tyvm!

Ok I couldn't sleep and nothing else was going on so I spent another hour in front of a L60 test dummy testing weapon enchant proc rates and I came to the same conclusion I did before. The sim is still doing it wrong, less wrong, but still wrong. This time I equiped my old OH with mongoose on it to seperate out the MH and OH weapon enchant procs.

I was a big dummy and forgot to start logging and reset recount, which for whatever reason failed to record the fight so I don't have as much detail as before but I do plan on redoing this test later this week on PTR to take advantage of the ISS mana regen to eliminate my OOM time so I'll be sure to grab both of those then. I did remember to clear both Proculas and Procodile before hand however and neither of the previous two omissions change the reason I think the sim is still doing it wrong. The sim is assuming a static 1.2ppm unaffected by haste, however I observed on this 1hr run 2.78ppm on berserking (MH) and 2.12ppm on mongoose (OH). This gave a 48% uptime on berserking and 42% on mongoose. The sim shows the expected uptimes to be 42% and 36% respectively, pretty much 6% lower on both accounts.

From what I've seen there's no hard evidence to support the 1.2ppm, no haste effect beyond somebody in some other thread said so. The other thread had zero testing done in game with varying degrees of haste from what I read. I've done one test with minimal haste showing ~1.1-1.2ppm rate there and numerous tests with varying degrees of haste all showing markedly increased proc rates. Assuming the static 1.2ppm rate devalues both haste and enchant contribution. If I'm wrong and just had 5 or 6 ridiculously lucky streaks in a row then by all means show me the money, err data. If I'm right and it is a 1ppm affected by haste then lets get the simulator updated. Lets face it, if close enough were good enough we'd be using spreadsheets and approximators like everybody else.

And on another note for the mana simulation part, JoW contribution is horribly low as well, the 3.0.8 version returns considerably more than 9 MRPS. Seems to be more on the line of 40-70 MPRS or 200-350 MP5 on static fights.

For example on a Raz fight I had 183 JoW procs @ 87-88 mana each over 3:31 with I believe 809 eligible attacks (assuming FT can proc it) giving either a ~50ppm proc rate or, more likely, ~20% chance per hit.

I have a better idea. How about you show us some data and not just 3 paragraphs of run-on sentences of your stream of consciousness. You're the one asserting something is wrong, the burden of proof is on you and not anyone else. And no, screenshots of your procodile/recount windows don't count as data.

May I ask what you honestly think showing raw logs or WWS reports will show different than recount and mod reports? Assuming recount is accurate and the proc monitor mods are working as advertised then that IS data, just a summary report that says the same thing that a 20MB log file would.

There's none of what you're asking for to support 1.2ppm, so why is that taken as gospel?

That said expect a log parse later this week, but I'd still like to know why the screenshots don't count in this situation when they seem to in others. If photoshopping is the worry well logs can be manipulated just the same, that's why I've been asking for somebody else to give it a try.

For one thing someone walks in here and says "lul my mod says 4.8 ppm" we have no idea how long he tested and how many procs were recorded to make it statistically relevant. There's a lot of difference between a 5min test, an hour test, and a 5 hour test.

Dragon, I think you are right. It's what I posted back when I was checking things for the Weapon Enchant thread. I think 1ppm affected by haste is how the mechanic works, but I don't have any proof of that. I have an opinion but everyone has one of those.

You are spouting that your opinion is the right one and we better get off our asses and fix the Sim because it's inaccurate because it doesn't work how you think it actually does. For that to happen we need a bit more then an opinion, we need verifiable fact. That's where the combatlogs and time spent testing and conditions and all that start to come into play.

Everybody got opinions, but if the facts are on your side you should be able to prove that your opinion is the right one easily enough and then we will be more then happy to support your request to modify the Sim (notice the support aspect, in the end the Sim is Tukez's so he is the one you need to convince).

It does look like there may be something wrong with the current model, but it is really hard to say from your data. You have a ton of variables changing, your base case is difficult to work with, and there probably isn't enough data.

Without knowing how many hits & procs you had, we can't calculate the statistical certainty levels in any statements about these results. The fact that you had a 30% higher OH ppm on your autoattack test suggests that an hour isn't even close to long enough to get statistically significant results, but without the actual data I can't possibly say for certain.

