Mind Flay coefficient Edit
Hey! I've just noticed that you changed the Mind Flay coefficient from 57% to 80%. Are you sure you tested it without any damage increasing talents? I'm quite sure that Mind Flay does not receive so much, and I really doubt Blizzard has changed it in the last patch. --bfx 06:47, 8 February 2008 (UTC)
Mind Flay Edit
You are right about not removing the extra damage talents. I have recalc'd after removing those; however, the coefficients for dots on the wiki page are fundamentally flawed because the posters used total damage rather than per tick damage.
You can't calculate spell coefficients on total DoT damage. This method would be skewed by things like Improved SWP or armor pieces that add ticks to the spell (Absolution set bonus). The only thing that makes sense is to calculate the coefficient per tick. After running a number of tests and compensating for Darkness, Misery and Shadow Weaving. The coefficient per tick on Mind Flay is about 22% per tick. Calculating based on total damage gives ~65.89%.
I calculated the following coefficients by removing 141 bonus damage and nothing else from my stats. Crits were ignored.
Mind Blast 50.28%
Shadow Word: Death 52.38%
Mind Flay 22% per tick
Vampiric Touch 23% per tick
Plague 11.28% per tick
Touch of Weakness 13.48%
Shadow Word: Pain 21% per tick
Shadow Fiend 6.43% or 5.85% if the fiend is affected by darkness (did not test).
Shadow Fiend generates 2.5x mana per damage point and crits 2x over base damage.
- Personally, I don't feel there's anything wrong with the coefficients. They represent the amount of extra damage a 'naked' spell gains, hence they are applied before talents are. Any talents that increase damage (Darkness) or duration (Improved SW:P) need to be calculated afterwards. Per tick notation does not change the fact that the coefficient does not reflect the actual damage gained (a 6 tick and an 8 tick SW:P have the same coefficient per tick, yet one gets a considerably higher amount of spell damage in total than the other).
- Furthermore, your numbers confuse me. Spells like MB, SW:D, VT, and DP should follow normal rules and receive 43%, 43%, 100% (20% per tick), and 80% (10% per tick), respectively. It's very likely that you forgot some factor in your calculations, thence getting a wrong coefficient on Mind Flay as well. The best way to solve this problem is to unlearn all talents and then only take those which do not affect damage done. I'll probably run some tests as soon as I get access to the PTR. --bfx 09:52, 12 February 2008 (UTC)
"Per tick notation does not change the fact that the coefficient does not reflect the actual damage gained (a 6 tick and an 8 tick SW:P have the same coefficient per tick, yet one gets a considerably higher amount of spell damage in total than the other)."
Thank you for making my point. You get considerably more damage with more ticks. Do the math yourself; having more ticks changes the apparent coefficient, which is why you can't do it that way. Why in the world would a blizzard programmer base the damage addition (coefficient) on total damage for spells that have talents that change ticks? It makes absolutely no sense. Add 20 ticks to any of the DoT spells and see how far off your coefficient gets. In fact, take the shadow fiend as an example, it gets 10 or 11 hits before dying. Which "coefficient" are you going to use, the one where he gets 10 in or 11? Is the 10 hit fiend the "naked" spell or is the "naked" fiend 11 ticks? The only value these coefficients have is to calculate by formula how much more damage a toon will get without actually doing hundreds of spells in tests. Having per tick damage cofficients and adding additional ticks based on your build will give you that total damage number. Having overall damage coefficients that are based on the authors personal tick count, "or what he considers naked," will only work for that one person. How is that helpful?
Here is the math laid out:
SWP damage per tick: 430, add 141 more bonus damage and damage per tick is, 460
6 tick coefficient: 430 * 6 = 2580. 460 * 6 = 2760. (2760-2580) / 141 = 1.2766
8 tick coefficient: 430 * 8 = 3440. 460 * 8 = 3680. (3680-3440) / 141 = 1.7021
1.2766 <> 1.7021 If I post either of these numbers, no one would be able to calculate their build's total damage unless I told them how many ticks I used. However, if I post the unchanging damage per tick value, they can. I'm not infallible, show me where I'm wrong.
At some point when I have the time, I can rerun hundreds of spells with a "naked" build to see if there is some reason that simple math can't subtract out their (darkness,misery,shadow weaving) affects. I'm curious why you think the coefficients must follow your "43%, 43%, 100% (20% per tick), and 80% (10% per tick)." Based on what? Did you run hundreds of spells or are you just repeating the numbers of others? --diabolus
Update 2/12/2008 4:43UTC ---
I rebuilt without any damage talents and ran the tests again on the spells that don't require hundreds of casts to get accurate numbers and the values were nearly identical (differed in the hundredths except ToW). The values on the wiki page are just flat out wrong. Couple that with the fact that they are not trying to find the value the programmer used as an actual constant in the WoW code, but some "total damage" nonsense, making them wrong and confusing. Here are the updated numbers for the spells I tested.
