[Rogue/Warrior] Weapon Skill Adjustment Discussion - Page 9

Karmon · 07/27/07, 7:36 AM

Originally Posted by songster

<can o'worms>We're absolutely sure that your hit chance varies linearly with hit rating, yes?</can o'worms>

Uh oh! He used that word nonlinear

No we can't but:
with 355 skill from 1 point in WEx we need 316 hitrating to never miss
with 360 skill form 2 points in WEx we need 308 hitrating to never miss.

8 hitrating is 0.507% and the 5 skill difference translates to 0.5%.
If there is a nonlinearity it is pretty near linear for those two cases

Olgas · 07/27/07, 10:45 AM

Originally Posted by Karmon

I am not convinced that you really can sum the misses for autoattacks and specials. You only can
add the samples of identicaly statistical processes. Specials and autoattacks may have a different
expected hitrate.

Also comparing melee and ranged may not be proper.

I tend to agree about combining apecials and auto-attacks. And that is why I first listed my pet's auto-attacks separately. Without including the specials, we still have a fairly large sample size (2402 regular melee attacks!) with a realized miss rate of 3.997%. Adjusting for the 4% from Animal Handler, this, by itself, still really suggests a base miss rate of very close to 8% exactly.

Also, when you consider the specials and the auto attacks separately, you still see a very similar result:

Autos: 2402 attacks, 96 misses, 3.997% miss
Specials: 1451 attack, 63 misses, 4.342% miss

_______________________________

As for comparing, melee and ranged, here I don't agree. You are of course correct that it "may not be proper." However, I tend to believe it is. A Hunter's ranged damage is physical damage just like a Rogue's or Warrior's. And Blizzard already has two fairly complex combat systems -- one for physical damage and one for magical damage. I doubt that they'd create yet another whole system just for ranged physical damage, which after all, would be a system that really only applied to one class. It's possible, yes, but improbable in my view.

Karmon · 07/27/07, 11:00 AM

Originally Posted by Olgas

[..]
Without including the specials, we still have a fairly large sample size (2402 regular melee attacks!) with a realized miss rate of 3.997%. Adjusting for the 4% from Animal Handler, this, by itself, still really suggests a base miss rate of very close to 8% exactly.
[..]

Well using the data from your pet the 5% confidence intervall is [3.213% 4.78%]. Adjusting with the 4%
from Animal Handler we get [7.213%, 8.78%].
So based on that sample we cannot differentiate between 8% and 8.6%.

Punscho · 07/27/07, 11:08 AM

I agree to a certain point, but then again they seperate melee and ranged hit talents and so on so they might handle melee and ranged miss rates differently (the rogue +hit talent doesn't apply to ranged attacks afaik, probably not the warrior one either).

Also I don't think pets necessarily have the same mechanics as players, although knowing the pet's miss rate would be of interest aswell.

The best thing is probably to wait for more data from someone running without extra skill. I feel that it would be nice to really be sure and settle on the hit aspect.

Then there's the supposed reduced avoidance from extra skill, which interests me the most. First we have to conclude the base parry/dodge/block ratings and my data points to parry being twice as likely to happen as dodge or block. I have a couple of thousand swings recorded but we havn't raided in a month so I don't know when I'll have enough to draw any conclusions.

Read my previous request: http://elitistjerks.com/412424-post117.html

Of course other people than the tanks can /combatlog if they know the tank's weapon skill at all times (racial, or using weapon X all the time).

EDIT: another thing struck me. It might be interesting for a hunter to tame a pet 3 levels below the servants in blasted land (they are 50-54 or so) and let the pet hit the mob in the face for a while. We don't know if pets got the same hit/miss mechanics but the avoidance ratings would be the same as a player, wouldn't it? Using the level 70 mend pet it wouldn't be hard to go on for quite a while I suppose? Just don't stun them and don't /combatlog before the "banish" effect is gone.

Olgas · 07/27/07, 11:15 AM

Originally Posted by Karmon

Well using the data from your pet the 5% confidence intervall is [3.213% 4.78%]. Adjusting with the 4%
from Animal Handler we get [7.213%, 8.78%].
So based on that sample we cannot differentiate between 8% and 8.6%.