What would be really nice to see, and would help immensely in deriving the actual mechanic as opposed to just showing that the current mechanic is wrong, is a base case test with no haste, no flurry, no imbue, only MH weapon equipped, and a screenshot or log indicating (a) the total number of hits and (b) the total number of procs.

You can then increase haste using WF totem and gear and record the same data.

At that point, there will be enough data to make an indisputable statistical statement about the results.


After that we can worry about how all our abilities affect things, first we need to prove or disprove the fundamental mechanic.

To give some idea of how long you need to test for to get meaningful results (and to help show why people tend to be so dismissive when someone shows up with small sets of test data):

Let's assume a 2.5 speed weapon. If the real PPM is 1.0, each swing has a 4.2% chance to proc. If the real PPM is 1.2, each swing has a 5% chance to proc.

Given some observed data (X swings proc'd, Y swings didn't), we can use statistics to determine a range for the actual chance to proc (e.g. 4% to 6%) and our confidence in that range (e.g. we're 90% certain it's somewhere in that range).

The formulas I'm using are from Checking if a coin is fair - Wikipedia, the free encyclopedia.

Let's say we want to find the PPM value within plus or minus 0.1 PPM, with a certainty of 99%. 0.1PPM corresponds to about a 0.4% change in the chance to proc per swing on our 2.5 speed weapon.

The number of events needed to achieve this level of certainty is given by n = p(1-p) * Z^2/E^2. E is the 0.004 (0.4%) change in proc chance. The Z-value for 99% certainty is 2.576. p is the 'real' proc chance, which we'll take to be 0.04 (just to be nice; closer to 0.5 gives a higher stddev).
Thus, the number of swings you'd need is over 15000, which is about 10 hours of swinging that 2.5 speed weapon.

So, if you record 15000 swings with no talents or haste gear and you observe a PPM of 1.0, you can say with 99% certainty that the 'true' PPM value is in the range 0.9-1.1, which would indicate that the currently accepted value is wrong.

Someone correct me if my math is wrong, but that is the gist of it as I understand it.



It does somewhat suck that the burden of proof is on you when the EJ TTT articles never seem to source their data. This wouldn't be the first time that an accepted theory was really just based on insufficient anecdotal evidence.

I was thinking should I make WF cd a normal distributed random number? Mean 3.0s with standard deviation 0.Xs?

Could people help me with Spirit Wolf mechanics? Just unload all the info regarding them here, obvious ones too. I am not very familiar how pets work. I also need to know the buffs, which could affect wolfs. Thanks for the help.

The macro to bring up the pet window is:
They gain 30% of your AP.

Mousing over their character sheet using the pet window, they clearly gain the benefits of:
WF totem
SoE totem
Heroism

It seems likely that they are affected by any and all buffs that other (e.g. warlock) pets are affected by - I've heard of people getting warriors to battleshout after they pop them.

It appears that they gain the benefit of any AP buff you gain even after you've popped them (after a slight delay). For example, my wolves started at 1777 AP, then after dropping untalented SoE they immediately went to 2087 AP (i.e. they gained the SoE buff), then after a short delay they went up to 2183 (30% of my benefit from the SoE buff).


The hard part will likely be getting accurate base crit and miss rates for them.

I was also thinking Tukez that we will possibly need support for more than one set bonus. It is quite likely that people will soon be running round with naxx_melee_2 and ulduar_melee_2 simultaneously. Would this be a case of coding tier7_8_melee_2_2 and tier7_8_nuker_2_2 or some such variable name? It's possibly the simplest way of tackling it. Or would you do two variables set bonus 1, set bonus 2 and just have each take same values as valid inputs?

I'm looking at how to get the Rawr.Enhance model to deal with set bonuses, so guidance would be appreciated. Having two separate lines of set bonus would possibly be easier for end users, and definitely easier for Rawr.Enhance coding if that makes a difference

Whilst I'm posting :

Support for glyph4 can be dropped, as could some of the really old items. Also any chance the default config could be more or less the BiS config file it seems to be a Enh talent, Ele rotation hybrid at present. Thus causing new users issues with setting rotation priorities correctly.

