Effects of ATK

[DiosT]

Post by Coprolith, edited by DiosT

0. Foreword

This post is a compilation of the results I presented in the original 'Effects of ATK' post, the two updates after that, and several pages worth of discussion spread out in two threads. I made this compilation by request of the moderators for the beastlore section. It follows the same structure as the original post, but the discussion of the results in Chapter 5 has been changed significantly to incorporate the new insights obtained from the various discussions.


1. Introduction

Since SOE was trying to lure me back into playing EQ by giving me a free 14 day trial subscription I decided to use that time doing some parses i've always wanted to do on the effects of ATK rating on melee dps. Not the actual value of my dps, but rather its effect on the two main factors that make up melee dps, average damage per hit and the hit ratio. For the latter it has been known some time that strength and pure attack buffs have no effect on it, but i've seen several posts recently of people that didn't know/believe this. Plus, it never hurts to check this these things because you never know if SOE have changed the underlying mechanics since the last investigation.
The effect of ATK rating on the average damage per hit is more complex. A good description of it would considerably help with making theoretical dps predictions for weapon comparisons, understanding the effects of lower 1HB, 2HB and piercing skill caps, and buff/focus effects.


2. Theoretical background

It's long been thought that ATK is actually the sum of two separate numbers, one representing a 'to hit' rating and one representing a 'damage' rating. The 'to hit' rating is determined by the offensive skills only, buffs have no effect on it. Once the weapon skills are capped there is no way to vary this part of the ATK rating except thru skill modifiers.
There are two formulas i will refer to later on. The general formula for the melee dps of a weapon is

   melee dps = swings/sec * Hit ratio * average dmg/hit.   (equation 1)

The average damage/hit itself can be represented as

   average damage/hit = n(ATK) * base damage + damage bonus.   (equation 2)

Base damage and damage bonus are weapon- and level-dependent. With weapon skills at the cap, the ATK rating enters into dps only thru the multiplier n. The relation between n and ATK is the main focus of this investigation. Swings/sec can be easily calculated. Hit ratio depends on whether or not you are tanking or hitting from behind and the level of the mob, but approcimate values for the hitratio are known for pretty much every condition. It's the relation between average damage and ATK that has always been the limiting factor in theoretical dps calculations. Apart from this practical application, the underlying mechanics are interesting by itself.


3. Parse set-up and details

All parses were done with my Lvl 65 beastlord (Magelo profile in sig). Two series of parses were performed. The first, and largest, series was done attacking from behind against the banker in Katta, Heracus Helsin, a Lvl 1 unkillable NPC that doesnt fight back. The second series of parses was done tanking Broken Skull Bloodguards in Torgiran Mines. These are level 50-52 mobs with approx. 10k hps. Only weapon melee damage was parsed, no kicks or slams. The only relevant offensive AA skill is BF5, which only function in this case is to increase the sample rate by 15% anyway.
Weapons used are:
HtH: Waning Light Katar, 16/22 +13
1HB: Club of the Ice Ocean, 14/27 +13
2HB: Staff of Flowing Water, 42/34 +38
All weapon skills are at the cap, i.e. 250 HtH and 225 1HB, 2HB.

Variation of the ATK rating was obtained with various buffs and swapping out equipment. Unbuffed parses were performed overnight. Parses with buffs on require some user interaction, and in those cases i also used Celerity simply to speed up the sample rate. The Bloodguard parses were done during in normal exp sessions. Sections of the log file where an ATK buff had faded were removed before parsing the file, so each parse is done at constant ATK.

Since i was tanking the bloodguards the hit ratios cannot be directly compared with the Katta banker results because of dodges, parries and ripostes. I therefore choose to show (1 - miss ratio) instead. For the Katta banker that number is the same as the hit ratio, for the Bloodguards (1 - miss ratio) is what the hit ratio would have been had i been hitting them from behind instead of tanking them



4. Results

The first series consists of a total of 10 parses. Parse details and results are given in table 1.

Table 1: Katta banker parse results

Test #	1	2	3	4	5	6	7	8	9	10
Buffs(*)	n	S	S+F	n	S+F	n	S+F	S+FrW	n, -g	S+FrWF
Strength	175	272	272	175	272	172	269	280	115	280
MH	HtH	HtH	HtH	1HB	1HB	2HB	2HB	1HB	1HB	1HB
MH ATK	1129	1270	1471	1061	1404	1059	1401	1293	987	1494
Swings	14774	9130	7856	15366	7843	13721	9216	7704	21162	8335
Hits	10568	6516	5599	10682	5402	9551	6395	5378	14628	5813
Misses	4206	2614	2257	4684	2441	4170	2821	2326	6534	2522
Hit ratio	71.53%	71.37%	71.27%	69.52%	68.88%	69.61%	69.39%	69.81%	69.12%	69.74%
- error	0.74%	0.95%	1.02%	0.74%	1.05%	0.79%	0.96%	1.05%	0.64%	1.01%
Min hit	15	15	15	14	14	42	42	14	14	14
Max hit	92	109	111	78	99	228	289	99	71	99
Avg hit	48.71	56.88	65.69	42.14	58.62	123.79	173.02	54.46	37.54	61.14
- error	0.47	0.70	0.88	0.41	0.80	1.27	2.16	0.74	0.31	0.80
n	2.232	2.743	3.293	2.081	3.259	2.043	3.215	2.962	1.753	3.438
- error	0.022	0.034	0.044	0.020	0.044	0.021	0.040	0.040	0.014	0.045
OH	1HB	1HB	1HB	HtH	HtH	-	-	HtH	HtH	HtH
OH ATK	1061	1202	1404	1129	1471	-	-	1360	1052	1561
Swings	10115	6317	5429	15498	7788	-	-	7662	21418	8257
Hits	6926	4435	3733	10994	5505	-	-	5409	15191	5888
Misses	3189	1882	1696	4504	2283	-	-	2253	6227	2369
Hit ratio	68.47%	70.21%	68.76%	70.94%	70.69%	-	-	70.60%	70.93%	71.31%
- error	0.92%	1.15%	1.26%	0.73%	1.03%	-	-	1.04%	0.62%	1.00%
Min hit	1	1	1	2	2	-	-	2	2	2
Max hit	65	81	86	79	98	-	-	98	69	98
Avg hit	29.31	36.67	45.84	35.50	53.67	-	-	48.61	30.66	55.65
- error	0.35	0.55	0.75	0.34	0.72	-	-	0.66	0.25	0.73
n	2.093	2.620	3.275	2.219	3.354	-	-	3.038	1.916	3.478
- error	0.025	0.039	0.054	0.021	0.045	-	-	0.041	0.016	0.045

