Hunter Spreadsheets -- In development - Page 91

• Intermission · 02/02/08, 3:39 AM

edit: im blind

Cheeky · 02/02/08, 8:35 AM

Both Excel and OO versions posted.

Change Log:
- Added in new gear ([Demon Stalker Shoulderguards], [Charmed Amani Jewel])
- Fixed a bug with how pet specials factored into Ferocious Inspiration uptime.
- Changed Hunter attacks to a 2-roll system, where hit and crit are determined separately.
- Made very clear in Hit/Crit chances what the target level is.
- Changed default target level back to 73. There are just as many people who couldn't figure out why their hit changed from previous versions as there were people who didn't know where 0.6% of their crit went.
- Corrected some item names ([Prowler's Strikeblade], [Sun-Gilded Shouldercaps]).
- Fixed the bug where the Rotation for different profiles wasn't being saved properly (display in 'Settings and Results' bug only.)
- Corrected some item stats ([Gladiator's Chain Helm]).
- Provided a better model for the Curse of Shadows Debuff.

Midnight · 02/02/08, 9:54 AM

Looks like there´s a bug with critical hit damage calculations now.

On the calculations tab for auto shot where it used to be 236,90% with version 40 it says 133,90% now (same crit chance of 36,40%). Also the Crit/Miss Adjustments went down from 149,34% to 111,84%.

left Side v40, right side v41

edit: included some screenshots for better illustration. This also affects Steady Shot and Multi Shot but whereas it´s a hit of over 200 DPS for auto shot it´s less than 9 dps for Steady and Multi. I guess this is propably somehow related to the changes for the 2 roll system that differs between white and yellow hits.

Koroshiya · 02/02/08, 10:06 AM

Hopefully Cheeky won't mind... but here's an alternative place to download the latest spreadsheets. It took me a while to download from his site so just offering an alternative.

Excel: http://www.altoholic.com/wow/spreads...S_Calc_v41.xls
Open Office: http://www.altoholic.com/wow/spreads...alc_v41_OO.xls

Dec · 02/02/08, 10:16 AM

Cheeky, on the 'Gear' tab, cell L88 is not picking up the +hit set bonus.

Cheeky · 02/02/08, 12:36 PM

Originally Posted by Midnight

Looks like there´s a bug with critical hit damage calculations now.

On the calculations tab for auto shot where it used to be 236,90% with version 40 it says 133,90% now (same crit chance of 36,40%). Also the Crit/Miss Adjustments went down from 149,34% to 111,84%.

edit: included some screenshots for better illustration. This also affects Steady Shot and Multi Shot but whereas it´s a hit of over 200 DPS for auto shot it´s less than 9 dps for Steady and Multi. I guess this is propably somehow related to the changes for the 2 roll system that differs between white and yellow hits.

Turns out I forgot to move Auto Shot to the 2-roll system, but decreased the critical strike damage ratio from 2.3x to 1.3x (because crits are now just a ratio of extra damage per shot.)

I'll get an updated version posted in the next few minutes.

Cheeky · 02/02/08, 12:48 PM

Change Log:
- Have Auto Shot use the 2-roll system and have proper critical strike damage.
- Use +hit gained from set bonuses.

OO version updated too.

Sorry about those bugs guys.

pharzo · 02/02/08, 1:18 PM

[Jin'rohk, The Great Apocalypse] is mispelled in the spreadsheet, and does not import from armory correctly. Missing the 'h'

alienangel · 02/04/08, 10:32 AM

Originally Posted by Cheeky

- Changed Hunter attacks to a 2-roll system, where hit and crit are determined separately.

Could you tell us what kind of 2-roll system you're using Cheeky?

A small nit-pick I had is that we all seemed to assume that if a single-roll system is discounted we'd be left with a "first roll for hit/miss, then roll for hit/crit" system. While agree that that's the most likely alternative to the 1 roll system, all that Lactose's recent test with the banshee showed is that we're not working with a 1 roll-system, not that we're actually working with a 2-roll system (e.g. blocks could be on a 3rd roll) or if we are, what kind of 2-roll system, since the rolls could for instance be "roll for hit/crit, if hit roll for hit/miss" (and unless I'm misunderstanding, that would again change the relative values of hit/crit).

Again, I think it's a safe assumption that if it's 2-roll it'll be roll for hit, if hit roll for crit, and just to be complete I'll probably go test that when I get a spare few hours, but stating what model you went with is probably good for the spreadsheet. And afaik we still don't know when the roll for blocks happens in the 2-roll system (edit: although for your dps simulation blocks aren't relevant).

Cheeky · 02/04/08, 10:42 AM

Originally Posted by alienangel

Could you tell us what kind of 2-roll system you're using Cheeky?

A small nit-pick I had is that we all seemed to assume that if a single-roll system is discounted we'd be left with a "first roll for hit/miss, then roll for hit/crit" system. While agree that that's the most likely alternative to the 1 roll system, all that Lactose's recent test with the banshee showed is that we're not working with a 1 roll-system, not that we're actually working with a 2-roll system (e.g. blocks could be on a 3rd roll) or if we are, what kind of 2-roll system, since the rolls could for instance be "roll for hit/crit, if hit roll for hit/miss" (and unless I'm misunderstanding, that would again change the relative values of hit/crit).

Again, I think it's a safe assumption that if it's 2-roll it'll be roll for hit, if hit roll for crit, and just to be complete I'll probably go test that when I get a spare few hours, but stating what model you went with is probably good for the spreadsheet. And afaik we still don't know when the roll for blocks happens in the 2-roll system (edit: although for your dps simulation blocks aren't relevant).

