Elitist Jerks Hunter Spreadsheets -- In development

 02/11/08, 10:08 AM #1441 Seven Glass Joe     Lagortil Night Elf Druid   Sen'jin (EU) I'm sorry to disturb your legendary discussion. Bug Report: Spreadsheed v.42 seems to switch 2 gems in my bow stitched leggins when it collects the data automaticaly from the armory. Because of this the socket bonus appears as "not achieved" which actually is not true. It switches the gems in the blue and yellow socket. (10AP 4Hit vs. (AP 2MP5). Thanks for your great work!
02/11/08, 10:38 AM   #1442
Elimbras
Don Flamenco

Dwarf Priest

Eitrigg (EU)
 Originally Posted by Cheeky I compute both simulatenously, as I did before. I just use slightly different equations. The order of the rolls is independant, just whether or not a single roll is used is important.
Actually, one can imagine different 2 rolls systems.
The one you use is most likely the one Blizzard use, but here is a "Crits dont miss" 2-roll system, with different results :
- roll for crit or normal.
- Crits always hit. If normal, roll for hit or miss.

If crit and hit are your crits and hits "paperdoll corrected" values (meaning that they are the character sheet ones, corrected by defense/ level /etc, except the roll system), then this systems leads to Crit chance of critical strikes, Hit*(1-Crit) chances of hit, and (1-Hit)*(1-Crit) chances of miss.

The 2-roll system you use leads to Hit*Crit chances of critical, Hit*(1-Crit) chances of hit and (1-Hit) chances of miss.

The old 1-roll system just gave the Crit, Hit-Crit and (1-Hit) probability.

It is cleary weird, but it is possible to have a different 2-roll system, and I've not always been able to understand blizzard logic (why is the amno dps replaced by a fixed bonus in steady-shot ???).

The good point is that with a little more datas like the one Lactose provided to move to 2-roll system, it will be quite easy to find which system is accurate, since Hit and Crit "paperdoll" probability are far away from 0 and 1.

02/11/08, 1:15 PM   #1443
awstick
Glass Joe

Orc Hunter

Zul'Jin
 Originally Posted by Thorongil So basically it´s a matter of whether relative or absolute values are the ones that are of interest for you.
It seems like relative gains should be the only ones of interest. Obviously against bosses with lower armor, the raw DPS gain for any amount of armor pen higher than they are on bosses with high armor. But I don't think that matters at all, because the percentage increase would be the same with 1000 armor pen whether you are facing a boss with a net 1000 armor, or one with 9999 armor. If you have enough armor pen to increase your damage by 10% it doesn't really matter whether you are doing 1000 DPS to a higher armor boss, or 2000 DPS to a low armor boss, it's still just 10%. Against the low armor bosses any increase in stats would produce a higher DPS increase than they would against high armor targets.

So to me it seems like every point of armor pen should have the same value, no matter how much or how little you have it, unless of course you get targets down to zero armor.

02/11/08, 2:33 PM   #1444
Thorongil
Piston Honda

Human Priest

Alexstrasza (EU)
 Originally Posted by awstick It seems like relative gains should be the only ones of interest. Obviously against bosses with lower armor, the raw DPS gain for any amount of armor pen higher than they are on bosses with high armor. But I don't think that matters at all, because the percentage increase would be the same with 1000 armor pen whether you are facing a boss with a net 1000 armor, or one with 9999 armor. If you have enough armor pen to increase your damage by 10% it doesn't really matter whether you are doing 1000 DPS to a higher armor boss, or 2000 DPS to a low armor boss, it's still just 10%. Against the low armor bosses any increase in stats would produce a higher DPS increase than they would against high armor targets. So to me it seems like every point of armor pen should have the same value, no matter how much or how little you have it, unless of course you get targets down to zero armor.
Basically this seems convenient but keep in mind that diminishing returns on armor do exist, they are simply not as high as one would assume on first sight. Hence there is also a bigger benefit from more ArP/low Armor bosses. For exact calculation concerning this please consult the above linked Warrior-Thread (though, the calculation found there naturally uses a tank-PoV).

However, there is also the case that AP or Crit scales to another proportion at given amounts of armor than ArP therefore making not only the relative- but also the absolute-value interesting. For this Cheekys Spreadsheet provides quite useful answers.

