Anwig Salade WSACC test

[CDF]

It's an interesting result... strictly speaking, the statistical power associated with the comparison of the two proportions (first sample of size 1,000 with Anwig Salade, the other of size 400 with Walahra Turban) is just ok (under .70 for a one-sided test), assuming the difference in hit rate is really around 12.5% and there are no other unaccounted sources of variability.

Nothing really pops out as a possible confounding factor, though.

[Shinryuu]

Shouldn't the test be only about the 15 WS ACC (7.5% hitrate)? So O.Hat vs Salade or Turban vs Salade depending on what you picked as the second stat (can make a case for both WS damage +2% or ACC+10/ATK+5).

[CDF]

In this example the total accuracy +25 was compared against no accuracy in head slot. It would take larger sample sizes to examine only the WS accuracy +15 for an effect.

[Kyte]

To be honest I forgot about the accuracy mod on Asuran fists (although that in and of itself would be a moderately interesting test)

In that case, dual wielded Hexa-Strike (so 7 hits) w/o DA gear would be a good candidate. If you're manually checking TP return, Kclub might make it harder to see how many hits are landing.

Capped, you should hit all 7 times ~70% of the time. If the full +25 accuracy brought you just barely to cap, then you will know the extra +15 isn't working if you hit all 7 times closer to 40% of the time. This is a pretty big difference (30%) that should be easily detectable given an appropriate sample size, even with the extra accuracy on the first hit messing things up a bit (making it a ~43% chance of hitting all 7 I think? I dunno I'm too tired for teh maths atm)

[Kaope]

What does all this mean? Due to the nature of the ws and sample size, the test is inconclusive? (aka lul-gorget-test).

PS: Still waiting for someone to translate them. Where are you~ Mr.Taru-with pup & jukebox icon???

[Shamaya]

Just throwin out more crazy ideas,

Accuracy bonuses should be easier to spot at lower acc percentages. Floored acc of 20% and floored acc + wsAcc+15 (27.5%) would be a large difference fairly easy to spot. You'd be getting almost 1.5* the number of its just by adding the Kupo head.

There's the alleged acc bonus on the 1st hit of the WS, so using SA would take that hit out of the equation. But then, requiring SA also takes a lot out of the equation. If we tested Thf/war using vorpal in -accuracy gear (behemoth rings, etc) on high level/evasion mobs (lurker, gnat?), this would be one way to do it. But thieves can triple attack. Nonetheless you can mathematically predict your avg hits per WS at a given accuracy. With capped triple merits thieves get 20% more hits/time on WS. /war means singlewield. So over time vorpal would be 4.8 hits avg for thf/war (4 hits + 0.1*2*4 -- 10% chance of each hit to add 2 more attacks). Only 3.8 of these hits matter. At 20% acc you're getting 0.76 hits/ws (after first hit), and at 27.5%, 1.045 hits/ws.

Eh this is prolly too complicated -.-. Ideal situation going with a low-accuracy test would be a 2-hit+ WS, on a job with low skill level in the weapon for the WS being used, with the test including SA, and also not including double/triple attack. -.-

[Argettio]

Eh this is prolly too complicated -.-. Ideal situation going with a low-accuracy test would be a 2-hit+ WS, on a job with low skill level in the weapon for the WS being used, with the test including SA, and also not including double/triple attack. -.-

DRK/THF and Sturmwind?

[Taru Taru Stool]

That won't work because Sturmwind is automatically 100% accurate with sneak attack.

But! Maybe NIN/THF with Enpi?

[Argettio]

That won't work because Sturmwind is automatically 100% accurate with sneak attack.

I saw that on wiki, but assumed it was a lolwiki comment.

Enpi seems like a viable choice.

[Max™]

Can't get the people I was doing MKD last fight with, but my GKT skill is like 120, so if it hasn't been tried since then I'll go fuck around on Nin/Thf whenever I get my hat.

[Balrock246]

Eh this is prolly too complicated -.-. Ideal situation going with a low-accuracy test would be a 2-hit+ WS, on a job with low skill level in the weapon for the WS being used, with the test including SA, and also not including double/triple attack. -.-

Pup/thf and howling fist?

P.S. not 100% sure they get howling fist but it's worth a suggestion.

[Ragnus]

Blu/Thf Fast Blade seems good also

[dantaro]

Pup/thf and howling fist?

P.S. not 100% sure they get howling fist but it's worth a suggestion.

We do, so that'd work

[Kirschy]

Wish I had the Helm so I could test myself... but I'm interested in seeing some more test results.

I'd suggest the easiest testing method is DRK or BST /w Baha Scythe using blind pots on Kuftal crabs. Determining if it's exactly WS ACC+15 would be tough, but proving if it's applied to all hits or just the first hit should be fairly easy.

I guess on a related note, anyone ever mess with Hachiman set on SAM? Would be interesting to see if it works the same way~ (Even if it doesn't serve much practical purpose!)

[requim]

Isn't the goal of the test to see if ws accuracy applies to more than just the first hit? That's essentially all or nothing testing.

Because we don't care about the first hit, /thf would just be convenient to drop any issues due to 95% first hit accuracy. Next would just be to find a 2 hit weapon and a mob where you ideally have at most 87.5% accuracy. If you find that your SA WS consistently gives higher than 87.5% accuracy on the second hit you have an answer.

Another easy combo (though no SA) would be bard/warrior using a rune chopper. You'ld definitely have low accuracy! You're really only looking for TP return to be varied by 1 anyways, so SA isn't even needed.

I get that multi-hit weapons magnify the effects of accuracy, but it seems like trying to count that much TP gain would be a lot harder than 1/2 vs 2/2 hits.