(*) explanation of buffs:
n = none
S = Spiritual Vigor + Infusion of Spirit + Girdle of Karana + Celerity
F = Ferocity
Fr = Frenzy
W = Shared Form of the Great Wolf
-g = all gear except weapons and haste belt removed


The second table gives the results of the Bloodguard parses

Table 2: Bloodguard parse results

Test #	11	12	13
Buffs(*)	S+Fr	S+FrF	n
Strength	280	280	180
MH	1HB	1HB	1HB
MH ATK	1209	1411	1067
Swings	5888	6258	6322
Misses	2163	2356	2337
Hit ratio	63.26%	62.35%	63.03%
- error	1.26%	1.22%	1.21%
Min hit	14	14	14
Max hit	92	96	77
Avg hit	48.85	57.64	40.44
- error	0.86	0.99	0.69
n	2.561	3.189	1.960
- error	0.045	0.055	0.033
OH	HtH	HtH	HtH
OH ATK	1276	1478	1134
Swings	5686	6068	6025
Misses	1996	2170	2153
Hit ratio	64.90%	64.24%	64.27%
- error	1.27%	1.23%	1.23%
Min hit	2	2	2
Max hit	94	95	77
Avg hit	42.27	51.02	33.58
- error	0.74	0.87	0.58
n	2.642	3.189	2.099
- error	DiosT Wildblood - 70 to 74 Beastlord Savage Lord Posts: 802 Location: Xegony Server Logged More on ATK #1 January 03, 2004, 02:10:00 AM Continued from above - Hit ratio depends on offensive skills only, ATK buffs have no effect on it - Confirmation that the dualwield rate at lvl 65 is higher then previously assumed - In the ATK range between 1060 and 1470, the multiplier for the base damage, n, depends linearly on ATK. A master curve can be constructed to account for differences in weapon skills. - A change in ATK gives a corresponding change in n equal to (0.00338+/-0.00006) times the change in ATK - With these results its possible to calculate the melee dps of any weapon in a particular situation once you've got a parse of a single weapon in that situation - There's no truth to the statement that you need high ATK for 2HB to be effective. Dualwielding will generally benefit more from high ATK then 2HB More On Attack: Posted by Coprolith, I got Frenzy of Spirit last night and parsed out the result of 15 FoS sessions against the Katta banker. HtH ATK was 1911 and 1HB ATK was 1844 with FoS (1575/1508 w/o). Statistics arent very good (only 1133 MH swings and 845 OH) but any longer and my brain would have dribbled out of my ears. Anyways, here's the updated graph: The return you get from ATK keeps diminishing at high ATK (as expected) but its still a very significant increase. /hug DiosT Wildblood - 70 to 74 Beastlord Savage Lord Posts: 802 Location: Xegony Server Logged Even more on ATK #2 January 03, 2004, 02:10:49 AM Posted by Coprolith, Okies, here's what i've got so far - For my baseline i use test #9 from the original series: no gear except for haste belt and weapons STR=115 MH - CotIO, 1HB, ATK = 987 min hit = 14, max hit = 71 #hits = 14628 Avg dam = 37.54+/-0.31 n = 1.753+/-0.014 OH - WLK, HtH, ATK = 1052 min hit = 2, max hit = 69 #hits = 15191 Avg dam = 30.66+/-0.25 n = 1.916+/-0.016 I did 2 4h parses against the Katta banker today: - STR gear only, buffs: Infusion of Spirit, Furious Strength, Frenzy STR = 280 MH - CotIO, 1HB, ATK = 1135 min hit = 14, max hit = 84 #hits = 7293 Avg dam = 46.14+/-0.54 n = 2.367+/-0.028 OH - WLK, HtH, ATK = 1202 min hit = 2, max hit = 85 #hits = 7370 Avg dam = 39.36+/-0.46 n = 2.460+/-0.029 - No STR gear, Veng2, Spiritual Vigor, Form of the Great Wolf STR = 115 MH - CotIO, 1HB, ATK = 1138 min hit = 14, max hit = 84 #hits = 7656 Avg dam = 46.06+/-0.53 n = 2.362+/-0.027 OH - WLK, HtH, ATK = 1205 min hit = 2, max hit = 85 #hits = 7745 Avg dam = 39.82+/-0.45 n = 2.489+/-0.028 I had intended to perform 2 more overnight tests at intermediate values, one with only STR gear and one with only Vengeance items, but after seeing the results from the first 2 parses that exercise seems moot. Combining these results with my previous analysis of max hits and avg dmg the end result is unambiguous: There is no difference whatsoever between ATK gained from STR and from pure ATK; both max hit and avg dmg are exactly the same. The new data points fit exactly onto the linear behaviour i found before: It would certainly be pleasing if they changed the system so that ATK from strength was treated no differently than "pure" ATK - it was always annoying and cumbersome to have to consider the two differently. I couldn't agree more. It makes more sense this way. There is now only two parts of the ATK-rating to consider: the 'to hit'-rating, determined by skill only, and a 'damage'-rating, determined by skill, STR and pure ATK Weapon skill goes 50/50 towards 'to hit' and 'damage'. I cannot say what the exact effect of the Offense skill is, but i assume its similar to weapon skill. STR and pure ATK have the exact same effect if the increase in the listed ATK is the same If it turns out that a +50 strength buff is close to a +50 ATK buff (for those who aren't capped), that may change - I know that uber players/raiders reach the stat caps fairly easily, even unbuffed, but for casual players it takes a LONG time before that is feasible (I know that I have under 200 unbuffed strength - partly because I've believed it to be viable for ignoring). To be precise: - 1 point of STR adds 0.9 points