Any progress on trinkets, relic etc taking itemid's rather than text string identifiers?

Wolves also reportedly gain ~30-35% of player Hit Rating Tukez.

From ShockAndAwe Export.lua

if enchantID == "2673" then
enchantText ="mongoose"
elseif enchantID == "3225" then
enchantText ="executioner"
elseif enchantID == "1900" then
enchantText ="crusader"
elseif enchantID == "3273" then
enchantText ="deathfrost"
elseif enchantID == "3789" then
enchantText ="berserking"

Probably the best way to model the Windfury cooldown (as I understand it to work) would be to just set it to 3 +/- 0.3 seconds. I'm 99% sure there is a latency component to the spread factor but that it would be capped at some value that would be completely server based. If you wanted to be as accurate as possible 3 +/- (0.2 + Latency)*

* = capped at 0.5 seconds

That would fit with all the data I've seen although the issue is that the spread value works as a bellcurve model where the extremes are less likely to be encountered which is why the simplest is probably the best way to go (3 +/- 0.3 seconds).


As for the Feral Spirits;

Base Stats
785 AP
331 Str
113 Agi
402 Stam
65 Int
109 Spirit
9778 Armor

Their attacks each do 331-456 damage at 1.5speed (262.2 dps)

They gain 30% of the shaman's AP and Stamina and 35% of his Armor.

There is a rumor that they may utilize the shaman's hit rating (but not expertise or crit rating) but that needs to be confirmed (I'll try to work on it later).

They receive all raid buffs if they are activated after the wolves are summoned. So when affected by Unleashed Rage they will gain 30% of their own AP in addition to whatever transfers from you.

In my normal gear my Wolves have 1759 AP. When the totem only affects me they have 1869. If they are buffed with SoE and I'm not they have 2115 AP. If we are both of us are buffed they have 2225. So it would seem they gain either 2AP per Strength or 1AP per Str and 1AP per Agi. Gonna get some Kibbler's bits and check on that part.

edit
Here are 3 short runs on the Boss Dummy. All are 2 summons so the sample size is a bit small to draw conclusions but the numbers are reasonable.

No gear
Wow Web Stats

Geared with 438 +hit
Wow Web Stats

Geared with 271 +hit (8.26% melee hit)
Wow Web Stats

For some reason on the last test they attacked from the front so they got parried a lot but even with just 271 +hit there wasn't a single missed attack. Their crit rate was the same 1-2% for each test and they got dodged/parried the same whether I was dodge capped or not. Opening the pet tab (macro is on the TTT) shows that they gain no benefit from haste. But from my really small sample it appears that having enough hit rating to melee cap (without dual wield penalty) removed all misses from their table.

So that would mean they get
30% of your AP/Stam
35% of your Armor
All your hit rating
None of your other stats/ratings.

Spirit wolwes hit scaling: My theory is that hit scaling was 30% but hotfixed to 100%.

Last week 10man satharion wipes.
http://wowwebstats.com/mhmvfngaoqof1...023d1&s=bosses
Wow Web Stats
This week 25man satharion.
Wow Web Stats

One miss is propably when puppies attacked satharion when immunity was up.

Well that makes perfect sense if the assumption is that they gain 100% of your hit because the wolves are not dual wielding. They're treated just like a hunter pet or a druid's attacks, which is the same hit table as a 1H or 2H weapon.

Statistically that is wrong. Even if they use the same mechanism, they are two separate random variables. You can not combine those two variables together to obtain a higher confidence level.

And personally I would say that 80% confidence intervals are too low, even 90 could be too low (99% is too high though). Generally 95% is what is used in most studies.

Statistically speaking, how is doing two 1hr runs with one weapon any different than doing one 1hr run with two weapons that use the same mechanics? The point of statistics is combining multiple variables we know are related and looking at them overall to try to determine how they're related.

One weapon for 2 hours yielding 65 procs hour one and 55 procs hour two "combines" for an average of 1ppm.

Two weapons for 1 hour yielding 120 total procs for 2ppm / 2 = 1ppm per weapon. Do we know exactly which weapon proced which? No, and we don't care so long as the mechanics behind how they proc are consistent.

