimated.
Also, VS is listed ingame as a 4-hits WS. With subhand that's five hits (yes/no ?). However it seems the VS definition in the spreadsheet is only counting 2 extra hits (it's as if the spreadshet was already counting the subhands as part of the 4 hits). I hope this is clear. Is the spreadsheet wrong ?