to listed ATK - 1 point of pure ATK adds 1.33 points to the listed ATK (melee class bonus) So a Vengeance 1 item is worth only 7.4 points of STR, provided your STR isn't maxxed out. For a L65 beastlord, maxxing out STR is easy to do even for the non-ubers by just using our own buffs. For non-uber paladins, warriors, rogues and monks who have very little in the way of melee buffs and have to keep up other stats as well (e.g. AC and WIS for the paladin) I imagine this changes the way (s)he looks at gear selection more drastically then a BL /hug Romidar Brute Posts: 2 Logged Effects of ATK #3 January 04, 2004, 05:02:55 AM Thank you for the additional data and analysis - I actually thought that sort of parsing was fun 4 years ago but find it too tiresome now. To be precise: - 1 point of STR adds 0.9 points to listed ATK - 1 point of pure ATK adds 1.33 points to the listed ATK (melee class bonus) So a Vengeance 1 item is worth only 7.4 points of STR, provided your STR isn't maxxed out. For a L65 beastlord, maxxing out STR is easy to do even for the non-ubers by just using our own buffs. For non-uber paladins, warriors, rogues and monks who have very little in the way of melee buffs and have to keep up other stats as well (e.g. AC and WIS for the paladin) I imagine this changes the way (s)he looks at gear selection more drastically then a BL To your point, checking my Magelo, I'm at 183 strength, including stat food. I include my Magelo just as indication of what the pretty much highest end paladin equipment is for a non-raider. Given my normal group, which includes a shaman, I still wouldn't bother with increasing the strength on my gear; however, it'll now at least be a factor in evaluatiing equipment whereas it truly was not previously. Your findings also make the Planar Power ability a bit more appealing than it had been, in that the ~22 increase in ATK will actually be meaningful. One thing I haven't checked - just to be certain - is to make sure that your relationship of 0.9 ATK per 1 strength holds up for other classes. I would be surprised if there was a difference between beastlords and paladins in that regard, but I wouldn't really be floored, either. Coprolith Savage Lord Posts: 787 Location: Netherlands Logged Effects of ATK #4 January 04, 2004, 06:52:17 PM One thing I haven't checked - just to be certain - is to make sure that your relationship of 0.9 ATK per 1 strength holds up for other classes. I would be surprised if there was a difference between beastlords and paladins in that regard, but I wouldn't really be floored, either Well that's easily checked. My lvl 60 paladin with all his gear on has 161 STR and 993 ATK. Stripped of his gear his STR=113 and ATK=950. So 43 ATK per 48 STR equates to 0.896. With roundoff error in the listed ATK i'd say the relation holds. It may be that classes that are on a higher damage table get more ATK per STR. BLs and paladins are on the same damage table afaik. It might be that monks get 1.0 ATK per pt of STR. Im running a few tests with my pally as well, mainly to see if i can extract the effect of the Offense skill from the difference in the parses and skill caps with my BL. It'd be interesting to see if i can get all classes onto the same mastercurve. /hugs Elder Coprolith III Trollie ferrul lawd of 65 levels (retired) Merescata Arch Animist - 75 Beastlord Beastlord Posts: 199 Location: Oklahoma, US Logged Effects of ATK #5 January 04, 2004, 07:56:49 PM One thing I haven't checked - just to be certain - is to make sure that your relationship of 0.9 ATK per 1 strength holds up for other classes. I would be surprised if there was a difference between beastlords and paladins in that regard, but I wouldn't really be floored, either My lvl 60 paladin with all his gear on has 161 STR and 993 ATK. Stripped of his gear his STR=113 and ATK=950. So 43 ATK per 48 STR equates to 0.896. With roundoff error in the listed ATK i'd say the relation holds. It'd be interesting to see if i can get all classes onto the same mastercurve. /hugs Sounds like class balancing to me  :) Merescata 75 Vah Shir BST Tribunal Clan McEye Sylviania Guest Logged Effects of ATK #6 January 05, 2004, 04:29:38 PM wow great job... a few things to note : 1° be careful when stating ATK values for OH when using another skill than MH : the displayed ATK is valid only for MH ... valid for OH too only if Skill values are the same. I guess (I hope) Coprolith took this into account when he did his parses, but be sure noone forgot that when they do their own parses 2° You say the deviation from linear behaviour start to happen when max damage hits cap.. is is true ? it seems (hardly reading from the plots )max damage cap is in the 1250 - 1300 range, and diminishing n return in the 1400 - 1500 range. what did I miss ? I am not discussing the fact diminishing n return is linked to the fact max hit can't be beaten ( 'worst case' scenario : when all hits are for max hit, you will hardly increase n ...) Anyway, thanks ! those parses give more knowledge about PC damage output Coprolith Savage Lord Posts: 787 Location: Netherlands Logged Effects of ATK #7 January 05, 2004, 04:57:21 PM Tank yous, Sylviana. Replies to your post 1) Aye, i did. As you can see in the table, i've jotted down the ATK numbers for both MH and OH. I always used a HtH and a 1HB on purpose so i could study the exact effect of the lower skill cap. That's how i found out that by a simple rescaling of the ATK axis, you can obtain a 'master'-curve which is independent of skill. 