We're not studying an individual weapon or combination we're studying the underlying mechanic. Dual wielding achieves the same results in half the time since you can do two tests in parallel.

I run EnhSim with 4 threads to finish in 1/4 of the time and statistically the results are the same if I had done one thread for 4x the time. If Draenei had prehensile tails I'd be doing 3 weapons at once.

I noted in http://elitistjerks.com/1093112-post75.html that glancing rate is now assumed to be 24%. I know its configurable in the sim but perhaps time to change the default in config.txt?

I also noted the following 3.1 changes :

* Armor Penetration Rating: All classes now receive 25% more benefit from Armor Penetration Rating.
* Haste Rating: Shamans, Paladins, Druids, and Death Knights now receive 30% more melee haste from Haste Rating.

We also now get +25% expertise from expertise rating.

So whereas the old cap was 214 expertise rating with 3/3 UR the new cap is 17 expertise * (32.78998947/4)/1.25 = 112 expertise rating. So we gain 102 expertise from these changes. That's a huge amount.

Careful with that expertise change

http://elitistjerks.com/1137280-post376.html

Looks like someone may have coded the change wrong and affected expertise when they actually wanted to just change haste rating.


Edit:
Tukez can you add Frostbrand as an option in the Sim. Just use current values (442 base damage per proc with 10% spellpower coefficient) with an 8 PPM mechanic that is affected by haste (proc chance determined by base weapon speed).

That PPM value may change as more testing comes in but 8 seems like a decent compromise position for now.

Edit2:
Make it 8.5 ppm or roughly an average of a proc every 7 seconds (I wonder if that's how they code PPM, not by procs but by average time between procs since that would work out to 8.57 ppm which fits exactly with the current long running test).

Why am I getting past cap warnings for Expertise? If I'm reading this right the sim thinks I'm over by 15.79 expertise rating?
As far as I've been able to figure reading these forums, expertise rating caps at 214, but I'm only at 197. I'm using Snapper Extreme food for Hit, not a expertise food. What am i missing? this can get a bit complex as we get close to the caps is seems...



#############

Mh expertise + EP range goes past cap by 1.9262 expertise (15.79 expertise rating).
Oh expertise + EP range goes past cap by 1.9262 expertise (15.79 expertise rating).

Calculating EP values...
<snip>

EP value DPS total DPS difference

baseline 2971.96
ap 1.00 0.37 2995.95 23.99
crit rating 2.00 0.75 3006.49 34.53
hit rating 2.99 1.12 3005.56 33.60
expertise rating 1.30 0.49 2987.88 15.92
haste rating 1.46 0.55 2990.04 18.08
armor penetration rating 1.24 0.46 2978.90 6.94
spellpower 0.97 0.36 2982.11 10.15
mh dps 5.82 2.18 2993.78 21.82
oh dps 2.80 1.05 2982.44 10.48
strength 1.10
agility 1.84
intelligence 1.42
spirit 0.01 0.00 2972.28 0.32
mana 0.00 0.00 2972.09 0.13
mp5 0.18 0.07 2973.28 1.32

elapsed simulation time: 140000.00h
elapsed real time: 350.45s
simulation speed: 1438138x



######### my config file.


simulation_time 10000
simulation_time_combatlog 30
combat_length 6
report_count 80
threads 2
min_lag 40
max_lag 60
simulate_mana 1

ep_precision 2
ep_base_stat ap
ep_ap 64
ep_crit_rating 46
ep_hit_rating 30
ep_expertise 4
ep_haste_rating 33
ep_armor_penetration_rating 15
ep_spellpower 28
ep_dps 10.0
ep_mana 200
ep_spirit 100
ep_mp5 20

mh_auto_attack 1
oh_auto_attack 1
wait_ss_with_wf_cd 0
cast_ll_only_if_wf_on_cd 0
bloodlust_casters 2
sync_bloodlust_with_trinkets 0
cast_lvb_only_if_ed_left 15
cast_lvb_only_if_fsdots_left 4
cast_lvb_only_if_fs_active 1
cast_fs_only_if_dots_left 0
cast_ls_only_if_charges_left 0
cast_sr_only_if_mana_left 700
use_mana_potion_if_mana_left 3000

