Lvl 85 weapon comparisonsFollow

Maybe we don't know the exact formula, but you can certainly make a histogram from the data summary in Kparser over a massive number of fights and see what the distribution is (is it normal? it is evenly distributed across the range? is it skewed high or low?). Probably someone has done this. You could then approximate it with something closer to reality than just an average. Either way, it doesn't change that your model is, well, without much value. You can say you don't know, but that doesn't mean you don't need to know.



OK, now assume you always punch at maxpdif, then you will have a fight where you kill quicker. Now, calculate the kill time for every possible average damage per fight across the distribution and multiply by the percent of the times each average will pop up. Sum those up to get an average kill time. That average kill time will vary at a smaller step size than the delay between attack cycles (meaning you won't see these widely stepped tiers of kill times like in your model), and it will vary with your attack.



Which is a massive flaw in the model. Crits and regular hits operate differently enough that you need to consider probabilities of them occurring X number times per fight and for a distribution of damage.

"Reduced variance" in this sentence means nothing. "Reduced" from what? You're just trying to sound smart by using terms you don't understand. And the variance ISN'T that small. Scrolling in my log right now, my last 5 hits on MNK: 119, 108, 121, 133, 105. I did a Tornado Kick of 2428 (probably a double attack) then one of 622 (obviously a missed hit). Seems to be pretty high variance to me.



I'm not talking about Kinematic's model. I am talking about flaws in yours. Whether Kine has flaws or not, doesn't change that your model is worthless.



If it were realistic at all, it would pass the common sense test that more attack = more damage = faster kill speeds. Common sense and every day experience tells us that our fights DON'T have durations which are quantized to our weapon delay.

lynn, none of this math is hard. you seem very self-congratulatory in your error. stop posting.

Ok, I've been able to replicate the base idea lynn is asserting. I'm posting this to help explain the overall effects.

Simplified conditions (no kicks, 95% acc, 5% DA)

For a given amount of HP remaining, n hits are required on average based on average damage per hit (standard model with cRatio, crits, etc).

For the n hits needed to kill the mob, r rounds are required on average. Same principle as the weaponskill TP model.

Given that, how does the average number of rounds needed vary based on mob HP and player attack?

In general, for varying mob HP, the number of rounds changes in increments roughly matching the average damage per hit (as might be expected). If average damage per hit is 100, about every 100 mob HP (mob HP can vary based on mob type/level/outside sources of damage/etc) will have a different average number of rounds required.

For player attack, there are a number of factors that determine the extent of the scaling (eg: crit rate, level correction, etc). In general, however, changing attack can be considered a 'shift' in perceived mob HP for the purposes of calculating average number of rounds needed. That shift appears to be approximately 5 to 8 HP per 1 point of attack for a few common variances (higher effectiveness with lower crit rates; 8 HP was at 10% crit rate, 5 HP was at 54% crit rate), and drops with higher attack values (got down to 4 HP at near capped attack).


So, for example:

Setup:
Attack: 480
Defense: 350
Level Correction: 0.25
Crit rate: 54%
Crit damage: +30%
Base damage + fStr: 79

Gives:
cRatio of 1.121
Avg damage per hit of 158.403


A mob with 3009 HP will take an average of 9.802 rounds to kill
A mob with 3010 HP will take an average of 10.293 rounds to kill
A mob with 3169 HP will take an average of 10.803 rounds to kill

You'll note that there's a gap of about 159 damage where the average number of rounds remains constant (3010 HP to 3168 HP), which matches the average damage per hit.

Now, add 1 point of attack. Total attack of 481, avg damage per hit of 158.665. The mob now needs 3015 HP for it to take an average of 10.293 rounds to kill, and the HP range of 3010-3014 which used to take 10.293 rounds now takes 9.802 rounds.

5 HP out of 159 is about 3%, so adding 1 attack gives about a 3% chance of dropping about half an attack round on average, and out of 10 rounds has an average fight time reduction of 0.14%. Compare that with the .165% increase in cRatio due to the added attack.

Edit: The value scales with the HP of the mob such that you always get fairly comparable total reduction in average fight time.

For most purposes, mob HP has to be treated as infinitely variable. Other DDs doing damage to the mob (melee or nukers), DOT spells such as Dia, counters/Perfect Counters, level variation (can have a dramatic effect inside Abyssea, though a somewhat more limited impact outside), job variation (eg: different types of Mamool Ja or Dynamis mobs), weaponskills (as we're only looking at the melee side), mobs healing themselves and so on and so forth all mean that we can never expect the target amount of damage done to always stay within a predetermined range. That means there's always a chance for that shift in HP to have an impact in average number of rounds to kill.


Overall: Yes, lynn is correct in that adding 1 attack will often have no impact on average fight length. He is incorrect, however, in asserting that it will *never* have an impact, and that very large changes in attack are necessary in order to manifest any difference in fight length. This is because he was looking at a fixed value for HP rather than considering the variability of it, or the 'jitter' as milich notes.

Higher starting attack and crit rate and more crit damage (as we have in Abyssea) all dampen the value of additional attack, but never negate it entirely (until attack is capped, of course).


It was interesting to look at the value of attack from a different perspective, but I'm going to maintain my current methodology for DPS calculation.



Edited, Dec 6th 2010 11:11am by Kinematics

Tantra Cyclas +2 Rare Exclusive
Body
DEF:64 STR+12 VIT+12 Accuracy+15
Attack+15 Augments "Impetus"
Enhances "Martial Arts" effect
Set: Augments "Kick Attacks"
Lv89 MNK


Usukane what?

Figured I would post here, since this has been the main numbers thread, but I was wondering if either of these atmas may become worthy and used in conjunction with our new three limit

Atma of the Hybrid Beast

Double Attack+:Minor
Triple Attack+:Minor
Enhances "Arcana Killer" Effect:Minor


Atma of Alpha and Omega
Triple Attack:Major
Attack:Major (+50)
HP-:Major (-25%)