Ok, got it. Good point.

Olgas · 07/27/07, 11:26 AM

Originally Posted by Gruntle

I think it's pretty reasonable to assume that the mechanics for weapon skill/defense is the same for both DW and 2h/1h+shield. The only difference in the miss chance between these two should be the DW penalty of 19%. If (yeah, I'm ofc not sure) this is true the results from the 2-paladin testing (earlier in this thread) could also be used to narrow this down.

Looking at the paladin with 350 weapon skill he had 7.971% misses (with a std of 0.291% not sure how to correctly compete confidence limits), this is (asfaik there is no way to turn it off) with the Draenei +1% tohit aura, but without any other +hit equip (I think?). This means that the 2h miss rate vs +3 boss mobs is 8.97 (0.291) % from that study. This again points to a miss chance being even higher than the 0.2 per skill difference (for a difference of 15 at least).

If we assume that +19% is all that happens when going to DW, we get a DW miss rate of 27.97 (0.291) % for a 15 skill difference.

edit: my computed confidence limits were crap, removing them.

Gruntle,

Unless I misunderstood those Pally tests, I'm pretty sure they were done against Venoxis -- a level 63 mob in ZG. So, unfortunately, I don't think we can draw any real conclusions from them with respect to the effect of Weapon Skill against a Level 73.

• sp00n · 07/27/07, 11:31 AM

Originally Posted by Olgas

Gruntle,

Unless I misunderstood those Pally tests, I'm pretty sure they were done against Venoxis -- a level 63 mob in ZG. So, unfortunately, I don't think we can draw any real conclusions from them with respect to the effect of Weapon Skill against a Level 73.

Well, the whole reason they started this test is that boss mobs (those with the skull) are always regarded as being 3 levels above you.
In terms of weapon skill and therefore miss/hit/etc, they are regarded level 63 back before BC, and now level 73 when you engage them with level 70.

Olgas · 07/27/07, 11:32 AM

Originally Posted by Punscho

EDIT: another thing struck me. It might be interesting for a hunter to tame a pet 3 levels below the servants in blasted land (they are 50-54 or so) and let the pet hit the mob in the face for a while. We don't know if pets got the same hit/miss mechanics but the avoidance ratings would be the same as a player, wouldn't it? Using the level 70 mend pet it wouldn't be hard to go on for quite a while I suppose? Just don't stun them and don't /combatlog before the "banish" effect is gone.

Hmm...

Problem with this is that there is no way to know exactly what your pet's Attack Skill is. I mean, you can know what your Pet's character Level is. But the interface does not reveal what the pet's Weapon Skill level is, and so my hunch is that it improves "invisibly" through 5 skill levels before the pet dings to the next level.

I've had my pet at level 70 for so long, I can be very confident that it has a 350 Weapon Skill, but with a newly tamed pet, you'd have no real way of knowing...

Gruntle · 07/27/07, 11:51 AM

Originally Posted by sp00n

Well, the whole reason they started this test is that boss mobs (those with the skull) are always regarded as being 3 levels above you.
In terms of weapon skill and therefore miss/hit/etc, they are regarded level 63 back before BC, and now level 73 when you engage them with level 70.

Yes, that's my understanding of boss mobs as well. If you look at their results you see that they have quite a lot of misses and not a lot of crits, it's certainly clear that it's not lvl 70 players vs 63 mobs at least. I'm pretty sure that the mobs are counted as 73, and I think the stats they have (8682 swings with 0 weapon skill) is a great test to use when figuring this out.

Going out on a limb here, it might be possible to combine the data sets from the paladin (with 0 weapon skill, but remember the draenei aura), Karmans test without weapon expertise (but correct for the 19% DW penalty and hit rating) and Olga's hunter tests (correcting for hit rating) to find the base miss chance vs level 73 mobs (i.e. 8% or 8.6% or whatever it is). If we assume that hit rating is understood, that the DW penalty really is just 19%, that ranged and melee are treated in the same way and that classes work the same this should be ok. I would perhaps leave the pet out for now though.

