Samurai Weaponskills: STR+ versus ATK+

[Kawfee]

The SA thing was a joke lol.

The pDIF isn't a bell curve? That is interesting :T Then you're right, the average will be to the left of the median.

When I go back to school tomorrow (left my graphing calculator at my apt at college lol) I'll plot it and see >.>

Something that the SA commentary person said rings sparked something in my mind:
If you can WS on Tiamat for 1000+, perhaps her VIT and DEF really aren't that high, just evasion and super high HP?

I've only fought Tiamat 4 or 5 times myself (and usually as RDM), so I'm really inexperienced in commenting on melee performance :/

[Kawfee]

Errr... anyone know if Excel can graph functions with 2 variables?

Edit: Only asking because on my new computer I don't have it installed yet and I left the discs at school :T

[archibaldcrane]

viper bite stuff

anyway, It's still doesn't show anything. your atk/def is already capped in the first place. If VB really does double your attack, you should show the numbers from fighting higher level enemy. (i.e. your corrected atk/def is about 1.0)

oh and this is an interesting thing you guys can test:
from the range 1.6~2.4, what's the most frequent number that pop in that range. I think it doesn't follow normal distribution graph (Gaussian). I've heard it's rather skewed to the left (lower number sides), so your average may not be sitting on 2.0, but rather (slightly) below it

It does show some things. It shows that Viper Bite has no effect beyond a regular melee hit when attack/STR are capped, which was the first thing I wanted to prove. Clearly you know that it does more than a regular melee hit when attack isn't capped, from leveling thief yourself, so we're pointing in the right direction. Patience VZX, I'll get the rest of the data

As far as the distribution 1.6-2.4, I am inclined to believe that it is a true random distribution. That is just my inclination, but in initial "looking at melee hits when I know where they can possibly land" makes me think they're pretty evenly distributed. Needs moar testing. I did notice hitting a ton of 129 crits, def. more than any other crit #.

[RKenshin]

Vs. Tiamat:

Assumptions:

• You are a SAM75/WAR37, because Sneak Attacking is stupid and Trick Attack has no bearing (unless Chiz chimes in).[/*:m:c3d1d]

Why? Get some good tanks. >_>

Think he was joking. No way people are going to waste their time doing 300-500 dmg WS when they can be doing 1-1.6k W

Actually on that topic, while yeah something like Tiamat is completely different and /THF is usually best, I wouldn't brush off /WAR completely for HNM fights (like most tend to do). For some of the HNMs outside of Wyrms (since they fly half the time) /WAR is actually better IMO if you're able to just spam TP at any opportunity, I ended up doing a lot more total damage between /WAR and /THF on something like Cerberus.

Last fight ended up looking like this as SAM/WAR, with one death in the beginning from me getting lava spit twice in a row without Shell. D:

http://img244.imageshack.us/img244/5345 ... rsekx5.jpg

[Rocl]

WHAT LOL
Is that thing saying
you did 48% of Cerebus' HP or am I reading it wrong!

[NynJa]

this thread definately needs some Chiz.

[Thitaubian]

WHAT LOL
Is that thing saying
you did 48% of Cerebus' HP or am I reading it wrong!

I think it would be 48% of the damage on Cerberus that he didnt filter?

[RKenshin]

In the example it's 48% (~25k) of the total damage delt to him (~51k) in a 35-40 minute fight (with a death at the start )... Only damage I filtered in that parse was damage taken by PT/ally members.

this thread definately needs some Chiz.

It does!

[Elmer the Pointy]

Hello, I'm little late but

If anyone cares about Viper Bite still, the Japanese description is

(( 2倍撃 )) and then blah blah add effect poison TP = poison duration

That beginning part means it's an attack multiplied by 2. This could be mistaken as 2 strikes or a "doubled attack", especially considering the animation. However, for a WS that actually strikes multiple times, it'll say 5回攻撃 or "5 times attack" in the case of Penta Thrust.

So the assumption that Viper Bite gives you Attkx2 is most likely correct, and the English description is a translation error.

/wave Kawfee

[Kawfee]

Hey Elmer

I'm inclined to agree with Archibald, where my gut reaction is that the pDIF is a normal curve... most things just automatically become one.

Even though I started it, I'd rather not let this thread get hijacked onto which sub is better, plzkthx ^^;

[VZX]

Clearly you know that it does more than a regular melee hit when attack isn't capped, from leveling thief yourself, so we're pointing in the right direction. Patience VZX, I'll get the rest of the data

I think that's what I mean 'it shows something'. If VB doesn't double your attack, you won't see significance damage difference on uncapped atk/def.
anyway, Let's not get stuck with VB and back to general formula

As far as the distribution 1.6-2.4, I am inclined to believe that it is a true random distribution. That is just my inclination, but in initial "looking at melee hits when I know where they can possibly land" makes me think they're pretty evenly distributed. Needs moar testing. I did notice hitting a ton of 129 crits, def. more than any other crit #.