rotation_priority_count 7
rotation_priority1 SR
rotation_priority2 MW5_LB
rotation_priority3 SS
rotation_priority4 FS
rotation_priority5 ES
rotation_priority6 LL
rotation_priority7 LS

miss 8.00
dodge 6.50
glancing 25.00
armor 13083
spell_miss 17.00
nature_resistance 0
fire_resistance 0
frost_resistance 0
arcane_resistance 0
shadow_resistance 0

armor_debuff_major 0/3925
armor_debuff_minor 0/1260
physical_vulnerability_debuff 0/2.0
melee_haste_buff 20.0/20.0
melee_crit_chance_buff 0/5.0
attack_power_buff_flat 0/688
attack_power_buff_multiplier 0/99.7
spell_haste_buff 0/5.0
spell_crit_chance_buff 0/5.0
spell_crit_chance_debuff 0/10.0
spell_damage_debuff 0/10.0
spellpower_buff 280/280
spell_hit_chance_debuff 0/3.0
haste_buff 0/3.0
percentage_damage_increase 0/3.0
crit_chance_debuff 0/3.0
stat_multiplier 10.0/10.0
stat_add_buff 52/52
agi_and_strength_buff 178/178
intellect_buff 60/60

replenishment 1
water_shield 0
mana_spring_totem 1
blessing_of_wisdom 0
judgement_of_wisdom 0

flask_elixir flask_of_endless_rage
guardian_elixir -
potion -
food snapper_extreme
misc_item -

###############################################################################
### Everything in the section below can be replaced by information obtained ###
### from your paper doll stats or exported by the ShockAndAwe addon ###
###############################################################################

###############################
## ShockAndAwe EnhSim Export ##
###############################
race tauren
mh_speed 2.50
oh_speed 2.60
mh_dps 156.6
oh_dps 130.2
mh_crit 25.92
oh_crit 25.92
mh_hit 11.22
oh_hit 11.22
mh_expertise_rating 197.00
oh_expertise_rating 197.00
ap 3107
haste 3.39
armor_penetration 5.39
str 190
agi 721
int 503
spi 187
spellpower 991
spell_crit 19.56
spell_hit 14.03
max_mana 11661
mp5 0
mh_imbue windfury
oh_imbue flametongue

mh_enchant mongoose
mh_weapon -
oh_enchant mongoose
oh_weapon axe

trinket1 grim_toll
trinket2 mirror_of_truth

totem stonebreakers_totem
set_bonus - ## not currently available in the addon export
metagem relentless_earthsiege_diamond

glyph_major1 windfury_weapon
glyph_major2 stormstrike
glyph_major3 lava_lash
glyph_major4 - ## slot empty

glyph_minor1 - ## no useful glyphs current implemented in the sim
glyph_minor2 - ## no useful glyphs current implemented in the sim
glyph_minor3 - ## no useful glyphs current implemented in the sim


#############
## Talents ##
#############
ancestral_knowledge 2/5
improved_shields 0/3
mental_dexterity 3/3
shamanistic_focus 1/1
flurry 5/5
elemental_weapons 3/3
unleashed_rage 5/5
weapon_mastery 3/3
dual_wield_specialization 3/3
mental_quickness 3/3
improved_stormstrike 2/2
static_shock 3/3
maelstrom_weapon 5/5
convection 0/5
concussion 5/5
call_of_flame 2/3
elemental_devastation 3/3
reverberation 0/5
elemental_focus 1/1
elemental_fury 5/5
call_of_thunder 0/1
unrelenting_storm 0/3
elemental_precision 0/3
lightning_mastery 0/5
elemental_oath 0/2
lightning_overload 0/5
lava_flows 0/3
storm_earth_and_fire 0/3
shamanism 0/5

Tukez perhaps you should change the warnings from the sim, because we're having to answer this question twice a day it seems. If the person is over the cap because of their stats, then they should be told that their gear exceeds the cap. If the EP Step Size will put them over the cap during simulation then the message should be worded differently.