New Monk Weapon skills

[MarkovChain]

Final results
447 skill, brew, SA+ shijin en Ep mobs in abyssea, no gorget, 4% TP return
the WSs range between [4478,5086] (24 ws)

So if DEX=100% => ftp is greater than 1.061

447 skill, brew, SA+ shijin en Ep mobs in abyssea, flame gorget, 4% TP return
the WSs range between [4907,5556] (20 ws)
confirms values above.

[Motenten]

Gorget results require at least 1.159 fTP. Subtracting 100/1024 from that means base fTP has to be at least 1.0614, which is right on top of the 1.061 of your first test.

Closest higher fraction of 1.0614 in units of /1024 would be 63/1024, but I'm inclined to stay with 64/1024 as a more even unit, and the probability that you didn't get -all- the way to the max value in just a couple dozen weaponskills. Thus we can consider 1.0625 fTP as stable, along with the 20/40/60/80/100% dex mod.


Next issue is what Gokku mentioned. A number of the new weaponskills are showing up with attack penalties (but of course not GK, which gets an attack boost). The great sword weaponskill is -40%. I haven't seen values confirmed on any of the others.

If SS has a built-in attack penalty, that lowers its usefulness considerably. It could be tested by getting just enough attack to be at 2.0 cRatio on a particular mob (eg: 654 att vs Greater Colibri) and making sure that the range of damage values produced matches the expected distribution, or if it seems generally lower. If it's lower, then you'd want further testing to refine just how much it was penalized. If you save TP to weaponskill in the last 5-10% you should only land 1-2 hits, which would reduce the amount of variability to account for. Also, don't use a gorget for this test, due to the last issue.


And the final issue to be checked on is the possibility of fTP being carried over across all hits. This is the case for some of the new weaponskills such as Stardiver (polearm), where it's 0.75 fTP on all hits rather than 0.75 on the first hit and 1.0 on the remaining hits.

Since SS's fTP is fairly close to 1.0, its primary interest would be in the value of a gorget or belt. To test this you'd probably want to find mobs with about 6k HP (I know dolls in Altepa-A have about that much HP since that's what's needed to one-shot them using Blizzard V on blm) and do the brewed SA test again, as that should allow one additional hit to land. You'd then need to narrow down the range of possible values of the second hit based on the tests you already did for the first hit.

[MarkovChain]

The "expected distribution is bullshit", just saying. Nobody knows the distributions between min and max. Sometimes it's around the average between max and min sometimes it's concentrated on one side. I've personally conducted interesting tests on chigoes, getting hit by 3 of them with various level of defense including counterstance and stacked defense.

For instance I've got data on chigoes hitting me while counterstance is up and having beserk (40 def only)

counterstance_40 := ["86" = 6, "87" = 10, "88" = 20, "89" = 22, "90" = 41, "91" = 34, "92" = 32, "93" = 38, "94" = 26, "95" = 30, "96" = 35, "97" = 35, "98" = 42, "99" = 21, "100" = 31, "101" = 39, "102" = 29, "103" = 28, "104" = 24, "105" = 28, "106" = 38, "107" = 36, "108" = 48, "109" = 28, "110" = 34, "111" = 30, "112" = 24, "113" = 32, "114" = 32, "115" = 47, "116" = 39, "117" = 25, "118" = 40, "119" = 35, "120" = 38, "121" = 34, "122" = 31, "123" = 30, "124" = 28, "125" = 29, "126" = 32, "127" = 28, "128" = 37, "129" = 29, "130" = 38, "131" = 25, "132" = 27, "133" = 33, "134" = 33, "135" = 42, "136" = 40, "137" = 36, "138" = 34, "139" = 46, "140" = 41, "141" = 38, "142" = 33, "143" = 35, "144" = 18, "145" = 20, "146" = 10, "147" = 16, "148" = 10, "149" = 5, "150" = 2];BarChart(counterstance_40);

which produces the following distribution


Then I've got plenty of intermediate data. For instance if I got 390 defense

def_390:=["15" = 963, "16" = 48, "17" = 52, "18" = 50, "19" = 23, "20" = 106, "21" = 41, "22" = 35, "23" = 57, "4" = 47, "25" = 38, "26" = 45, "27" = 47, "28" = 32, "29" = 43, "30" = 56, "31" = 48, "32" = 49, "33" = 47, "34" = 56, "35" = 45, "36" = 32, "37" = 25, "38" = 10];BarChart(def_390);

which produces this distribution

So really .. good luck figuring out what the real distribution is it's clearly no uniform even at low defense, then you have intermediatory distributions like above where a fraction of the hit seem to floor at max(pdif). I've ran plenty of test and in this situation the proportion of hit that are = max(pdif) is proportionnal to the invert of def and therefor cratio.

