Hunter Spreadsheets -- In development - Page 122

Cheeky · 04/12/08, 1:05 PM

Originally Posted by Tsook

From: Proc Mechanics

There's a lot of good discussion of the effect of haste on ppm here: http://elitistjerks.com/f40/t15553-p...chanics_haste/

My current understanding is that haste effects reduce your chance to proc in order to maintain the same ppm from autos -- however, this reduction also carries over to any specials/instants you use while under haste. That means when Quick Shots procs it makes it less likely that DST (or whatever) will proc off any specials you use while under the effects, reducing your total number of procs (though the procs from autos remains constant).

To be mostly correct you could probably calculate the ppm for each of the different shot rotation tabs since they're all under different amounts of haste and then use the uptime % you already have to come to a weighted average proc %. I'm not sure how far off that would be, though -- obviously your shots fired in the 10% of the time you have DST and Quick Shots both active is going to be more than 10% of your total shots fired.

Ok, based on all that, the model reduces to a 20% uptime for the DST, in all cases. Here is my math:

Baseline: 1:1 Auto:Steady rotation, no latency issues, 2.8s weapon, 41/20/0 spec.

Dragonspine Trophy, Take 2		
Normal Auto Shot Speed		2.03
Proc per Minute Rate		1
% Chance per Auto Shot		3.38%
% Chance per Special Shot       3.38%
Total Shot Frequency		1.01
Shots to Proc		        29.57
Time to Proc		        30.00
Cooldown		        20
Proc Duration		        10
Uptime for Normal Speed		20.00%

Quick Shots Shot Speed		1.76
% Chance per Shot		2.94%
Total Shot Frequency		0.88
Shots to Proc		        34.01
Time to Proc		        30.00
Uptime for Quick Shots		20.00%

Now, I don't have a DST to check, but does 20% uptime seem reasonable? I think most of the equations I use are self-explanatory, but I can provide details if needed. If you introduce latency into the equation the uptime goes down. Using the same rotation, but adding 500ms of latency in a /castsequence macro yields a 19.66% uptime.

Using a 3:2 rotation with a slower weapon can increase the uptime, as you take more advantage of your specials. A 3.0s weapon using 3:2 and little latency gets 21.36% uptime in normal shots, 20% in quick shots (as a 1:1.)

I'd like to release a new spreadsheet before I'm out of town next week. Any peer review of this model is appreciated. Adding in Quick Shots to the mix is fairly trivial.

Lactose · 04/12/08, 1:14 PM

Originally Posted by Cheeky

Ok, based on all that, the model reduces to a 20% uptime for the DST, in all cases. Here is my math:

I'm pretty sure that's incorrect.
Special attacks use the unhasted weapon speed to determine proc chance, not the hasted speed.

With a PPM of 1, and a 2.8 weapon speed, you'd see the following:

Dragonspine Trophy, Take 2		
Normal Auto Shot Speed		2.03
Proc per Minute Rate		1
% Chance per Auto Shot		3.38%
% Chance per Special Shot		4.67% //Changed
Total Shot Frequency		1.01
Shots to Proc		        29.57
Time to Proc		        30.00
Cooldown				20
Proc Duration		        10
Uptime for Normal Speed		20.00%

Quick Shots Shot Speed		1.76
% Chance per Auto Shot		2.94% //Chance per AUTO shot
% Chance per Special Shot		4.67% //Added
Total Shot Frequency		0.88
Shots to Proc		        34.01
Time to Proc		        30.00
Uptime for Quick Shots		20.00%

I only changed the proc chance stuff (bolded and commented), so you'll have update the rest of the values based on that.

alienangel · 04/12/08, 1:32 PM

Originally Posted by Cheeky

Now, I don't have a DST to check, but does 20% uptime seem reasonable? I think most of the equations I use are self-explanatory, but I can provide details if needed. If you introduce latency into the equation the uptime goes down. Using the same rotation, but adding 500ms of latency in a /castsequence macro yields a 19.66% uptime.

The post you were quoting had misread the thread, as Lactose said, so I don't think you can get away with a "Total Shot Frequency" method - you'd need one shot frequency for autos, and one for specials, each with different proc chances.

20% uptime doesn't seem right. I was doing some log parsing at http://elitistjerks.com/708890-post3395.html, and had 56% uptime during Nalorakk, and 39% uptime on Halazzi. Both are short fights though, so the variance is large.

If anyone has a larger sample though (say a night of brutallus attempts, going all out every attempt to keep the shot frequency representative) it would be nice to run through the parser to see what kind of values to expect the model to come up with.

Actually, I'll try to get a shadowpriest to come with me and nuke Dr. Boom for an hour, that would remove RF and Heroism's effects on the logs since the spreadsheet model probably isn't factoring those in.