I think it's better if you test on non-crit condition
who knows... they probably code the pDIF ranging from 2.6 to 3.2, but when it's got multiplied by base damage, they change all value from 3.001-3.200 to 3.000 (i.e. you see more occurece on high end side rather than on low-end side)

[aurik]

On this subject, I wrote a damage distribution tracker to my parser and I noticed a definite amount of clustering around particular values during merit. Particularly, my base damage (WD+fSTR) on the monsters.

From a few parses I came to the conclusion that it is actually an even distribution over an interval, and then any numbers below a floor or above a ceiling are set to the floor/cieling. A hypothetical example: if your attack is ~1.25xdef, it is evenly chosen from the interval [0.8-1.8], and then any numbers below 1.0 are set to 1.0.

[VZX]

On this subject, I wrote a damage distribution tracker to my parser and I noticed a definite amount of clustering around particular values during merit. Particularly, my base damage (WD+fSTR) on the monsters.

From a few parses I came to the conclusion that it is actually an even distribution over an interval, and then any numbers below a floor or above a ceiling are set to the floor/cieling. A hypothetical example: if your attack is ~1.25xdef, it is evenly chosen from the interval [0.8-1.8], and then any numbers below 1.0 are set to 1.0.

Your pDIF is ranging from [1.0-1.5] if you have cRatio = 1.25

anyway, if you say it's evenly distributed, then the value that occurs the most frequent is (1.0+1.5)/2 = 1.25 ?

Edit for some wrong number

[aurik]

On this subject, I wrote a damage distribution tracker to my parser and I noticed a definite amount of clustering around particular values during merit. Particularly, my base damage (WD+fSTR) on the monsters.

From a few parses I came to the conclusion that it is actually an even distribution over an interval, and then any numbers below a floor or above a ceiling are set to the floor/cieling. A hypothetical example: if your attack is ~1.25xdef, it is evenly chosen from the interval [0.8-1.8], and then any numbers below 1.0 are set to 1.0.

Your pDIF is ranging from [1.0-1.5] if you have cRatio = 1.25

anyway, if you say it's evenly distributed, then the value that occurs the most frequent is (1.0+1.5)/2 = 1.25 ?

Edit for some wrong number

Was a hypothetical example, I don't remember the exact numbers. But there was a huge spike in the frequency graph at WD+fSTR.

Like, there was a 60% chance that the multiplier was between (1.0-1.5], and a 40% chance that the multiplier was exactly 1.0.

[VZX]

Was a hypothetical example, I don't remember the exact numbers. But there was a huge spike in the frequency graph at WD+fSTR.

Like, there was a 60% chance that the multiplier was between (1.0-1.5], and a 40% chance that the multiplier was exactly 1.0.

yeah,
that's what you wrote in wiki, right?, that the distribution is more like this

     ++ +
   +        + 
  +            +
  +                +
 +                     +
 +                           +
+                                 +
min                               max

which is slightly skew to the left (average < median)
rather than

                 +
               +  +
              +    +
            +         +
         +               +
      +                      +
 +                                 +
min                              max

which is evenly distributed at average (average = median)

[aurik]

yeah,
that's what you wrote in wiki, right?, that the distribution is more like this

     ++ +
   +        + 
  +            +
  +                +
 +                     +
 +                           +
+                                 +
min                               max

which is slightly skew to the left (average < median)
rather than

                 +
               +  +
              +    +
            +         +
         +               +
      +                      +
 +                                 +
min                              max

which is evenly distributed at average (average = median)

More like

 +
 +
 +
 +
 +
 +++++++++++++++++++++++++++++++++++
                                   +
min=WD+fSTR                      max

for moderately high attack.

[Kawfee]

Hmmm, that's very interesting o.O
If it's not a normal distribution, my math skills at this point in life prevent me from knowing how to take that into account lol.

I'll let you guys take over <.<

[Genosync]

You know, I don't know if it was mentioned in a post because I don't care enough to read this thread, but I want to make sure that the OP fucked up his numbers because 2STR = 1ATK and isn't in the data.

[VZX]

You know, I don't know if it was mentioned in a post because I don't care enough to read this thread, but I want to make sure that the OP fucked up his numbers because 2STR = 1ATK and isn't in the data.

it's been fixed on page 4

[Faranim]

WHAT LOL
Is that thing saying
you did 48% of Cerebus' HP or am I reading it wrong!

Parsing Cerberus is pretty silly to begin with. So much damage gets absorbed by Stoneskin it can easily skew the results.

Also, lolSMN Damage doesn't show up in the chat log if they release their Avatar quickly (bug been around forever... it's like how SC damage doesn't appear if you change equips).

But I'm still pretty sure Ruke did the most damage by a fairly large margin.