Actual Relic information thread?

**Vodou** · 2010-02-25 18:08

Originally Posted by Kirschy

~20% is the floor for melee acc. I've done various blind pot testing in general. In particular my first Catastrophe testing was done on Steelsheels. I had 25% melee accuracy after ~750 hits, increasing my accuracy by ~+30 I still had 25% melee acc (showing that it was a lower bounded cap.) If you can find a scenario where your melee acc is below 20% for an extended period of time (and parsed), it would be fairly big news.

This also applies to mob acc in most cases. You can't evade over 70-80%~

I was at 14.1% (37/262) on ranged attacks at 10'. I wonder if either ranged attacks have a lower bound or if the accuracy adjustment is a separate calculation that is additive to the floor.

Any thoughts?

**Shinryuu** · 2010-02-25 18:36

If I were to wager a completely baseless guess it's a seperate -% from distance adjustment. Easily tested though.

Masamune · 2010-02-26 08:44

Originally Posted by Vodou

I was at 14.1% (37/262) on ranged attacks at 10'. Any thoughts?

Confidence interval calculation:

Error @ 95% confidence = 1.96*sqrt(14.1%*(1-14.1%))/sqrt(262), assuming 262 is the total number of shots included crits and misses.

Error = +/- 4.2% => Confidence interval: 9.9% < 14.1% < 18.3%

Indeed that makes 20% out of this C.I., but, from personal experience on sample sizes needed for parsing, 500hits is really a strict minimum to only start to have a "good point".
That result isnot "lost" though: if you still have the parse, you can just continue it, i wouldnot be surprised to see your accuracy raised @ 17% +/- 3.3% after 500hits (which includes 20% in the C.I.).

Btw, i'm interested how did you kept mob @ 10' while shooting at it ? Shadowbind has a too long recast to make such test "bearable", while bind spell well... i don't know honestly...
Asking that to know how i could reproduce for ranged testings.

**Shinryuu** · 2010-02-26 09:18

Worms, silly.

Masamune · 2010-02-26 12:10

Originally Posted by Shinryuu

Worms, silly.

o yes my bad, forgot those worms XD silly me ^^

**Vodou** · 2010-02-26 16:14

Originally Posted by Masamune

Confidence interval calculation:

Error @ 95% confidence = 1.96*sqrt(14.1%*(1-14.1%))/sqrt(262), assuming 262 is the total number of shots included crits and misses.

Error = +/- 4.2% => Confidence interval: 9.9% < 14.1% < 18.3%

Indeed that makes 20% out of this C.I., but, from personal experience on sample sizes needed for parsing, 500hits is really a strict minimum to only start to have a "good point".
That result isnot "lost" though: if you still have the parse, you can just continue it, i wouldnot be surprised to see your accuracy raised @ 17% +/- 3.3% after 500hits (which includes 20% in the C.I.).

Btw, i'm interested how did you kept mob @ 10' while shooting at it ? Shadowbind has a too long recast to make such test "bearable", while bind spell well... i don't know honestly...
Asking that to know how i could reproduce for ranged testings.

Late, but yes, I can confirm it was worms. I suppose forts would have worked as well, but not sure.

I'm not clear on why you would not be surprised at a result that "grew" to 17%. Regardless of what the next test batch would show, this would have no impact on the prior batch. The independent sample has to stand on it's own as an extreme outlier if we assume the underlying rate is 20%.

Ignoring for the moment the possibility of researcher error, and allowing for the assumption that the hit rate does not follow a normal distribution (being truncated), there is seemingly still no reasonable way to anticipate a sample rate of 14.1% (262 attempts) out of a normal population.

While its certainly highly likely that the rate would be higher if added sampling occurred IF we assume the underlying population is at least 20%, it does not however explain the result of 14.1% unless we are disagreeing with the very nature of confidence intervals.

I don't regard sample size to be an impediment unless you are questioning the same stats which you posted (Error = +/- 4.2% => Confidence interval: 9.9% < 14.1% < 18.3%) which are derived from accepted statistical methodology. (As an example, if we toss a 2-headed coin only 10 times and we get 10 tails, it's clearly not random, despite doing "only" 10 tests.)

The most likely scenario is that either the rate is floored at 20% but distance racc is an added adjustment, or that racc is not floored at 20%.

Having said all that (long version), I'll try to do some more testing and see if that sheds some light. I leave open a wide door to researcher error and I did not realize my results were surprising.

**FailureMidgard** · 2010-02-27 00:22

Originally Posted by Masamune

That parse is on which mob and lvl?

Also could you provide from the "Offense details" tab the normal+crits frequencies on each particular mob+lvl combination please ? (don't even bother posting numbers when filter is set on "ALL mobs").