 02/11/08, 4:06 PM #1445 Cheeky Great Tiger   Cheeky Troll Hunter   No WoW Account Preview With all the new gear in 2.4 I'm thinking of releasing a new version tonight. Only problem is I'm still not 100% satisfied with my model of the 3:2 (Steady Spam) rotation yet, so I'm going to leave that out. Here is what is done: - Change Band of the Eternal Champion proc duration to 10 seconds. - Corrected some bugs in the profiles matrix. Everything should display accurately now. - Corrected Draenei base Intelligence to 78 (from 82.) - Corrected some item stats (Embrace of the Twisting Nether, Bow-Stitched Leggings) - Fixed a bug in Shoulder socket bonuses that was sometimes showing bonuses for gear that did not have sockets. - Corrected some item name spellings (Ancient Draenei War Talisman, Wicked Flame Spessarite) - Remodeled Mark of Conquest with a 17 second internal cooldown and 13% chance to proc. This is based on data collected by Xeno of Suramar. - If you select Thor'idal, the Stars' Fury as a ranged weapon the ammo is defaulted to None. (This is not applicable for OO versions.) - Added new gear (Thor'idal, the Stars' Fury, Blackened Naruu Sliver, Apolyon, the Soul-render, Fang of Kalecgos, Muramasa, Crus of the Apocalypse, Shiv of Exsanguination, Hard Khorium Choker, Shoulderpads of the Silvermoon Retainer, Breeching Comet, Edge of Opression, Sunrage Shoulderpads, Hauberk of the Warbringer, Bladed Chaos Tunic, Leggings of the Immortal Night, Shoulderpads of Vehemence, Duplicitous Guise, Truestrike Crossbow, Bombardier's Blade, The Sunbreaker, Shattered Sun Pendant of Might, Gronnstalker's Belt, Gronnstalker's Bracers, Gronnstalker's Boots, Fletcher's Gloves of the Phoenix, Embrace of the Phoenix). - Provide for three Buff Profiles. These are Loaded/Saved on the 'Buffs' tab, the names are controlled on 'Settings and Results' and they are Imported/Exported properly. Does anyone else have any quick suggestions?
 02/11/08, 4:47 PM #1446 Lybohske Glass Joe   Lybohske Tauren Hunter   Khadgar I don't think this is a quick suggestion, but I have been wondering if you could add the Rift Stalker 4 piece bonus to the modeled items that proc off of a crit (Tsunami Talisman and Hourglass of the Unraveller)?
02/11/08, 5:28 PM   #1447
Cheeky
Great Tiger

Cheeky
Troll Hunter

No WoW Account
 Originally Posted by Lybohske I don't think this is a quick suggestion, but I have been wondering if you could add the Rift Stalker 4 piece bonus to the modeled items that proc off of a crit (Tsunami Talisman and Hourglass of the Unraveller)?
We can add that to the bug list. It'll take a bit more work than what I plan to spend before packaging this up tonight. I'll have to track different crit rates per shot. (In reality I'd be surprised if it made more than a 2 AP difference.)

 02/11/08, 6:04 PM #1448 Rokh Piston Honda   Rokh Orc Hunter   Emerald Dream Cheeky, could we get some remedial "3:2 support" included? All you'd have to do is remove latency from affecting Steady Shot, and the reduced GCD's you talked about before in this post. We can build our own rotations from that. The only time latency affects steady shot anymore is when using the cast sequence macro, pretty much, and so many people have switched over that I think it's worth taking out now.
 02/11/08, 6:13 PM #1449 Vaxum Von Kaiser   Møbius Human Death Knight   Spirestone Here are a few more items to include in the 2.4 gear list: Shadowed Gauntlets of Paroxysm, Sunrage Shoulderpads, Glyph of the Gladiator, Carapace of Sun and Shadow, Hard Khorium Band, Mounting Vengeance, Starstalker Legguards, Surestrike Goggles 3.0, Tunic of the Ranger Lord The shoulder enchants have a typo - Inscription of Vengence/Greater Inscription of Vengence should be spelled Vengeance. Last edited by Vaxum : 02/11/08 at 8:32 PM. Reason: added Tunic of the Ranger Lord
02/11/08, 9:22 PM   #1450
Stinkerbelle
Von Kaiser

Spayed
Draenei Hunter

Stormscale
 Originally Posted by Midnight Five of the eight test series produced an almost perfect 3:2 rotation
I must be misunderstanding something about what we mean by a 3:2 rotation. I thought it meant an alternating pattern approximating

Hence 3 steadies per 2 autos.

Looking at Midnight's test summaries, it looks like all but 1 of the quivered tests resulted in a 2:1 rotation (2 steadies per auto), with the exception of test 8, which comes pretty close to a 3:2 ratio.