• sp00n · 07/27/07, 11:54 AM

So the fellow testers here (great work!) have found out that you need 316 hit rating for 5 weapon skill to (presumably) never miss, and around 361 for 0 weapon skill.

Let me draw a small chart. The hit ratings have been calcualted with 10*82/52, which gives a more accurate values than 15.769 (so it's 307.5 instead of 307.4955, which is much "nicer" in my opinion).
It's a comparison from weapon skill to defense skill. 0 +Def means you are fighting an equal mob with 0 weapon skill, +5 is a level 73 mob with 10 points weapon skill, +10 with 5 points of weapon skill, and +15 without any points of weapon skill (so a "true" 3 level difference).

Def  Hit      Miss rate         Difference
+ 0  299.61   24%		+ 0%
+ 5  307.5    24.5%		+ 0.5%
+10  315.38   25%		+ 1%
+15  360-361  27.83 - 27.89%	+ 3.83% - 3.89%

+15  362.69   28%		+ 4%

I've included the last one because I admit I don't like odd numbers. 4% is just much nicer than 3.83%, and the difference is basically only 1.69 hit rating than the reported results (which is nothing).

Assuming this, the miss chance is indeed only depandant on the weapon skill to defense ratio, and has nothing to do with level difference.

I have to admit, this seems more and more appealing to me.

// Edit
Following this road, we would face the following situation:

Def   Hit      Miss    Total    Increase
+ 0   299.61   24%       0      0
+ 1   301.19   24.1%   + 0.1%   0.1%
+ 2   302.77   24.2%   + 0.2%   0.1%
+ 3   304.35   24.3%   + 0.3%   0.1%
+ 4   305.92   24.4%   + 0.4%   0.1%
+ 5   307.5    24.5%   + 0.5%   0.1%
+ 6   309.08   24.6%   + 0.6%   0.1%
+ 7   310.65   24.7%   + 0.7%   0.1%
+ 8   312.23   24.8%   + 0.8%   0.1%
+ 9   313.81   24.9%   + 0.9%   0.1%
+10   315.38   25%     + 1%     0.1%
+11   324.85   25.6%   + 1.6%   0.6%
+12   334.31   26.2%   + 2.2%   0.6%
+13   343.77   26.8%   + 2.8%   0.6%
+14   353.23   27.4%   + 3.4%   0.6%
+15   362.69   28%     + 4%     0.6%

So basically it is 0.1% per point of weapon skill if you are 1-10 skill points below the target, and 0.6% for range of 15-10.
Also this means that the first point in WEx yields 3% tohit, and the second only 0.5% (against a level 73 mob).

Trazhenko · 07/27/07, 12:51 PM

Ok, so up to 5 skill (1 pt in weap/ex or ~20 rating), weapon skill is awesome. After that, it's not quite as awesome, but probably still good. I think it's time to turn our efforts toward +crit, -dodge, and -parry (there appeared to be 0 -parry in the paladin testing, bummer).

Dralmoo · 07/27/07, 1:06 PM

Originally Posted by sp00n

So basically it is 0.1% per point of weapon skill if you are 1-10 skill points below the target, and 0.6% for range of 15-10.
Also this means that the first point in WEx yields 3% tohit, and the second only 0.5% (against a level 73 mob).

This really resembles the spell hit rates for casters, which scale very non-linearly as mob level increases, from 6% miss against a +2 mob to 17% miss against a +3 mob.

It therefore makes quite a bit of sense that effectively making the mob +2 instead of +3 (the effect of 5 points of weapon skill) would have a much larger effect than making the mob +1 instead of +2.

• Aldriana · 07/27/07, 1:13 PM

So, next question... how does this apply to dodge rating? From the pally test we can say that there doesn't seem to be much (if any) effect on block/parry, but it would still be interesting to know what happens with dodge.

Trazhenko · 07/27/07, 2:21 PM

Yes, based on +.1% hit and +.04% crit (we know we get at least that) for each point of skill, skill rating is pretty a pretty weak stat after you get 5 skill.