Also, on "Buffs by time" tab, check if you got significant time% under a particular buff to be able to report corresponding numbers assuming that buff were up 100% of the time. Classic example: i've seen poeple kindly give me their parses which looked to have potential interesting infos, until i verify on that tab the buffs they had... Hasso+ 56,73% and Seigan+ 43.27%... needless to say half the corresponding hits becomes unexploitable (i should say each belongs to a different pDIF analysis)... other problem is the combination of that with another asumption with other buff up X% of the time like Berserk, which KParser can't do for now (doable technically but obscene in terms of display). In other words, please check if can make asumption buffs were up 100% of the time (easiest being of course 0 buff at all XD).

So can maybe have a bit more rigorous analysis on Relic procs, pDIF "boost" and Kaiten.

Btw no gear swaps were a great idea /clap, too bad you don't remember how much str & attack you used ;;

As stated, was vs Kuftal Crabs, though I didn't pick and choose 1 level of mobs. My goal was not to figure out the exact # of melee accuracy added on Kaiten, but rather confirm that Kaiten gets a significant enough accuracy boost that I could ignore said stat on my WS set still (basically, I looked around, found no proof, and wanted proof before I went with the accepted Byrnie+1/etc).

Also, I'm visiting my parents at the moment, so I don't have access to my parse until Sunday night, I can provide more information then. I don't recall if I used Hasso or Seigan, but I am 99% sure I kept it consistent throughout the test.

As for gear, I want to say it was;
Amano-x-x-x
Turban-Orochi-Novia-Musical
Arhats-Dusk-Rajas-Jelly
Boxers-VBelt-Byakkos-Usu

I could be off, but I'm pretty sure that was it.

Originally Posted by Zirk

Can you also include info from the "Extra Attacks" tab, I'm going to assume that with such a low melee accuracy %, many of the melee hits that actually landed were zanshin hits, which benefited from zanshins accuracy bonus, so your actual melee accuracy % would have been lower than what that parse actually states.

This is just for purposes of determining how much of an accuracy boost kaiten/maybe all single hit weapon skills actually gets, I initially sorta arbitrarily thought it was +50 accuracy, but being able to overcome a blind pot that well makes me think it could be anywhere like +200-300 accuracy.

Also to the above poster, I think you missed Failure's sarcasm, they are not claiming to have any kind of a hidden pDIF boost on Kaiten.

Again, can try to get you the extra info when I get home, I admittedly didn't look much at it, though I am sure I had 0% DA during the test, and as such had 0 1%TP Kaitens. And yes, the pdif comment was sarcasm.

Originally Posted by Vodou

I'm not sure how (or why) you eyeballed +200-300, but it was documented from tests at approximately +100 quite a while ago (and +50 has been debunked for a couple years at least.)

Failure's post is clearly supporting/enhancing the initial findings and in no way does it indicate anything over +100.

100 sounds like a good ballpark guess...but ya, I'm not trying to counter it at all, I just posted my data if anyone cared, and the only point I had in the test was to properly gear myself without any assumptions.

Originally Posted by Byrthnoth

Instinctively, I'd say:
If Blinding Potion is -256 Acc, 20% is the floor, and he parsed 56% melee accuracy, then he must have been way over cap before potion-ing up because +35 Acc would only raise you to ~37% Acc on Zanshin'ed hits or 27.5% Acc overall.

But if Blinding Potions weren't on all the time (only last 3 minutes) then who knows.

Either way, Kaiten (like any other first hit) has at least a +90 Acc bonus, and potentially more.

I had exactly 1 melee swing (Hit) that wasn't blinded. I blame Kirschy! I generally turned around at ~2min56sec, waited for it to wear off, reapplied, the continued to melee. I missed exactly 1 swing due to typing on aim.

Originally Posted by Kirschy

Failure and I both used crabs in kuftal and only tested for about ~30-45 minutes. Hasso/Blind pots were on full time. It was just a simple/basic test to see if the accuracy bonus was there or not.

This, but don't recall if I used Hasso. Will check on Sunday.

**FailureMidgard** · 2010-03-01 03:09

Code:

Failure
Buff                      Type      Min/Max    Average   Cr.Min/Max     Cr.Avg     Buff (In)Active%
+               Hasso    Melee       29/387     116.65      118/612     230.20             100.00 %
-               Hasso    Melee          0/0       0.00          0/0       0.00               0.00 %
+               Hasso   WSkill      189/784     577.14                                     100.00 %
-               Hasso   WSkill          0/0       0.00                                       0.00 %

Code:

Player               # Melee Attacks    # Melee Rounds    Attacks/Round    # Extra Attacks
Failure                          516               459                1                 57
Player               # Missed First Attacks    # DA w/Missed First   Possible Zanshin %
Failure                                 212                     57              26.89 %

Melee Hits:

Spoiler: show

Melee Crits:

Spoiler: show

By Level
C-42

Spoiler: show

C-45

Spoiler: show

C-47

Spoiler: show

C-49

Spoiler: show

Thread: Actual Relic information thread?

Thread Tools

Latest Threads

Up & Coming Threads

Hottest Threads