 Originally Posted by Midnight Test #8 - 2,6s Weapon, 15% quiverhaste, 5/5 SS talent - 1,88s actual attack speed. Number of Auto Shots fired: 405 Number of Steay Shots fired: 594 Total number of shots fired: 999
I realize that maintaining 3 steadies per 2 autos is not necessarily the goal of these tests, but rather to puzzle out optimal rotation/speed/haste combinations given the new Auto Shot mechanics. Just confirming what you mean by 3:2.

 02/11/08, 9:48 PM #1451 • sordee Priest for Hire     Sordee Tauren Priest   No WoW Account Huh, if they leave the damage as such and not change anything with the mechanics of Steady Shot, doesn't it make the Legendary Bow a little OP, such that it won't be replaced until way,way into lvl80 Raiding? I.e Three benefits so far: -Free Arrows, -Free Bag slot, and - Rough increase of 200dps over Bristleback/Timeless? (And still not factoring in the other stats) This would almost prompt me to believe the first balancing change for WotLK would be for changing Steady Shot Mechanics to factor in Ammo DPS. (a la many nerfs to Thunderfury)
02/11/08, 10:52 PM   #1452
Stinkerbelle
Von Kaiser

Spayed
Draenei Hunter

Stormscale
 Originally Posted by Midnight Haste doesn´t seem to have much of an impact on those 3:2 rotations. The average overall shot throughput is almost the same for all of them with faster attack speeds just clipping auto shot more than slow weapons. This favors slow weapons over fast ones for absolute damage.
Looking at your numbers, I'm assuming that "general shot delay" means average time per shot (regardless of type), and that this is what you're looking at to evaluate "average overall shot throughput"?

I do find it helpful to measure time per shot as you have done (or the inverse shots/time) rather than DPS results. I think a next logical step would be to compare your results with what you'd expect to achieve from legacy 1:1 and 1.5:1 rotations.

For example, taking the results of your quivered tests:

 Test #2 - 3,0s Weapon, 15% quiverhaste, no SS talent - 2,61s actual attack speed. Number of Auto Shots fired: 322 Number of Steay Shots fired: 677 Total number of shots fired: 999 Auto Shot delay (min/max/avrg): (2,625s/4,610s/3,187s) Steady Shot delay (min/max/avrg): (1,203s/2,000s/1,541s) General Shot delay (min/max/avrg): (0.312s/1.766s/1.043s)
1:1 Expected shot delay*: 1.304 *(i.e. average time per shot)
Improvement: 20%
1.5:1 Expected shot delay: 1.043
Improvement: 0%

 Test #4 - 3,0s Weapon, 15% quiverhaste, 5/5 SS talent - 2,17s actual attack speed. Number of Auto Shots fired: 331 Number of Steay Shots fired: 669 Total number of shots fired: 1000 Auto Shot delay (min/max/avrg): (2,000s/4,813s/3,119s) Steady Shot delay (min/max/avrg): (1,094s/2,110s/1,537s) General Shot delay (min/max/avrg): (0.265s/1.687s/1.031s)
1:1 Expected shot delay: 1.087
Improvement: 5%

 Test #6 - 2,6s Weapon, 15% quiverhaste, no SS talent - 2,26s actual attack speed. Number of Auto Shots fired: 331 Number of Steay Shots fired: 669 Total number of shots fired: 1000 Auto Shot delay (min/max/avrg): (2,922s/4,672s/3,216s) Steady Shot delay (min/max/avrg): (1,063s/2,000s/1,545s) General Shot delay (min/max/avrg): (0.344s/1.625s/1.044s)
1:1 Expected shot delay: 1.130
Improvement: 8%

 Test #8 - 2,6s Weapon, 15% quiverhaste, 5/5 SS talent - 1,88s actual attack speed. Number of Auto Shots fired: 405 Number of Steay Shots fired: 594 Total number of shots fired: 999 Auto Shot delay (min/max/avrg): (0,078s/4,625s/2,275s) Steady Shot delay (min/max/avrg): (1,031s/2,219s/1,541s) General Shot delay (min/max/avrg): (0.031s/2.188s/0.919s)
1:1 Expected shot delay: 0.942
Improvement: 2%

The blue numbers are the average shot times (or "delay" as you've called it) that I'm using to evaluate improvement over expected legacy rotation shot times.

So, if I understand your results correctly, of the four quivered tests, the best improvement you achieved in terms shots/time (or time/shot) was actually not with the slower bow, but in Test 6 (2.6 speed bow, no Serpent's Swiftness), for which you saw a reduced average time per shot of about 8%.

Last edited by Stinkerbelle : 02/11/08 at 11:02 PM.