At that point, 1 skill rating is ~.025 hit + ~.01 crit (at least) + some unknown amount of -dodge.

With 1 hit rating worth ~.063 hit, and 1 crit rating (which is weak for rogues already) worth ~.045 crit, skill rating looks terrible.

However, if you are hit capped, you can lose a few hit rating and also eat into dodge with skill rating items, so how much dodge you can eliminate will determine if having much more than 5 skill is worth it.

• sp00n · 07/27/07, 7:18 PM

I went through the postings in this thread, gathering information on dodges, parries and crits and weapon skill.
Unfortunately I don't have the motivation to analyze the numbers now, so bear with me.

But it looks like that won't be easy, especially dodge varies quite a lot, and crit is highly dependent on buffs and procs, which makes it very unreliable to test in a raid environment. Parry on the other hand can only be really measured when attacking from the front, which leaves us with a) tanks or b) mobs outside raid instances (or c) such crazy things as running to Venoxis

).

Karmon #1
Bossmobs
355 Skill
2175 attacks
6.161% dodges

Karmon #2
Bossmobs
350 Skill
1607 attacks
5.974% dodges

Karmon #3
Bossmobs
350 Skill
964 attacks
7.469% dodges

Nalisol #1
Against Level 62
360 Skill
1580 attacks
1 dodge (0.06%)

Nalisol #2
Against Level 60
350 Skill
1138 Attacks
0 dodge

Punscho
At 45 skill more I did experience parries and dodges but at 50 skill plus I didn't notice anything for some 3000 swings

XP-Dolphin
On PTR against Zul'Gurub Bossmob
Bossmob
Attacking from front

350 Skill (Beaglej) [Character Screen]
8682 attacks
4.826% dodges
11.702% parries
2.258% crits [5.25% crit chance]

365 Skill (Therasiv) [Character Screen]
9975 attacks
3.238% dodges
11.509% parries
3.278% crits [5.97% crit chance]

koaschten
Bossmobs
369 Skill
2153 attacks
3,44% doges

Gogge
Against Level 54

350 Skill
10002 attacks
8.378% Crits

360 Skill
9954 attacks
8.75%

Karmon · 07/27/07, 7:27 PM

Originally Posted by songster

Surely it's vanishingly unlikely that Blizzard would put an exception in their entire code for melee hit rates, based on one talent in one tree of one class! It would set them up for hell when itemising weaponskill and balancing classes/specs.

Seems you are right.

I got hold of total +27 daggerskill items.
Under the wowwiki theory you should not miss against bossmobs with 313 hitrating:
(24 + (365 -(350+27*52/205))*0.1 - 5)*205/13 = 312.469

So against The Curator I did not miss once. Albeit 530 total hits is a bit too few for my taste:
15/41/5, 0 Wex, +27 daggerrating, 313 hitrating
129 crits, 247 hits, 0 blocks, 130 glances, 0 misses, 22 dodges, 2 parries
530 total attacks
181493 damage, 342.439623 damage per swing
72076 hit damage, 291.805668 hit damage per swing
31680 glance damage, 243.692308 glance damage per swing
24.340% crits
46.604% hits
0.000% blocks
24.528% glancings
0.000% misses
4.151% dodges
0.377% parries

Olgas · 07/27/07, 8:53 PM

Originally Posted by sp00n

Well, the whole reason they started this test is that boss mobs (those with the skull) are always regarded as being 3 levels above you.
In terms of weapon skill and therefore miss/hit/etc, they are regarded level 63 back before BC, and now level 73 when you engage them with level 70.

Oh, yeah -- duh! -- I don't know what I was thinking -- ignore my note. Sorry, Gruntle.

• Vulajin · 07/27/07, 9:19 PM

As this thread continues on, I'd just like to salute you guys for continuing to unravel the mysteries of this game. It's amazing how much we can still learn about this game. Keep up the great work.