[Kyte]

I get that multi-hit weapons magnify the effects of accuracy, but it seems like trying to count that much TP gain would be a lot harder than 1/2 vs 2/2 hits.

Ya, that was why I suggested WS with more hits; it seems like you could get away with a smaller sample size to support the hypothesis. You could count how many hits you got in with every individual WS, but really all you'd have to do is count whether they gave full TP return or not (so it's still a binary result).

[Shinryuu]

WHM/THF, SA Hexa. Fin~

[Niiro]

Just need to find a WHM with WS ACC on their helm that cares enough to test this......

Much easier to just do it with rampage, evis, vorpal, guillo, drakesbane... only 1-2 less hits and many more people will have the WS ACC on melee jobs.

Need a multi-hit WS (the more the better) with no ACC mod. Can use as low as a 2-hit (fast blade) if you want but you'll get results faster with multi-hit as long as you're careful so you don't fudge your own results (no STP, no DA/TA).

WHM/THF with hexa would be perfect tho, 6 hits, no ACC mod, can sleep mobs for SA.

RDM/THF and THF would be next best with evis, 5 hits and sleep(bolts).
(RDM has the same problem WHM does, who gets WS ACC on a mage helm?)

BLU and DRK work too, 4 hits and sleep spells (gogo baha scythe).

I'm certainly not re-getting my helm for RDM and my THF isn't 75 yet, so someone go test this for us lazy people.

[RyaWHM]

Okay, here we go... I did some testing, but didn't feel like blowing 4 hours
and 120k on blind potions or whatever. At least I wanted to contribute
some, rather than nothing -- I really hope this helps!

The only difference in ACC during WS, was the Selenian Cap WSacc bonus.

This assumes Blinding Potion is exactly -256acc!

Note: I started by using Blinding Potions with only ACC+3 from gear (swift),
but my total resulting acc was 49, and I was very concerned I was
below minimum, where +15 WSacc might have no effect; 64 acc seems
awfully low still!

EDIT: Almost forgot to mention (just for fun) -- although my overall
ACC, for non-1st-hit, and regular melee, was something like 40%,
my first hit of the WS never missed across 67 ws's. :P

==========

http://i685.photobucket.com/albums/v...whm/kuftal.gif

WHM75/SMN37
279 skill, 56 dex, 53 gear
should be 352acc
so, with Bpot:
~96acc

==========

Lv.60 Robber Crab (56exp)

Spoiler: show

10xWS (50 non-first-hit samples)
146 hits
177 misses
=323 swings
146/323 = 45.201%acc

WS TP returns:
8,9,11,10,10,12,10,10,9,11

Arranged:
8,9,9,10,10,10,10,11,11,12

Subsequent hit counts:
0/5 hit: 0 (sum=0)
1/5 hit: 1 (sum=1)
2/5 hit: 2 (sum=4)
3/5 hit: 4 (sum=12)
4/5 hit: 2 (sum=8)
5/5 hit: 1 (sum=5)
total=10ws sum=30

Subsequent hit acc:
30/50 hits = 60%acc

Lv.61 Robber Crab (60exp)

Spoiler: show

9xWS (45 non-first-hit samples)
108 hits
167 misses
=275 swings
108/275 = 39.273%acc

WS TP returns:
9,10,7,9,10,9,8,8,9

Arranged:
7,8,8,9,9,9,9,10,10

Subsequent hit counts:
0/5 hit: 1 (sum=0)
1/5 hit: 2 (sum=2)
2/5 hit: 4 (sum=8)
3/5 hit: 2 (sum=6)
4/5 hit: 0 (sum=0)
5/5 hit: 0 (sum=0)
total=9ws sum=16

Subsequent hit acc:
16/45 hits = 35.556%acc

(2x) Lv.62 Robber Crabs (63exp)

Spoiler: show

22xWS (110 non-first-hit samples)
244 hits
468 misses
=712 swings
244/712 = 34.270%acc

WS TP returns:
8,9,8,9,8,9,8,8,9,10,9,9,9,8,9,10,8,10,8,11,9,8

Arranged:
8,8,8,8,8,8,8,8,8,9,9,9,9,9,9,9,9,9,10,10,10,11

Subsequent hit counts:
0/5 hit: 0 (sum=0)
1/5 hit: 9 (sum=9)
2/5 hit: 9 (sum=18)
3/5 hit: 3 (sum=9)
4/5 hit: 1 (sum=4)
5/5 hit: 0 (sum=0)
total=22ws sum=40

Subsequent hit acc:
40/110 hits = 36.364%acc

(2x) Lv.63 Robber Crabs (66exp)

Spoiler: show

26xWS (130 non-first-hit samples)
319 hits
665 misses
=984 swings
319/984 = 32.419%acc

WS TP returns:
8,8,9,10,10,9,11,7,9,9,11,7,12,9,10,9,8,9,10,10,8, 11,8,7,9,11

Arranged:
7,7,7,8,8,8,8,8,9,9,9,9,9,9,9,9,10,10,10,10,10,11, 11,11,11,12

Subsequent hit counts:
0/5 hit: 3 (sum=0)
1/5 hit: 5 (sum=5)
2/5 hit: 8 (sum=16)
3/5 hit: 5 (sum=15)
4/5 hit: 4 (sum=16)
5/5 hit: 1 (sum=5)
total=26ws sum=57

Subsequent hit acc:
57/130 hits = 43.846%acc

Overall (combined)
818/2294 melee hits
= 35.658% melee acc

143/335 subsequent (non-1st) WS hits
= 42.678% WS acc

==========

Conclusions
... too small sample size, is what it looks like to me. :(


EDIT: fixed typos/colors; added "(2x)" for clarification that there were 6 crabs in total

[Starr]

small sample size but if it turned out like that after more testing it would be giving the 15 acc to extra hits.