2) It's tricky to see where the deviation from linearity occurs in data with statistical noise. Especially since the deviation is very gradual; when you're just above the cap the number of max hits is still low and its effect on average damage tiny. Check the graph i made in the second post of this thread. I added a dotted line as a guide to the eye in that one. The onset of the deviation of linearity is easier to see in that graph, it starts just below ATK-skill*1.33 = 1000 (i.e ~1300 ATK), but as i said, very gradual at first. /hugs Elder Coprolith III Trollie ferrul lawd of 65 levels (retired) Coprolith Savage Lord Posts: 787 Location: Netherlands Logged Effects of ATK #8 January 07, 2004, 10:21:56 PM Did some parsings, with my somewhat dusty L60 paladin and even dustier L49 ranger, mainly to see if i could get a better feel of how the Offense skill works. ==== edit: Duh, forgot to include the basic parse details ===== target: katta banker, rogue position 60PAL (225 weapon skill, 225 offense) Windblade (50/44 +37) 1) STR = 113, no buffs, no gear ATK = 950 swings = 10837 hitratio = 0.680 +/- 0.009 avg hit = 120.8 +/- 1.4 n = 1.676 +/- 0.020 2) STR = 203 (STR gear + GoK) ATK = 1030 swings = 7810 hitratio = 0.676 +/- 0.011 avg hit = 137.4 +/- 1.9 n = 2.008+/- 0.028 2) STR = 203 (STR gear + GoK+Wolf) ATK = 1114 swings = 7090 hitratio = 0.681 +/- 0.011 avg hit = 153.3 +/- 2.2 n = 2.325+/- 0.034 49RNG (200 weapon skill, 210 Offense) weapons: 1hb and 1hp 1) STR = 100, no buffs, no gear ATK = 850 swings = 14191 hitratio = 0.650 +/- 0.009 n = 1.275 +/- 0.013 2) STR = 196 (STR gear + GoK) ATK = 936 swings = 13564 hitratio = 0.642 +/- 0.008 n = 1.561 +/- 0.017 3) STR = 196 (STR gear + GoK + wolf) ATK = 1020 swings = 14000 hitratio = 0.649 +/- 0.008 n = 1.815 +/- 0.019 ============ end edit ================ First i tried to find some sort of relation between Weapons Skill (WES), Offense Skill (OFS), Strength (STR) and ATK This gives the following list: Class WES   OFS   STR   ATK   ATK* 60PAL 225   225   113   950   949 60PAL 225   225   203   1030   1030 65BST 225   252   115   987   987 65BST 225   252   280   1135   1135 65BST 250   252   115   1054   1053 65BST 250   252   280   1202   1202 49RNG 200   210   196   936   937 49RNG 200   210   100   850   851 ATK* is calculated as follows: ATK* = a*WES + b*OFS + c*(STR-d) We already know 2 of the constants, a=2*1.33 and c=0.9. As an initial guess i took b=1.33. I ran a quick check with a new lvl 1 BST, maxxed WES at the guild trainer and fought until i got some Offense skillups. You certainly dont get more then 2 ATK per point, so 1.33 seems as good a guess as any. That leaves the constant d, which can be considered a threshold strength below which you actually take a penalty to ATK (much like the 75 AGI threshold). Fitting d to the list i obtain d=56, which is a bit of a strange number for a threshold value. I checked the numbers against the lvl 1 BST. I then get the following numbers: Class WES   OFS   STR   ATK   ATK* 1BST 5   0   90   37   44 1BST 10   0   90   49   57 2BST 15   2   90   65   73 It's a bit off, but this could simply be the result of a roundof error in the number 2*1.33 i found for the WES factor. When fitting a, b and d to the entire set of data i get a=2.692, b=1.334 and d=65 and get a good match for the entire set. I'm gonna stick with a = 2*4/3 and b=4/3 for now, as it gives good results for the high level chars The next question is then how much of the listed ATK is 'to hit' rating and how much is 'damage rating'. We already know that WES splits evenly between to hit and damage rating. STR and straight ATK buffs are damage rating only. My initial assumption is that OFS also splits evenly between to hit and damage rating.  We then have: to hit rating = 4/3*WES + 2/3*OFS damage rating = ATK - 4/3*WES - 2/3*OFS If this is true then there should be a relation between the to hit rating and hit ratio, and between damage rating and average damage. There are 2 complications however. Firstly, hit ratio also depends on the difference between PC level and mob level (or perhaps the difference between PC WES and OFS and mob Defense skill). Secondly as you increase in levels you also go to higher damage tables. The improved damage tables at L51 and L60 for instance are very noticeable in-game. Let's have a look at hitratios first. Plotting hit ratio vs (to hit rating + level) yields very nice linear behaviour (but it doesn't eliminate the effect of mob level) Next is a plot of n vs damage rating I've omitted all the high ATK data from the 