Cheeky · 04/12/08, 3:37 PM

Originally Posted by alienangel

The post you were quoting had misread the thread, as Lactose said, so I don't think you can get away with a "Total Shot Frequency" method - you'd need one shot frequency for autos, and one for specials, each with different proc chances.

20% uptime doesn't seem right. I was doing some log parsing at http://elitistjerks.com/708890-post3395.html, and had 56% uptime during Nalorakk, and 39% uptime on Halazzi. Both are short fights though, so the variance is large.

If anyone has a larger sample though (say a night of brutallus attempts, going all out every attempt to keep the shot frequency representative) it would be nice to run through the parser to see what kind of values to expect the model to come up with.

Actually, I'll try to get a shadowpriest to come with me and nuke Dr. Boom for an hour, that would remove RF and Heroism's effects on the logs since the spreadsheet model probably isn't factoring those in.

With a 10 second duration and a 20 second cool down, how is a 50%+ uptime possible?

I'm working through amortizing the chance per shot, based on the ratios of Autos to Specials. Not sure it will be 100% accurate, but it should suffice.

alienangel · 04/12/08, 5:02 PM

That's a good question :S

the portion of the parser (linked in the procs thread) that measures uptime is the simplest part though, just summing together the times between aura application and aura removal, I'll poke around the logs to see if they match the summary.

alienangel · 04/12/08, 9:02 PM

Originally Posted by alienangel

That's a good question :S

the portion of the parser (linked in the procs thread) that measures uptime is the simplest part though, just summing together the times between aura application and aura removal, I'll poke around the logs to see if they match the summary.

All right, I found out what the problem was. I think Tunah originally wrote the script to parse logs from his own solo testing of procs, since while it does check actor IDs to make sure it's only counting one person's attacks, it doesn't do the same for counting procs - since I was parsing logs where multiple people had DSTs, it was recording both my DST and the rogue's DST procs.

With the parser changed to only count my own procs, the numbers work out to:

DST:
Uptime: 28.77% (57.7s/200.6s)
Uptime2: 33.50% (67.2s/200.6s)
Internal cooldown: 20
Proc chance: 7.50% (6/80)

Madness:
Uptime: 40.46% (64.2s/158.7s)
Uptime2: 46.78% (74.3s/158.7s)
Internal cooldown: 0.0
Proc chance: 5.26% (7/133)

QuickShots:
Uptime: 39.75% (72.0s/181.1s)
Uptime2: 46.17% (83.6s/181.1s)
Internal cooldown: 0.0
Proc chance: 11.54% (6/52)

The reason I've listed two uptimes for each is because the script discards the last proc duration from the calculation. I'm not entirely sure why. The Uptime2 figure uses the full uptime, which gives a result closer to manually going through the log, adding up the uptime and dividing by the time between first and last shots.

I uploaded my hack of the original script at Procalyzer for non-personal logs [ruby] [wow] [procs] if anyone else wants to run some logs through it - just change the line near the bottom to reflect the buff you want to track for the person who recorded the log (you may want to change the filter too if the proc only occurs on specific types of attacks (e.g. Ashtongue/Quick Shots)).

Cheeky · 04/12/08, 11:39 PM

Originally Posted by Lactose

I'm pretty sure that's incorrect.
Special attacks use the unhasted weapon speed to determine proc chance, not the hasted speed.

With a PPM of 1, and a 2.8 weapon speed, you'd see the following:

Dragonspine Trophy, Take 2		
Normal Auto Shot Speed		2.03
Proc per Minute Rate		1
% Chance per Auto Shot		3.38%
% Chance per Special Shot		4.67% //Changed
Total Shot Frequency		1.01
Shots to Proc		        29.57
Time to Proc		        30.00
Cooldown				20
Proc Duration		        10
Uptime for Normal Speed		20.00%

Quick Shots Shot Speed		1.76
% Chance per Auto Shot		2.94% //Chance per AUTO shot
% Chance per Special Shot		4.67% //Added
Total Shot Frequency		0.88
Shots to Proc		        34.01
Time to Proc		        30.00
Uptime for Quick Shots		20.00%

I only changed the proc chance stuff (bolded and commented), so you'll have update the rest of the values based on that.

Ok, carrying that through:

Base Weapon Speed		2.8
Normal Auto Shot Speed		2.03
Proc per Minute Rate		1
% Chance per Auto Shot		3.38%
% Chance per Special Shot       4.67%
Auto Shot Frequency		2.03
Special Shot Frequency		2.03
Auto Shots to Proc		29.57
Special Shots to Proc		21.43
Time for Auto		        60.00
Time for Steady		        43.48
Weighted Delay		        25.21
Auto Shots to Proc		6.38
Cooldown		        20
Proc Duration		        10
Uptime for Normal Speed		22.12%

To compute the weighted delay, I combined them as 1/((1/Auto Time) + (1/Special Time)), sort of like resisters in parallel. Not sure if that's accurate, it just felt right, and I've been drinking too much to think clearly.