Used 141 delta and then tried 261 point bonus damage delta by adding food/wiz oil/flask of pure death, to get more accurate numbers from better spread. Both spreads yielded nearly identical results:
Mind Flay 22.06% per tick
Vampiric Touch 22.87% per tick
Plague 11.48% per tick
Touch of Weakness 12.26%
Shadow Word: Pain 21.02% per tick
Again, prove me wrong. --diabolus
- Regarding the per tick vs. per spell notation: It's mostly a matter of preference, both systems are valid. I prefer per spell notation, because I quickly get an overview of how much spell damage a spell gets. The per tick notation requires to determine the amount of ticks and then to calculate the coefficient. If you dislike the current system, feel free to start a new discussion here. IMHO, the only cases where per spell notation makes no sense is for spells like Shadow Fiend and Searing Totem (which either do not have a fixed amount of "ticks" or where using ticks is unhandy).
- Regarding the coefficients: My numbers are based on both the coefficient formula provided by Blizzard and blue posts from the official forums. The formula should apply to most spells, with only a few exceptions (including Mind Flay and SW:P). I'm in no way trying to prove you wrong, I merely was suspicious of your numbers, since I have never seen them before and they make no sense (speaking of the before-mentioned formula). If they are true, so be it. My intention was to alert you of a potential error. If you feel that your calculations are correct, don't hesitate to update the coefficients on the Spell damage coefficient page. --bfx 09:26, 13 February 2008 (UTC)
To Tick or not to Tick Edit
I guess our disagreement on the per tick versus total damage comes from my desire to know what constant Blizz used in the code. I am a computer programmer/scientist first, WoW player a distant 11th. The constant they used represents the truth, the other a pseudo truth. Can it be made to work except on the fiend - yes, but as you noted about your gold changes on your personal page, why not be "consistent" since the fiend does not work that way.
After all the tests I ran on MB and SWD I believe the error margin is still about +-2%, and I'd like to run more tests to squeeze that down before updating those; however, due to the lack of variability in the other spells, I'm confident they are quite close to the actual value used in the code. ToW is a problem because its number is so small. Since Blizz rounds to the nearest damage point, one point difference can sway the number nearly 0.5% in either direction using a 261 delta.
Once I get the error margin below 1% on MB and SWD, I'll update the page.
One of the reasons I started this is because I keep running into invalid numbers throughout the WoW data. On ShadowPriest.com they have a gear section where they claim [(CritRating/6.14) = The amount of bonus spell damage it equates to.] I wrote a computer program to generate a VT priorized max spell cycle. The program loops through all possbile starting spell squences to mazimize DPS in a 1min, 2min, up to 6min spell cycle. Looking at how many times SWD and MB (the only spells that crit), come up, that value doesn't work and it doesn't for practical play reasons in addition to computer generated perfection. Unfortunately no one is doing any of their own work and just continuing to post values they see elsewhere. It's the old, "if it's in print, it must be true" problem. --diabolus
- Maintaining consistency is important indeed, but in this case, multiple forms of concistency collide. Do we keep it consistent in terms of technical implementation (then we should use per tick notation). However, in this case, we'd probably have to be go further and update all DoT pages with per tick damage and not total damage. We can also stay consistent and keep to total base coefficient for all spells, regardless of whether they are DoTs or DDs (which would improve readability, since then the coefficient of each spell can be easily determined and compared to others'). In this case, I tend towards using the second approach, even though I entirely understand your point. This is just my personal preference, and I believe this issue qualifies for being discussed by more people. --bfx 08:03, 14 February 2008 (UTC)
Found the problem Edit
I kept going over and over the numbers, looking for anything that looked wrong or typos in the spreadsheet and could find nothing. Finally while driving to work; it struck me why the numbers are all slightly higher. I never fight out of Shadow Form. It's just not a "talent" I think about. All the tests were done with the 15% buff.
Since I find it unlikely that Blizz has a fascination with irrational numbers as the values are all extremely close, within the margin of error, of whole numbers, rounding off the Shadow Form compensated numbers gives the following:
Mind Flay 19% per tick (or 57% total)
Vampiric Touch 20% (or 100% total)
Plague 10% per tick (or 80% total)
Shadow Word Pain 18% (or 108% total)
Touch of Weakness 10.66 (I'll have to use a bigger delta to see if this converges on 10%. As it stands, it looks more likely to converge on 11%)
These are all very close to the posted coefficients with the exception of SWP. The per tick coefficient is small making the error margin larger like ToW and Plague, but this is a difficult gap to bridge. Tonight I'll try to increase the delta farther to see if this number holds. I suspect it will.
Many more MB and SWD's are still needed to get those numbers to settle down, but they are currently within the margin of error of the posted values. I suspect these coefficients are also whole numbers, and am figuring the posted values are from small sample sizes.
I will also attempt to determine if the Fiend is affected by Darkness. --diabolus
- Great, it does make a lot more sense now. Everything is as it should be. IIRC, a blue post claimed that SW:P's efficient was set to 110%, which comes very close. --bfx 11:48, 15 February 2008 (UTC)
SWP & ToW Edit
With a 772 point delta, SWP came out 18.3182% per tick or 109.909%, this is within the margin of error of 110%; however, ToW moved further away with a value of 10.7254%. I suppose it is possible that this coefficient is not a whole number, however 11% is within the error margin, but 10% is not. Should this coefficient also be a whole number like all the others, it would have to be 11%. --diabolus