65BST so that the 60PAL and 49RNG data are clearer. The 60PAL data seems to fit quite nicely with the 65BST data. The damage of the 49RNG is well below the other 2 tho, the result of being in a damage table several notches lower. The results makes sense and the model of ATK describes the available data pretty well, but let me emphasize that all this proves nothing. There's just too many factors involved. If anyone has any suggestions on how to proceed i welcome them (best suggestion would be to tell SoE to finally give up on listing just one ATK number and split it into 2 and while they're at it do the same with AC please) /hugs Elder Coprolith III Trollie ferrul lawd of 65 levels (retired) DiosT Wildblood - 70 to 74 Beastlord Savage Lord Posts: 802 Location: Xegony Server Logged Effects of ATK #9 January 08, 2004, 05:07:14 AM As usual, Coprolith, you make my head hurt ;-) Coprolith Savage Lord Posts: 787 Location: Netherlands Logged Yet another update #10 January 14, 2004, 10:11:35 AM Sempai was very kind to send me a logfile of him fighting a red con (L70) mob. This mob does not fight back, and Sempai was attacking from behind. Weapons used were Ethereal Destroyer in MH and Shinai o/t Ancients OH. MH: ATK = 1420 swings = 8399 hits = 4886 hit ratio = 0.582+/-0.011 avg dmg = 63.64+/-0.91 n = 3.165+/-0.045 OH: ATK = 1487 swings = 8142 hits = 4927 hit ratio = 0.605+/-0.011 avg dmg = 74.99+/-0.91 n = 3.260+/-0.046 Pictures say more then a 1000 words so here they are. I'll start with the n vs ATK graph this time: There's not much of a difference in n (or avg damage) between the L70 red con and the L51 blue con. Let's move on to hit ratio instead: A very clear decrease in hit ratio is seen here. This reinforces the idea that mob avoidance (which affects your hit ratio) depends mostly on the mob's level whereas mob mitigation (which affects your avg dmg per hit) does not and can vary considerably. The next plot shows hit ratios as a function of mob level: . Pretty much what you'd expect, a steady decrease in hit ratio as mob level increases. The straight lines i added don't fit particularly good, and there's no reason why the relation between mob level and hit ratio should be linear. But with 3 data points it's pretty much the best you can do. At least it gives an approximation of the effect that I can use to enhance my dps calculation sheet. Thanks again to Sempai for sending me the data. If anyone else has high quality data like his and would like to share it with me i'd be indebted to you. Hopefully I'll be able to add some more mob levels one day to better nail the effect of mob level on dps down /hugs Elder Coprolith III Trollie ferrul lawd of 65 levels (retired) Kashmiir Battlekat WoW Fanboi Savage Lord Posts: 755 Location: World of Warcraft - Hyjal Server - Name: Khrash Logged Effects of ATK #11 February 18, 2004, 10:40:07 PM I think my intellegence just increased. Evandar Feral Posts: 58 Location: The Cold North (Sweden) Logged Effects of ATK #12 April 14, 2004, 06:01:57 PM You have become better at understanding Coprolith! (2)! Soo, well, I'm neither a math genious nor superb in the english language. Any chance you can just tell me when the effect of atk really starts to diminish? Like, say 1800ish? (Pure guess) ;) I know I read somewhere about where the AC values "caps" out in the diffrent expansions, is there anything like that for ATK? (Like, I think it was 900ish AC for old world, 1000ish for Kunark etc. Above that you got very little in return.. ) I mean, is there a big diffrence between, lets say,  2k ATK and 2400 ATK? Evandar - Antonius Bayle - Retired - EQ Dalle - Tauren Warrior - EU Stormrage - WoW Coprolith Savage Lord Posts: 787 Location: Netherlands Logged Effects of ATK #13 April 14, 2004, 06:43:49 PM Well Evandar, I've only parsed up to 1900 ATK so its hard to say if the effect of adding ATK really starts to diminish at ~1800 ATK. Extrapolation of data is the #1 sin in science  :D In the range i've parsed it seems to be a very gradual proces. Technically diminishing returns start at ~1300 ATK but you'd hardly notice it until you get to ~1600 ATK. And even above 1600 the average damage keeps increasing notably, albeit slower then before. At ~2000ATK i expect that the multiplier n is about 4. Now n cannot possibly get higher then 6, the cap for melee damage, but that still leaves room for improvement in the 2000 - 2400 ATK range. /hugs Elder Coprolith III Trollie ferrul lawd of 65 levels (retired) Otuol Arch Animist - 75 Beastlord Primalist Posts: 225 Logged Effects of ATK #14 April 15, 2004, 01:09:27 PM My head hurts now.  :lol: Anyhow, I know when my attack is around 1500 (with Fero) I notice more hits for closer to max.  When I get a ranger buff to bump it over 1600 some, I don't see a buge difference, but it is a noticeable difference. Print Go Up Pages1 2 User actions The Beastlords' Den Everquest 1 Library Effects of ATK User actions Print Help | Terms and Rules | Go Up ▲ SMF 2.1.4 © 2023, Simple Machines Hextech Theme Made By : TwitchisMental