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. Here is an example using 95% hit rate, 30% crit rate, +130% crit bonus:

1-roll - (2.3 * 0.3) + (0.95 - 0.3) = 134% damage increase.
2-roll - ((1.3 * 0.3) + 1) * 0.95 = 132% damage increase.

Where hit = 100% they are even:
1-roll - (2.3 * .3) + (1.0 - .3) = 139%
2-roll - ((1.3 * 0.3) + 1) * 1 = 139%

Deatiled testing of above, 100 attacks for 100 damage each:
2-roll, 30 crit for 230, 65 hit for 100, 5 miss for 0 - 134 average damage
1-roll, 5 miss for 0 damage, 28.5 crit for 230, 66.5 hit for 100 - 132.05 average damage

Glaurong · 02/04/08, 11:56 AM

As far as I can tell from the testing that has been done so far:

Roll 1 (Avoidance):
Hit/Miss/Block

Roll 2 (Critical Hit):
Crit/Normal

Roll 3 (arcane shot only):
Partial Resist

Which is almost exactly how spells work (minus block).

Glaurong · 02/04/08, 12:01 PM

Originally Posted by Estala

Unless I'm insane/mistaken I am quite positive I've missed against frost mages with sufficient cumulative hit. I don't have any actual data to corroborate it so I may be incorrect. I'm also pretty sure I recall rogue's in PvE gear missing specials. If there is something funky going on with a rogue specials.

I may also be completely mistaken, I didn't mean to make such a strong assertion without any concrete evidence to prove it.

This could be a side effect of the order effects are applied when generating an attack table.

For instance (this is just an example):

Base table + talent modifiers + gear modifiers + self buff/debuff modifiers = Your attack table.

To keep the amount of computation down, this is only calculated when one of the above things changes, not on every for every attack.

On each attack, your final attack table could be computed by adding your table with an opponent defense table, which could consist of the same stuff, talents, gear, buffs/debuffs.

If things are actually calculated this way then you can have talents of your target overriding your gear.

Saryanne · 02/04/08, 1:11 PM

Hey Cheeky,
Is there any way to work out a rotation for the "new" macro that is going around? I tried looking back, but couldn't find anything.

Cheeky · 02/04/08, 2:12 PM

Originally Posted by Saryanne

Hey Cheeky,
Is there any way to work out a rotation for the "new" macro that is going around? I tried looking back, but couldn't find anything.

I haven't seen a solid mathematical model for it yet, and have had some difficulty coming up with one that fits all the logs I've seen myself. I think there is a heavy random factor (either latency or the jitter in combat logs and how fast you can spam the macro) that is a large component to the way the macro behaves.

If you have a suitable model, post it here and I'll see if I can get it incorporated.

Cheeky · 02/05/08, 12:10 PM

So I've taken a stab at the new macro and how to model it.

During my investigations I discovered something no one had been talking about, but which is part of why the macro works like it does. Auto Shot's cast speed is not 0.5s in practice. I analyzed almost 100 times when chained Steady Shots would clip an Auto and measured the time between the last Steady Shot and the Auto Shot. I understand there is jitter in the combat logs, so I'm hoping the sample size is enough to draw some conclusions. Here are the results:

1 - 70 of the 96 test cases showed a difference under 0.5s. 26 showed over.
2 - Over all test cases there was an average of 0.44823s between Auto and the previous Steady.
3 - This works out to an 11.55% haste effect over the tooltip 0.5s cast time.

Using that data, and what we've seen regarding client-only GCD, I've modified the Shot Rotation matrix as follows:
1 - Latency is no longer added to Steady Shot cast time.
2 - If the shot previous to a GCD-using shot was an Auto Shot, I readjust the GCD start to be the greater of previous GCD or previous Auto Shot cast time + latency.
3 - Auto Shot's cast time is adjusted by the 11.55% haste factor I've found above.

With those changes to both the regular Shot Rotation and Quick Shots I also added 2 new "Steady Spam" buttons. These check Auto vs. Steady cooldowns for each shot and pick whatever is next.

Doing all that, using a 0.15s latency and my gear, led to the following DPS totals (Hunter only):
[Serpent Spine Longbow] - 827 DPS, 17:33 Auto:Steady in regular, 22:28 in QS
[Sunfury Bow of the Phoenix] - 863 DPS, 20:30, 23:27
[Steelhawk Crossbow] - 840 DPS, 20:30, 24:26
[Wolfslayer Sniper Rifle] - 847 DPS, 20:30, 25:25
[Barrel-Blade Longrifle] - 848 DPS, 22:28, 25:25

If this holds true (see caveats below) then the sweet spot of weapon speed seems to be about 2.9s. This also seems to be what the anecdotal evidence is saying. We may still be weapon speed dependant, just on a different, much more narrow speed now.

Caveats:
0 - This was a modified version 40 of the spreadsheet, it was only to look at timings.
1 - We need much more testing under different constant haste effects to determine exactly how Auto Shot's cast time is modified.
2 - With jitter in latency, it will be almost impossible to ever model this in a way that can be replicated consistently in game. I saw variances in Auto Shot cast time from 0.17 - 0.8s.
3 - This whole mechanic is so fucking stupid I'm seriously thinking about giving up on modeling Hunters. I can't for the life of me figure out how any of this is "working as intended" and I really wish Blizzard would at least give us some indication of what they'd like our mechanics to be.
4 - I could be completetly wrong about all of this, but it's the only model I've been able to come up with that fits the facts I've seen. The logs with 4-5 Steady chains still worry me, and it may be related to a small window for clobbering available right before an Auto Shot is cast (maybe the time between the 0.5s base cast time and the hasted time?)

If we get some concensus on whether or not this model is sound I can release it. I hope to provide a spreadsheet of the Auto Shot data I collected later tonight.