The higher my defense the higher the % of hit towards the max distribution. Once def is low enough it's getting close to something that looks evenly distributed but it only looks so because if you look at the CC graph above it's not uniform nor symetric. So idk how you reversed attack from a few WS samples, but to say anything precise you'll have to know what proportion is in which interval and there doesn't seem to be anything trivial to describe it.

The only thing that seem to come from the countertatnce data above is that the average value is (min+MAX)/2 and there are as much value above and below this value (2K sample size). It's safe to say it's only correct at capped pdif and always incorrect to be safe at low pdif.

[Motenten]

Umm... That looks like a fairly ordinary distribution chart.

For the low defense chart:
Min damage is 86, max damage before accounting for 5% spread is 143. The low-end values (~86-89) have their frequency lowered because a lot of each damage value is moved upwards by the 5% spread, and isn't compensated for by having values below that spread into them. Midrange stuff (90-143) is mostly even, and the high-end (144-150) is taking the highest values' 5% spread.

86 * 1.05 = 90, which is the first value that gets to accumulate the full spread of the values below it, and is also the first value that's mostly even with the rest of the midrange. 143 * 1.05 = 150, which is the highest observed value, where 143 is the last of the mostly even midrange group.

The high defense chart has clear indication of a common phenomenon of the damage being capped. I'll often see values around 15-19 when I guard against mobs that I have relatively high defense on anyway. Somewhere in that range is the minimum damage that SE will allow. I haven't really tried to figure out how it's calculated, though.

The rest of that range is fairly even. 38 is 1.05 * 37, where you get the last bit of fall-off from the damage spread.

Edit: Oh, and the spike at 20 is most likely due to another part of the pDif equation that's a bit complicated to explain. Simple version: random pDif is generated between A and C; valid values are between B and C; values between A and B get reset to a 1.0 (though in this case it looks like 0.5; I never worked out the specifics for cRatios below 1.0). And of course values between B and C that fall below some arbitrary limit (looks like 75% of the 0.5 reset point, so 0.375) are capped at that value.

[MarkovChain]

Anyway since you don't control much, the only and best way to test pdif on WS is a lot of WSs then average the value and divide by the overall ftpx0.95. It's a 5-hit ws anyway so 1 mob = 5 sample. If there is a 40% drop in attack it's fairly easy to check just by examining the average WS. Also for the ftp happening maybe on all hits it will be impossible to test it with a brew like I did earlier. I had to do 20ish WS to get close to cap so I'll need 400 for the first 2 hits to get close to capx2. Again this will be easier to check with full WSs, for instance spamming brews without sneak attack.

[MarkovChain]

PUP99,432skill,D+39
No multiattack
219 DEX / 578 att / 173 str (irrelevant : you cap fstr)

I did sand sweepers ; only registered WS that didn't kill them. I stopped when they started to repop decent challenge.

max crit was : 476 (irrelevant.. I think, but maybe since I don't cap max(crit) I think)
shijin=[2128,2206,1963,2124,1676,1878,1943,2033,1912,2073,1950]

global ftp should be :
0.95*1.0625*(floor(432*0.11)+3+12+39+floor(0.85*219))+4*0.95*(floor(432*0.11)+3+12+39+floor(0.85*219))=1380
mean(shijin)=1989
stddev(shijin)=146
CI : 1989 +/- 86 ( more or less irrelevent since that's only 11 ws so 55 hits total).

1989/1380=1.44 average pdif on the WS hits.

Here we go, /check command returns low eva at 486 att and low def&eva @ 493 (easy preys), make it 490 => they got 392 defense, my cratio is therefore 578/392=1.47

Using masa formulae on the notes from wiki's pdif page
1 handed middle value @ 1.47 cratio is (1.14*cratio - 0.72+1.0888*cratio + 0.31)/2=1.438 (far from cap therefore)

It matches the average pdif found on the WS I did so far so yeah, nothing happened, no def penalty.