[DiosT]

Continued from above
- Hit ratio depends on offensive skills only, ATK buffs have no effect on it
- Confirmation that the dualwield rate at lvl 65 is higher then previously assumed
- In the ATK range between 1060 and 1470, the multiplier for the base damage, n, depends linearly on ATK. A master curve can be constructed to account for differences in weapon skills.
- A change in ATK gives a corresponding change in n equal to (0.00338+/-0.00006) times the change in ATK
- With these results its possible to calculate the melee dps of any weapon in a particular situation once you've got a parse of a single weapon in that situation
- There's no truth to the statement that you need high ATK for 2HB to be effective. Dualwielding will generally benefit more from high ATK then 2HB

More On Attack:

Posted by Coprolith,

I got Frenzy of Spirit last night and parsed out the result of 15 FoS sessions against the Katta banker. HtH ATK was 1911 and 1HB ATK was 1844 with FoS (1575/1508 w/o). Statistics arent very good (only 1133 MH swings and 845 OH) but any longer and my brain would have dribbled out of my ears.

Anyways, here's the updated graph:



The return you get from ATK keeps diminishing at high ATK (as expected) but its still a very significant increase.

/hug

[DiosT]

Posted by Coprolith,

Okies, here's what i've got so far

- For my baseline i use test #9 from the original series: no gear except for haste belt and weapons
STR=115
MH - CotIO, 1HB, ATK = 987
min hit = 14, max hit = 71
#hits = 14628
Avg dam = 37.54+/-0.31
n = 1.753+/-0.014
OH - WLK, HtH, ATK = 1052
min hit = 2, max hit = 69
#hits = 15191
Avg dam = 30.66+/-0.25
n = 1.916+/-0.016

I did 2 4h parses against the Katta banker today:

- STR gear only, buffs: Infusion of Spirit, Furious Strength, Frenzy
STR = 280
MH - CotIO, 1HB, ATK = 1135
min hit = 14, max hit = 84
#hits = 7293
Avg dam = 46.14+/-0.54
n = 2.367+/-0.028
OH - WLK, HtH, ATK = 1202
min hit = 2, max hit = 85
#hits = 7370
Avg dam = 39.36+/-0.46
n = 2.460+/-0.029

- No STR gear, Veng2, Spiritual Vigor, Form of the Great Wolf
STR = 115
MH - CotIO, 1HB, ATK = 1138
min hit = 14, max hit = 84
#hits = 7656
Avg dam = 46.06+/-0.53
n = 2.362+/-0.027
OH - WLK, HtH, ATK = 1205
min hit = 2, max hit = 85
#hits = 7745
Avg dam = 39.82+/-0.45
n = 2.489+/-0.028

I had intended to perform 2 more overnight tests at intermediate values, one with only STR gear and one with only Vengeance items, but after seeing the results from the first 2 parses that exercise seems moot. Combining these results with my previous analysis of max hits and avg dmg the end result is unambiguous:
There is no difference whatsoever between ATK gained from STR and from pure ATK; both max hit and avg dmg are exactly the same.

The new data points fit exactly onto the linear behaviour i found before:

It would certainly be pleasing if they changed the system so that ATK from strength was treated no differently than "pure" ATK - it was always annoying and cumbersome to have to consider the two differently.

I couldn't agree more. It makes more sense this way. There is now only two parts of the ATK-rating to consider: the 'to hit'-rating, determined by skill only, and a 'damage'-rating, determined by skill, STR and pure ATK
Weapon skill goes 50/50 towards 'to hit' and 'damage'. I cannot say what the exact effect of the Offense skill is, but i assume its similar to weapon skill. STR and pure ATK have the exact same effect if the increase in the listed ATK is the same

If it turns out that a +50 strength buff is close to a +50 ATK buff (for those who aren't capped), that may change - I know that uber players/raiders reach the stat caps fairly easily, even unbuffed, but for casual players it takes a LONG time before that is feasible (I know that I have under 200 unbuffed strength - partly because I've believed it to be viable for ignoring).

To be precise:
- 1 point of STR adds 0.9 points to listed ATK
- 1 point of pure ATK adds 1.33 points to the listed ATK (melee class bonus)
So a Vengeance 1 item is worth only 7.4 points of STR, provided your STR isn't maxxed out. For a L65 beastlord, maxxing out STR is easy to do even for the non-ubers by just using our own buffs. For non-uber paladins, warriors, rogues and monks who have very little in the way of melee buffs and have to keep up other stats as well (e.g. AC and WIS for the paladin) I imagine this changes the way (s)he looks at gear selection more drastically then a BL

/hug

[Romidar]

Thank you for the additional data and analysis - I actually thought that sort of parsing was fun 4 years ago but find it too tiresome now.

To be precise:
- 1 point of STR adds 0.9 points to listed ATK
- 1 point of pure ATK adds 1.33 points to the listed ATK (melee class bonus)
So a Vengeance 1 item is worth only 7.4 points of STR, provided your STR isn't maxxed out. For a L65 beastlord, maxxing out STR is easy to do even for the non-ubers by just using our own buffs. For non-uber paladins, warriors, rogues and monks who have very little in the way of melee buffs and have to keep up other stats as well (e.g. AC and WIS for the paladin) I imagine this changes the way (s)he looks at gear selection more drastically then a BL

To your point, checking my Magelo, I'm at 183 strength, including stat food. I include my Magelo just as indication of what the pretty much highest end paladin equipment is for a non-raider. Given my normal group, which includes a shaman, I still wouldn't bother with increasing the strength on my gear; however, it'll now at least be a factor in evaluatiing equipment whereas it truly was not previously.

Your findings also make the Planar Power ability a bit more appealing than it had been, in that the ~22 increase in ATK will actually be meaningful.

One thing I haven't checked - just to be certain - is to make sure that your relationship of 0.9 ATK per 1 strength holds up for other classes. I would be surprised if there was a difference between beastlords and paladins in that regard, but I wouldn't really be floored, either.

