Website Feedback
Closed Threads
IRC Web ChatTeamSpeak 3 (53 users)
Active: 6040 users | |
|
| Kwanny Germany. August 06 2012 03:49. Posts 219 | Profile # |
This has been really bugging me.
I'm trying to calculate the weapon damage on my weapon, which has following offensive stats.
1.4 A/s
404-914 Damage
+25% Damage
+293-611 Arcane Damage
Total: 923.3 Dps
calculated then: (404+914)/2*1.4 does not equal 923.3 DPS, but rather 922.6
while (404+914+1)/2*1.4 = 923.3 DPS
With
1485 Prim Stat (Dex),
12 Crit%
50% CritHit Dam,
+6% ias
+40-49 Dam on armor,
it should equal, (404+914+40+49)/2*(1.4*(1+0.06))*((0.12*(1+0.5)+(1-0.12))*(1485/100+1)
which gives me 17430.14 DPS,
but ingame it reads: 17552.61, which would again only work, if you add +1 damage anywhere to the weapon or armor.
Similar differences appear on several other 1 hand melee weapons, close to, but not always 1.
A wiz weapon with
312-678 and +11% attack speed shows 1.55 Aps, with 769.2 DPS,
1.4*1.11=1.554 actually, with (312+678)/2 equalling 769.23 DPS, which is perfectly right.
Same works perfectly with bows, ceremonial knives etc.
Any clue why this is happening?
|
|
|
| zipz0p United States. August 06 2012 04:34. Posts 111 | Profile # |
| I'm not sure about this, but noticed similarly that my wiz, who uses a wand has perfect dps prediction in my little spreadsheet (doing the calculation you outline), while my barb is always off by a bit. I hadn't figured out why, but I'll have to check the +1 damage difference. |
| |
|
| Kwanny Germany. August 06 2012 04:46. Posts 219 | Profile # |
I can imagine why to some extend, as weapons by themselves have a base damage,
such as
Heaven Hand
Fist Weapon
242.2–248.5
Damage Per Second
(104-107)–(242-248) Damage
1.40 Attacks per Second
Item Level: 63
and to those numbers, the affixes such as +dam% and +elemental is added/multiplied afterwards.
Still, I don't really understand, why it only happens for "dual wieldable" melee class weapons. |
|
|
| Cascade Australia. August 06 2012 11:03. Posts 2197 | Profile Blog # |
At a first glance I would guess that it is about rounding. Maybe the randomised (?) weapon damage is not integer only, so that the damage isn't 404-914 but maybe 404.4-914.4, so that the displayed damage is slightly lower than the actual damage. Then when multiplied with 1.4, the extra undisplayed (0.4+0.4)/2*1.4 = 0.4*1.4 = 0.56 (which rounds to 0.6, which is the difference you see) will seem to appear out of nowhere.
No idea why this wouldn't happen with other weapons though... Do you always get more than expected dps, or can it sometimes be less as well?Last edit: 2012-08-06 11:06:40 |
|
|
| Kwanny Germany. August 06 2012 12:55. Posts 219 | Profile # |
Sometimes it's more, sometimes it's less. Pretty random.
I'll look for a couple of examples tomorrow.
edit: Or you could check it out youself aswell  Last edit: 2012-08-06 12:56:00 |
|
|
| Cascade Australia. August 06 2012 14:21. Posts 2197 | Profile Blog # |
On August 06 2012 12:55 Kwanny wrote:Sometimes it's more, sometimes it's less. Pretty random. I'll look for a couple of examples tomorrow. edit: Or you could check it out youself aswell
Well, my computer is on a boat from Europe to Australia, so I cannot actually. 
The fact that you get both more and less dps (about equally often in each direction?) is consistent with rounding errors I guess. Dual wield remains a mystery for me though.
The only thing that really comes to my mind is if dual wield weapons for some reason would be the only kind of weapons that have non-integer damage. Maybe they can have different damage in different situations or something? And the different situations are related by some multiplier or something. Then the displayed damage could be the original (integer) damage multiplied with the non-integer multiplier resulting in a non-integer damage, and we would get rounding errors.
Or it is about something completely else.
edit
Actually, a good way to test the rounding hypothesis is to calculate the "actual" damage of a weapon from the dps. Now you calculate the dps from the displayed damage, but you can also use the dps and calculate the damage. Just do the same, but backwards. So in the first example you would get the calculated damage 404.4..-914.4... compared to the displayed damage 404-914. Which is consistent with a rounding error. If you do this for a few weapons, and all of them end up with calculated damage within +- 0.5 of the displayed damage, it is consistent with a rounding error. If we find even a single weapon that goes outside the +- 0.5, we can discard the rounding idea, or at least have to revise it.Last edit: 2012-08-06 14:27:35 |
|
|
| Kwanny Germany. August 07 2012 05:43. Posts 219 | Profile # |
Okay, I did a quick test using random samples from the AH (hardcore), I stumbled upon this:
Everytime when there is a weapon that has +%damage, or iA/s, the calculation has a very high probability (from my sample, over 95%) that the damage listed does not match the calculated DPS on the weapon.
Also, 1h Axes have a probability (suprisingly, or maybe not, exactly 50% from my sample) that, even if +%damage, or iA/s is NOT on the weapon, the listed damage is off by exactly +/- 0.05 from the calculated damage. Same occurs with Daggers. These are attack speeds of 1.3 / 1.5 respectively.
With 1.2/1.4/1.6 attack speeds, these 0.5 deviations don't occur, and the basic calculation of (min+max)/2 is always correct, again only in the cases where there is no +%damage or iA/s.
Wondering what I would find with other, weird attack speeds, I looked through polearms with 0.95 attacks per second:
There is a probability that a polearm without +%damage and iA/s can have a deviation by exactly +/- 0.025 or exactly +/-0.05. I have no clue about the distribution, as my sample size is too small.
So here's some food for though.
EDIT: Just reading through the previous posts:
Seeing that there can be a real, actual deviation by 0.05 to the upward direction would probably suggest, that it should not really be a rounding error, but something else. Where does blizzard hire all those people programming stuff like this...
2nd EDIT: Ah, nevermind, it's 0.05, not 0.5.
From my samples, the highest difference of calculated to listed is exactly 0.5, occuring twice. Once with higher calculated, and once with lower calculated.Last edit: 2012-08-07 05:55:45 |
|
|
| l3iRdMaN United States. August 07 2012 11:28. Posts 71 | Profile # |
| any particular reason other than curiosity as to why you're spending so much time worrying about fractions of a percent difference? |
|
|
| Cascade Australia. August 07 2012 11:57. Posts 2197 | Profile Blog # |
On August 07 2012 11:28 l3iRdMaN wrote:
any particular reason other than curiosity as to why you're spending so much time worrying about fractions of a percent difference?
Turns out that curiosity not only kills cats, but also a lot of time! |
|
|
| Kwanny Germany. August 07 2012 20:34. Posts 219 | Profile # |
| Nah, Curiosity is on Mars. I'm simply annoyed by the fact that you can't perfectly calculate stuff based on all the info you have. Trying to change that. |
|
|
| Markwerf Netherlands. August 07 2012 22:34. Posts 2916 | Profile # |
Probably a rounding error. I figure the game generates integer values for the weapon damage but if a weapon includes the stat +x% damage the damage won't actually be whole numbers again. So i figure that weapons with +% damage give slightly different DPS numbers than shown if you calculate them with the rounded numbers.
|
|
|
| |
|
|
|
|
|
|
| |
|
Sidebar Settings...
|
Store