[Motenten]

Also for the ftp happening maybe on all hits it will be impossible to test it with a brew like I did earlier. I had to do 20ish WS to get close to cap so I'll need 400 for the first 2 hits to get close to capx2. Again this will be easier to check with full WSs, for instance spamming brews without sneak attack.

To test the fTP carrying over:

SA+WS first hit with gorget maxes at about 5563.

Max damage on a non-crit 1.0 hit should be around 2261
Max damage on a non-crit 1.16 hit (fTP+gorget) should be around 2622

Max damage if fTP carries: 8185
Max damage if fTP does not carry: 7824

That's a 360 point gap in damage, which should be enough to make it fairly easy to check for. If fTP carries, you should see something over 7824 nearly half the time since 2261 would only be about a 2.05 pDif. If you needed 20 weaponskills to get in the 5550 range for the first hit, you could probably expect to see a hit over 7824 within about 50 attempts.

Even better would be to add an elemental belt to the mix (should be easy enough to brew Indrik for a light belt), and take off Sanguine Scythe to drop crit damage to 30% (to prevent killing the mobs on the first hit). That would change values to: 4901 max first crit hit, 2842 for max if fTP carries, and 2261 if fTP does not carry. Total of two hits if fTP does not carry would be 7162, while if it does it would be 7743, giving you a 581 point gap. 7162 would be about 1.9 pDif on the second hit if the first one was near max, and thus would show up well over half the time.

Plus, given that about half the first hit crits should be at 3.0 pDif or higher, it would be very common for the first hit to be at least 4668. 2842 on the second hit after a 3.0 first hit would total 7510, which is still 348 points above the max total if fTP does not carry. You should see a value within that range on the second hit about 40% of the time. Overall, you should expect to see a value above that 7162 limit about 20% of the time if fTP carries across hits. It should be apparent very quickly.

If you don't see a value within that range within 20 weaponskills, we can pretty much dismiss the notion of the fTP carrying over.

It matches the average pdif found on the WS I did so far so yeah, nothing happened, no def penalty.

OK, good to know. Thanks for that testing.

[Zhronne]

I'm feeling a bit lost now... minus the "a bit" part.
Thanks a lot for all your efforts and dedication guys, really hope you'll manage to reach a conclusion about how good/bad this WS is, because atm I really have no clue from the back & forth things going on D:

I have been very rash... just did 5-7 WS on normal mobs, on Kaggen, on Qilin (on both MNK and PUP) and judged the WS as an inferior option to Vsmite, but better than everything else for those who don't have Vsmite. Altough my gear wasn't top-notch (tbh I find it harder to gear for DEX on MNK compare to STR, I had better DEX ws gear on PUP lol)

[MarkovChain]

Sand sweeper, 571 att, 219 dex 173 str, 425 skill, gorget

shijin : [2061,2005,1965,2236,1622,2079,1964,1493,1550,1870,1947,1810,1977,2241,2339,1492]
Mean2 : 1915.68
stddev :263.8
CI : +/- 129

If I assume that the gorget bonus applies to all hits
ftp2=0.95*1.1625*(floor(425*0.11)+3+12+39+floor(0.85*219))+4*0.95*1.1*(floor(425*0.11)+3+12+39+floor(0.85*219))=1511.33

The previous test was done with 578 att 432 skill and no gorget. I extended it a bit and got
[2128,2206,1963,2124,1676,1878,1943,2033,1912,2073,1950,1853,2166,2240,1104,2200,2124,1289,1425,1968,1882,2015,2237,2395]
so Mean1=1949.33 and ftp1=0.95*1.0625*(floor(432*0.11)+3+12+39+floor(0.85*219))+4*0.95*(floor(432*0.11)+3+12+39+floor(0.85*219))=1380.29

the ratio Mean1/Mean2 (assuming the sample size is ok) should be
ftp1*578/(ftp2*571)=0.9244
Here we find mean1/mean2=1.0175

At best this fraction is, with our CIs,
(mean1-123)/(mean2+129)=0.893 and at worst (mean1+123)/(mean2-129)=1.160

so it's still borderline possible.. just need a bit more to infirm it. But I'm confident it's not the case.

[setting]

so whats better here 99 relic using new ws or a 99 vere with smite no clear anwsers are shown lol