Karmon · 07/28/07, 6:36 AM

Originally Posted by sp00n

[..]
Following this road, we would face the following situation:

Def   Hit      Miss    Total    Increase
+ 0   299.61   24%       0      0
+ 1   301.19   24.1%   + 0.1%   0.1%
+ 2   302.77   24.2%   + 0.2%   0.1%
+ 3   304.35   24.3%   + 0.3%   0.1%
+ 4   305.92   24.4%   + 0.4%   0.1%
+ 5   307.5    24.5%   + 0.5%   0.1%
+ 6   309.08   24.6%   + 0.6%   0.1%
+ 7   310.65   24.7%   + 0.7%   0.1%
+ 8   312.23   24.8%   + 0.8%   0.1%
+ 9   313.81   24.9%   + 0.9%   0.1%
+10   315.38   25%     + 1%     0.1%
+11   324.85   25.6%   + 1.6%   0.6%
+12   334.31   26.2%   + 2.2%   0.6%
+13   343.77   26.8%   + 2.8%   0.6%
+14   353.23   27.4%   + 3.4%   0.6%
+15   362.69   28%     + 4%     0.6%

So basically it is 0.1% per point of weapon skill if you are 1-10 skill points below the target, and 0.6% for range of 15-10.
Also this means that the first point in WEx yields 3% tohit, and the second only 0.5% (against a level 73 mob).

Almost:
24+(365-skill)*0.1 if (365-skill) <=10
24-5+(365-skill)*0.6 if (365-skill) >=10

• sp00n · 07/28/07, 7:15 AM

Jesus I'm stupid.
I wanted to go through my logs to find out my dodge rates and I had a large data set of combatlogs wearing the [Shoulderpads of the Stranger] and a fist/dagger weapon combo. On Wednesday though I switched to sword/sword, which do not profit from the +skill to dagger.

Anyway, I got all the dodge % from the combatlogs until I realized that I switched to sword/sword and dumped the old ones, leaving me only with these.

Well, I'm currently looking at 4502 total attacks with 245 dodges, which is 5.44%, and *if* I remember correctly, I was at around 4.9% while wearing the dagger off hand (with only the off hand having around 2.53 more skill -> ~362.53). Unfortunately I deleted all the old combat log files.
Never try to be smart late in the evening/early in the morning.

Olgas · 07/28/07, 9:43 AM

Originally Posted by sp00n

Def   Hit      Miss    Total    Increase
+ 0   299.61   24%       0      0
+ 1   301.19   24.1%   + 0.1%   0.1%
+ 2   302.77   24.2%   + 0.2%   0.1%
+ 3   304.35   24.3%   + 0.3%   0.1%
+ 4   305.92   24.4%   + 0.4%   0.1%
+ 5   307.5    24.5%   + 0.5%   0.1%
+ 6   309.08   24.6%   + 0.6%   0.1%
+ 7   310.65   24.7%   + 0.7%   0.1%
+ 8   312.23   24.8%   + 0.8%   0.1%
+ 9   313.81   24.9%   + 0.9%   0.1%
+10   315.38   25%     + 1%     0.1%
+11   324.85   25.6%   + 1.6%   0.6%
+12   334.31   26.2%   + 2.2%   0.6%
+13   343.77   26.8%   + 2.8%   0.6%
+14   353.23   27.4%   + 3.4%   0.6%
+15   362.69   28%     + 4%     0.6%

So basically it is 0.1% per point of weapon skill if you are 1-10 skill points below the target, and 0.6% for range of 15-10.
Also this means that the first point in WEx yields 3% tohit, and the second only 0.5% (against a level 73 mob).

Hmmm... couple points here.

First, I'm not yet convinced we have determined the definitive miss rate for a Level 73.

It certainly appears from the data that the miss rate is higher than 27%. But I'm not yet prepared to jump all the way to 28%, at least not yet. So far, the balance of the data suggests the miss rate is somewhere between 27 and 28. I do concede that the tests run by Sarlunas and the two pallies suggest it is much closer to 28 than 27, but I'd like to see more tests with +hit at 360 or 361, and ideally at 363, before agreeing to a 28% figure.