[Coprolith]

One thing I haven't checked - just to be certain - is to make sure that your relationship of 0.9 ATK per 1 strength holds up for other classes. I would be surprised if there was a difference between beastlords and paladins in that regard, but I wouldn't really be floored, either

Well that's easily checked. My lvl 60 paladin with all his gear on has 161 STR and 993 ATK. Stripped of his gear his STR=113 and ATK=950.
So 43 ATK per 48 STR equates to 0.896. With roundoff error in the listed ATK i'd say the relation holds.
It may be that classes that are on a higher damage table get more ATK per STR. BLs and paladins are on the same damage table afaik. It might be that monks get 1.0 ATK per pt of STR.
Im running a few tests with my pally as well, mainly to see if i can extract the effect of the Offense skill from the difference in the parses and skill caps with my BL. It'd be interesting to see if i can get all classes onto the same mastercurve.

/hugs

[Merescata]

One thing I haven't checked - just to be certain - is to make sure that your relationship of 0.9 ATK per 1 strength holds up for other classes. I would be surprised if there was a difference between beastlords and paladins in that regard, but I wouldn't really be floored, either

My lvl 60 paladin with all his gear on has 161 STR and 993 ATK. Stripped of his gear his STR=113 and ATK=950. So 43 ATK per 48 STR equates to 0.896. With roundoff error in the listed ATK i'd say the relation holds.
It'd be interesting to see if i can get all classes onto the same mastercurve.

/hugs

Sounds like class balancing to me  :)

[Sylviania]

wow great job... a few things to note :
1° be careful when stating ATK values for OH when using another skill than MH : the displayed ATK is valid only for MH ... valid for OH too only if Skill values are the same. I guess (I hope) Coprolith took this into account when he did his parses, but be sure noone forgot that when they do their own parses
2° You say the deviation from linear behaviour start to happen when max damage hits cap.. is is true ? it seems (hardly reading from the plots )max damage cap is in the 1250 - 1300 range, and diminishing n return in the 1400 - 1500 range. what did I miss ? I am not discussing the fact diminishing n return is linked to the fact max hit can't be beaten ( 'worst case' scenario : when all hits are for max hit, you will hardly increase n ...)


Anyway, thanks ! those parses give more knowledge about PC damage output

[Coprolith]

Tank yous, Sylviana.

Replies to your post

1) Aye, i did. As you can see in the table, i've jotted down the ATK numbers for both MH and OH. I always used a HtH and a 1HB on purpose so i could study the exact effect of the lower skill cap. That's how i found out that by a simple rescaling of the ATK axis, you can obtain a 'master'-curve which is independent of skill.

2) It's tricky to see where the deviation from linearity occurs in data with statistical noise. Especially since the deviation is very gradual; when you're just above the cap the number of max hits is still low and its effect on average damage tiny. Check the graph i made in the second post of this thread. I added a dotted line as a guide to the eye in that one. The onset of the deviation of linearity is easier to see in that graph, it starts just below ATK-skill*1.33 = 1000 (i.e ~1300 ATK), but as i said, very gradual at first.

/hugs

[Coprolith]

Did some parsings, with my somewhat dusty L60 paladin and even dustier L49 ranger, mainly to see if i could get a better feel of how the Offense skill works.

==== edit: Duh, forgot to include the basic parse details =====

target: katta banker, rogue position

60PAL (225 weapon skill, 225 offense)
Windblade (50/44 +37)
1) STR = 113, no buffs, no gear
ATK = 950
swings = 10837
hitratio = 0.680 +/- 0.009
avg hit = 120.8 +/- 1.4
n = 1.676 +/- 0.020
2) STR = 203 (STR gear + GoK)
ATK = 1030
swings = 7810
hitratio = 0.676 +/- 0.011
avg hit = 137.4 +/- 1.9
n = 2.008+/- 0.028
2) STR = 203 (STR gear + GoK+Wolf)
ATK = 1114
swings = 7090
hitratio = 0.681 +/- 0.011
avg hit = 153.3 +/- 2.2
n = 2.325+/- 0.034

49RNG (200 weapon skill, 210 Offense)
weapons: 1hb and 1hp
1) STR = 100, no buffs, no gear
ATK = 850
swings = 14191
hitratio = 0.650 +/- 0.009
n = 1.275 +/- 0.013
2) STR = 196 (STR gear + GoK)
ATK = 936
swings = 13564
hitratio = 0.642 +/- 0.008
n = 1.561 +/- 0.017
3) STR = 196 (STR gear + GoK + wolf)
ATK = 1020
swings = 14000
hitratio = 0.649 +/- 0.008
n = 1.815 +/- 0.019

============ end edit ================



First i tried to find some sort of relation between Weapons Skill (WES), Offense Skill (OFS), Strength (STR) and ATK

This gives the following list:
Class WES   OFS   STR   ATK   ATK*
60PAL 225   225   113   950   949
60PAL 225   225   203   1030   1030
65BST 225   252   115   987   987
65BST 225   252   280   1135   1135
65BST 250   252   115   1054   1053
65BST 250   252   280   1202   1202
49RNG 200   210   196   936   937
49RNG 200   210   100   850   851