 02/12/08, 5:45 AM #1453 Midnight Don Flamenco     Middi Night Elf Hunter   Ysera (EU) First of all - you´re right I confused the 3:2 rotation with a 2:1 rotation actually. 3:2 meaning Auto - Steady - Steady. That beeing said there seems to be a general uncertainty about those terms, so I guess it´d be a good idea to clarify them. To my understanding they´re used as follows: 1:1 = Auto - Steady (steady may be replaced with either multi or arcane, there´s only one special per auto though) 1:1.5 = Auto - Steady - Auto - Steady - Special (special is either multi or arcane) 2:3 = Auto - Steady - Auto - Steady - Steady (actually the same as 1:1.5 but no specials but steady are beeing use) 1:2 = Auto - Steady - Steady (kind of uncommon rotation I´ve never seen theorycrafted in depth about anywhere on these boards, but the most common rotation on my macrotests) To answer your questions Stinkerbelle: delay is the time between two shots. Steady shot delay is the delay between two steady shots whereas general shot delay is the delay between any two shots following on each other so the latter is the one determing the absolute shot throughput. Assuming all shots would do roughly the same damage the highest shot throughput would also yield the best dps result. I think I´m gonna do a few tests with a standard castsequence macro and see how it compares.
 02/12/08, 6:12 AM #1454 Rokh Piston Honda   Rokh Orc Hunter   Emerald Dream Yeah, like right now I have enough haste with a 3.0 weapon to spam the macro pretty much, under most conditions, and produce a 2:3 rotation. Your tests just didn't have that amount of haste, I guess. A 2.9 weapon + BM spec should have the right speed, it's around 2.08 - 2.1 to be able to spam the macro to produce a 2:3 rotation. And yes you can do a 2:3 rotation without any macro or without any special conditions, it's just harder to time everything 100% perfect and whatnot. So having a bit of haste + the 3.0 bows seems the best way to go, which won't be a problem at all after Sunwell's out, seeing as how there's only like 3 pieces of gear without haste so far. Anyway, yes, all the spreadsheet would have to do for MOST weapon speeds is reduce the GCD of a steady shot if the previous shot was Auto Shot by the time between the end of the previous shot and the auto shot shooting, up to a maximum of .5 seconds, and also remove the latency from affecting steady shot. From there you can just put in a button that does the "2:3" rotation (Auto-Steady-Auto-Steady-Steady) and it would be modeled for MOST situations, aside from the jitter we're seeing in auto-shot cast time. Anyway I'm tired and this got a little longer than I wanted but yeah, that's basically what's going on with this rotation.
02/12/08, 10:38 AM   #1455
Cheeky
Great Tiger

Cheeky
Troll Hunter

No WoW Account
Since I forgot to mail myself my working copy of the spreadsheet form work to home last night I didn't get this released. If I'm not a complete idiot today too it'll get done.

 Originally Posted by Vaxum Here are a few more items to include in the 2.4 gear list: Shadowed Gauntlets of Paroxysm, Sunrage Shoulderpads, Glyph of the Gladiator, Carapace of Sun and Shadow, Hard Khorium Band, Mounting Vengeance, Starstalker Legguards, Surestrike Goggles 3.0, Tunic of the Ranger Lord The shoulder enchants have a typo - Inscription of Vengence/Greater Inscription of Vengence should be spelled Vengeance.

 Originally Posted by Rokh Anyway, yes, all the spreadsheet would have to do for MOST weapon speeds is reduce the GCD of a steady shot if the previous shot was Auto Shot by the time between the end of the previous shot and the auto shot shooting, up to a maximum of .5 seconds, and also remove the latency from affecting steady shot. From there you can just put in a button that does the "2:3" rotation (Auto-Steady-Auto-Steady-Steady) and it would be modeled for MOST situations, aside from the jitter we're seeing in auto-shot cast time.
That's pretty much exactly what I have done. My first pass on the code was to intelligently pick Auto or Steady based on what the relative cooldowns were and prioritize Steady if it was at all available. That lead to a lot of Auto, Steady, Steady, Auto, Steady, Steady rotations, which isn't what we want.

I'll just force the Auto, Steady, Steady, Auto, Steady pattern now. My initial work shows it as being even more of a mana drain than a 1:1, Priority rotation using max rank Arcane and Multi. But that fits with what people are seeing in game.

Do we need the 3:2 rotation only for unhasted rotations? Or do people envision using it for Quick Shots and DST-based rotations too?

 Elitist Jerks Hunter Spreadsheets -- In development