Right now my working hypothesis is 27.6%, with the formula as follows:

Base miss rate is:

24% + (Defense Skill - Weapon Skill)*0.10 [if (Def Skill - Wep Skill) <=10]
24% + (Defense Skill - Weapon Skill)*0.24 [if (Def Skill - Wep Skill) > 10]

The .24 factor in the second line of the foregoing would be bumped higher if, after a little more testing, the actual base miss rate of a Level 73 proves out to be higher. Perhaps it is .25, .26 or .even .27, instead of .24.

Second, the way you presented the table above would not accurately reflect the discontinuity implied by this formula. It should, I believe, read more like this.

Def    Miss        Total    Increase
+ 0    24%          0         0
+ 1    24.1%     + 0.1%      0.1%
+ 2    24.2%     + 0.2%      0.1%
+ 3    24.3%     + 0.3%      0.1%
+ 4    24.4%     + 0.4%      0.1%
+ 5    24.5%     + 0.5%      0.1%
+ 6    24.6%     + 0.6%      0.1%
+ 7    24.7%     + 0.7%      0.1%
+ 8    24.8%     + 0.8%      0.1%
+ 9    24.9%     + 0.9%      0.1%
+10    25%       + 1%        0.1%
+11    26.64%    + 2.64%     1.64%   <------
+12    26.88%    + 2.88      0.24%
+13    27.12%    + 3.12%     0.24%
+14    27.36%    + 3.36%     0.24%
+15    27.6%     + 3.6%      0.24%

Notice the discontinuity as you move from +10 Def to +11 Def. (Which makes sense because that is the point when you move from fighting a Level 72 mob to fighting a Level 73 mob). Please also note how valuable this formula makes the FIFTH point in Weapon Skill. Essentially, as you add +Weapon Skill you reduce your chance to miss (more regularly described as gaining +hit) against a Level 73 mob as follows:

+1 Weap Skill =    +0.24% hit
+2 Weap Skill =    +0.24% hit (0.48% total)
+3 Weap Skill =    +0.24% hit (0.72% total)
+4 Weap Skill =    +0.24% hit (0.96% total)
+5 Weap Skill =    +1.64% hit (2.6% total)  <--------
+6 Weap Skill =    +0.10% hit (2.7% total)
+7 Weap Skill =    +0.10% hit (2.8% total)
...
+15 Weap Skill =   +0.10% hit (3.6% total)

• sp00n · 07/28/07, 9:50 AM

Originally Posted by Olgas

Right now my working hypothesis is 27.6%, with the formula as follows:

Base miss rate is:

24% + (Defense Skill - Weapon Skill)*0.10 [if (Def Skill - Wep Skill) <=10]
24% + (Defense Skill - Weapon Skill)*0.24 [if (Def Skill - Wep Skill) > 10]

I think we can rule out a 27.6% miss chance due to these observations:

Originally Posted by Sarlunas

I also decided to raid without WEx today for some testing. We wiped a couple times on Al'ar (don't ask) and each time I swapped in more hit rating since I was still getting misses. Started at 347, then 356 and finally got 2 misses in phase 2 with 360 hit rating.

Wow Web Stats

I swapped DFT for WSC to get 361 and killed VR and Solarian without a single miss so if there's still a miss chance at 27.89% +hit then it's pretty darn small.

Wow Web Stats
Wow Web Stats

Raiding without my DST is painful btw. :P

360 hit rating converts to 360/(10*82/52)= 22.83 + 5 = 27.83% to hit.
If the miss chance was 27.6%, 357 hit rating would have been enough to eliminate any miss.

I don't think these 2 misses came out of nowhere.

// Edit
To test how much exactly 1 point weapon skill gives, we would need to do tests with a skill lower than 5.
Mace specialization comes to my mind, since it allows nice jumps of 2, whereas weapon skill rating has an ugly modifier of 2.5*82/52= ~3.94.

So with 2 points in mace specialization, giving 4 additional weapon skill, under the 28% miss and 0.6% increase theory, you would need 28-5-(4*0.6)= 20.6% tohit to never miss, which is 20.6*(10*82/52)= 324.85 = 325 hit rating.
For the 27.6% and 0.24% increase the same would mean 27.6-5-(4*0.24)= 21.64 = 341.25 = 342 hit rating.