ATK* is calculated as follows:
ATK* = a*WES + b*OFS + c*(STR-d)
We already know 2 of the constants, a=2*1.33 and c=0.9. As an initial guess i took b=1.33. I ran a quick check with a new lvl 1 BST, maxxed WES at the guild trainer and fought until i got some Offense skillups. You certainly dont get more then 2 ATK per point, so 1.33 seems as good a guess as any. That leaves the constant d, which can be considered a threshold strength below which you actually take a penalty to ATK (much like the 75 AGI threshold). Fitting d to the list i obtain d=56, which is a bit of a strange number for a threshold value.
I checked the numbers against the lvl 1 BST. I then get the following numbers:

Class WES   OFS   STR   ATK   ATK*
1BST 5   0   90   37   44
1BST 10   0   90   49   57
2BST 15   2   90   65   73

It's a bit off, but this could simply be the result of a roundof error in the number 2*1.33 i found for the WES factor. When fitting a, b and d to the entire set of data i get a=2.692, b=1.334 and d=65 and get a good match for the entire set. I'm gonna stick with a = 2*4/3 and b=4/3 for now, as it gives good results for the high level chars


The next question is then how much of the listed ATK is 'to hit' rating and how much is 'damage rating'.
We already know that WES splits evenly between to hit and damage rating. STR and straight ATK buffs are damage rating only. My initial assumption is that OFS also splits evenly between to hit and damage rating.  We then have:

to hit rating = 4/3*WES + 2/3*OFS
damage rating = ATK - 4/3*WES - 2/3*OFS

If this is true then there should be a relation between the to hit rating and hit ratio, and between damage rating and average damage.
There are 2 complications however. Firstly, hit ratio also depends on the difference between PC level and mob level (or perhaps the difference between PC WES and OFS and mob Defense skill).
Secondly as you increase in levels you also go to higher damage tables. The improved damage tables at L51 and L60 for instance are very noticeable in-game.

Let's have a look at hitratios first. Plotting hit ratio vs (to hit rating + level) yields very nice linear behaviour (but it doesn't eliminate the effect of mob level)



Next is a plot of n vs damage rating



I've omitted all the high ATK data from the 65BST so that the 60PAL and 49RNG data are clearer. The 60PAL data seems to fit quite nicely with the 65BST data. The damage of the 49RNG is well below the other 2 tho, the result of being in a damage table several notches lower.

The results makes sense and the model of ATK describes the available data pretty well, but let me emphasize that all this proves nothing. There's just too many factors involved. If anyone has any suggestions on how to proceed i welcome them (best suggestion would be to tell SoE to finally give up on listing just one ATK number and split it into 2 and while they're at it do the same with AC please)

/hugs

[DiosT]

As usual, Coprolith, you make my head hurt ;-)

[Coprolith]

Sempai was very kind to send me a logfile of him fighting a red con (L70) mob. This mob does not fight back, and Sempai was attacking from behind. Weapons used were Ethereal Destroyer in MH and Shinai o/t Ancients OH.

MH:
ATK = 1420
swings = 8399
hits = 4886
hit ratio = 0.582+/-0.011
avg dmg = 63.64+/-0.91
n = 3.165+/-0.045

OH:
ATK = 1487
swings = 8142
hits = 4927
hit ratio = 0.605+/-0.011
avg dmg = 74.99+/-0.91
n = 3.260+/-0.046

Pictures say more then a 1000 words so here they are. I'll start with the n vs ATK graph this time:



There's not much of a difference in n (or avg damage) between the L70 red con and the L51 blue con. Let's move on to hit ratio instead:



A very clear decrease in hit ratio is seen here. This reinforces the idea that mob avoidance (which affects your hit ratio) depends mostly on the mob's level whereas mob mitigation (which affects your avg dmg per hit) does not and can vary considerably.

The next plot shows hit ratios as a function of mob level:

.

Pretty much what you'd expect, a steady decrease in hit ratio as mob level increases. The straight lines i added don't fit particularly good, and there's no reason why the relation between mob level and hit ratio should be linear. But with 3 data points it's pretty much the best you can do. At least it gives an approximation of the effect that I can use to enhance my dps calculation sheet.

Thanks again to Sempai for sending me the data. If anyone else has high quality data like his and would like to share it with me i'd be indebted to you. Hopefully I'll be able to add some more mob levels one day to better nail the effect of mob level on dps down

/hugs

[Kashmiir Battlekat]

I think my intellegence just increased.

[Evandar]

You have become better at understanding Coprolith! (2)!

Soo, well, I'm neither a math genious nor superb in the english language. Any chance you can just tell me when the effect of atk really starts to diminish? Like, say 1800ish? (Pure guess) ;)

I know I read somewhere about where the AC values "caps" out in the diffrent expansions, is there anything like that for ATK? (Like, I think it was 900ish AC for old world, 1000ish for Kunark etc. Above that you got very little in return.. )

I mean, is there a big diffrence between, lets say,  2k ATK and 2400 ATK?

[Coprolith]

Well Evandar, I've only parsed up to 1900 ATK so its hard to say if the effect of adding ATK really starts to diminish at ~1800 ATK. Extrapolation of data is the #1 sin in science  :D
In the range i've parsed it seems to be a very gradual proces. Technically diminishing returns start at ~1300 ATK but you'd hardly notice it until you get to ~1600 ATK. And even above 1600 the average damage keeps increasing notably, albeit slower then before. At ~2000ATK i expect that the multiplier n is about 4. Now n cannot possibly get higher then 6, the cap for melee damage, but that still leaves room for improvement in the 2000 - 2400 ATK range.

/hugs

[Otuol]

My head hurts now.  :lol:

Anyhow, I know when my attack is around 1500 (with Fero) I notice more hits for closer to max.  When I get a ranger buff to bump it over 1600 some, I don't see a buge difference, but it is a noticeable difference.