So we would need a rogue with 325 hit rating and 4 points additional weapon skill (2 points in mace specialization) beat some level 73 mobs. If he encounters a miss, the 28% 0.6% theory is disproven, but on the other side the 27.6% 0.24% theory is not proven to be correct either yet.

Btw, I *really* think the 27.6% have been disproven by the 360 hit rating above. Which of course doesn't mean that the 28% theory is the correct one, just the most appealing to me currently.

Olgas · 07/28/07, 10:29 AM

Originally Posted by sp00n

Btw, I *really* think the 27.6% have been disproven by the 360 hit rating above. Which of course doesn't mean that the 28% theory is the correct one, just the most appealing to me currently.

This may be correct.

But right now we only have one data point from a single person as evidence. I'd just like to see this test repeated by other observers before we reach this conclusion. (Please take no offense, Sarlunas.)

Olgas · 07/28/07, 11:07 AM

Originally Posted by sp00n

So we would need a rogue with 325 hit rating and 4 points additional weapon skill (2 points in mace specialization) beat some level 73 mobs. If he encounters a miss, the 28% 0.6% theory is disproven, but on the other side the 27.6% 0.24% theory is not proven to be correct either yet.

Let's not confound two issues -- or worse yet, try to test two different issues at the same time.

The first issue is miss rate against a Level 73. Let's see if we can get more data on that, so we can establish 27.6 or 28 (or whatever!) with a higher degree of confidence.

The second issue is how each point of Weapon skill incrementally affects miss rate. And this will certainly require additional testing. But I am confused by the formula that you and Karmon are proposing:

24-5+(365-skill)*0.6 if (365-skill) >10

Why did you come up with the formulation this way? Where did the -5 come from?

The working theory (pretty undisputed at this point) is that the base miss rate against an evenly leveled and skilled mob is 24%. Why are you reducing it to 19%, and then adding back .6% per Weapon Skill to get back up to 28%. That doesn't make sense to me.

It seems like the -5 number is just a plug number to re-establish linearity between +Def 11 and +Def 15. But why would there need to be linearity here. A Level 72 normally has +10 Def. And so anything higher than +10 Def must be a Level 73 mob. And this explains why there would be big discontinuous jump between +10 Def and +11 Def -- I don't see any need to smooth out that impact, the way your formula does.

Also, think about what your formula would mean for non dual-wielders. It would translate to:

5-5+(365-skill)*0.6 if (365-skill) >10

5-5? That couldn't be right.

There is a foundational base miss rate against all mobs of 5%. This foundational base is increased by 19% when dual-wielding. The Defense and Weapon Skill can work to adjust from this foundational base, but not to ignore it. Your proposal is that, when the difference between Defense and Weapon Skill is > 10, you ignore the foundational base and adjust solely based upon the Skill difference, times .6%. Sorry, I don't buy that.

Of course, we have to test it. But I'm very doubtful.

Dontmindme · 07/28/07, 11:50 AM

Originally Posted by Karmon

Well using the data from your pet the 5% confidence intervall is [3.213% 4.78%]. Adjusting with the 4%
from Animal Handler we get [7.213%, 8.78%].
So based on that sample we cannot differentiate between 8% and 8.6%.

One thing I think should be noted here, 95% confidence doesn't mean you can't differentiate. It means, that statistically speaking 95% of your results will fall within this range or 19 out of 20. Looking at how close 8.6% to the 8.78% limit, what are we talking? Something like 85-90% confidence that 8.6% can be ruled out? Let's not put too much value in this magical marker of 95% where data if thrown out randomly, and certainly it doesn't mean data showing a significant bias in one direction, like this set, should be completely ignored.

To me, when I look at data, anything within 1 standard deviation is fair game, meaning, it's likely indistinguishable. Slightly over 2/3 of data should fall within one standard deviation. After 1 standard deviation, that looks like bias and after 2, you start looking at whether the theory is wrong (or whether something is wrong with the method).

Let's not just dismiss this data as it still shows a strong trend away from 8.6%.