Cheeky · 04/13/08, 10:01 AM

Change Log:
- Armory scraping for Gear now correctly pulls most enchants too. Thanks to Drakier of Uther for providing the framework and the current set of supported enchants.
- Corrected some item stats (Band of Ruinous Delight, Hauberk of the Warbringer, Tunic of the Ranger Lord).
- Corrected some item names (Scale Leggings of the Skirmisher).
- Added some new items (Bracers of the Forest Stalker, Figurine - Shadowsong Panther, Cloak of the Coming Night, Sunrage Treads, Stalker's Chain Armor, Stalker's Chain Helm, Stalker's Chain Leggings, Stalker's Chain Spaulders, Stalker's Chain Gauntlets, Assassin's Alchemist Stone, Leggings of the Betrayed, Band of Celerity).
- Modeled the Shattered Sun Pendant of Might similar to other, similar trinkets (45 second internal cool down, 10% chance on hit.)
- Allow for selection of Aldor/Scryer factions (only applies to Shattered Sun Pendant of Might) on the 'Settings and Results' tab.

I'll be out of town next week on business. Please send all bug reports to Lactose.

Next up:
Gear sorting based on DPS (starting to put infrastructure in place)
Better modeling for PPM-based items
Better model for Scryer [Shattered Sun Pendant of Might] proc, once we have more data.
Various bug fixes (like auto calculation of stat->DPS changes on gear updates.)

• Relwin · 04/13/08, 3:33 PM

Being that I haven't seen any real data gathered about the new Oil, I decided to do some quick and dirty testing to check it out.

Righteous Weapon Coating Testing

About halfway through the log I switched from a Twinblade to a Silithid Claw/Brutality Blade combo. Procs never overlapped and seemed to share the ICD when using two weapons. ICD was 45 seconds and a rather high proc rate. Altogether it's a 66AP equivalent for hunters.

Rakhmamort · 04/13/08, 11:03 PM

I've downloaded this sheet a couple of months ago and have used it to improve my hunter. I have, after more than 2 years of being an MM build hunter, chosen to re-spec to BM due to what I got from this sheet. Anyway, after several versions of the worksheet, I'm tired of being unable to load my profile from the armory. I've always gotten the error message 'A connection to the server could not be established'. I didn't mind the first couple of times loading my gear manually, t'was fun actually since I was looking at the other gear I could use, but doing it after every version change is tiring.

Is there anything I have to set so I can get the Load Gear from Armory button to work for me?

Meursault · 04/13/08, 11:11 PM

I only have OpenOffice which I hate and is difficult to use well, and I didn't feel like digging through your code, but...

how exactly do you model shots under a 3:2? Do you force the user to use a 3:2 exactly, or do you go by the actual shots/time of macrospam using the /cast !Auto Shot /cast Steady Shot. I only ask because (as I'm sure you know) as haste increases macrospam quickly becomes a 5:4 macro and approaches 1:1 as auto shot speed approaches 1.5.

Indora · 04/14/08, 5:19 AM

@Meursault
At the moment it's a perfect 3:2 rotation, unlike the most player sees ingame, using their macros.
But Kamma provided an algorithm for calculating the steady:auto ratio.

More details under:
http://elitistjerks.com/f31/t11420-h...ble/p124/#3087

There are still mistakes..
-Auto Shot cast time is missing.
-GCD isn't delayed by latency.

I didn't have the time to fix them since two weeks now. -.-

Meursault · 04/14/08, 6:43 AM

I ask because I recently did some similar work to post a guide on the WoW forums, although I'm pretty sure I didn't make it accessible enough to the average forum-goer, so it'll fall through the forums in a day. The closed form for the actual shot rotation is actually quite simple, and the values look a lot like the work Kamma did.

Here's the napkin math, with useless events cut out since you won't need them. I calculated these with an "infinite" spam speed, adding an expected delay of spam speed/2 keeps it simple and approximates well, but isn't quite as pure as I like.

Time~Event
0 ~ Initial Steady and Auto begin casting ("A0")
.5~ Nil Auto releases. For the remainder of this section, "An" refers to the time of the beginning casting of an auto shot, it will always release .5 seconds later in the middle of the subsequent Steady channel.
1.5 ~ Second Steady begins casting
1.5+1.5/H ~"A1" time of starting casting of first auto shot
A1+B/H ~ "A2", second Auto begins casting
3.0-A1-B/H ~ "S2", or the spillover GCD time from the second steady.
A2+B/H ~ "A3"=1.5+1.5/H+B/H
4.5-A2-B/H ~ "S3"=4.5-1.5/H-2B/H
A3+B/H ~ "A4"=1.5+1.5/H+2B/H
6.0-1.5/H-3B/H ~ "S4"

It's easy enough from here to see the function for the "spillover" of GCD time, or
"Sn"=1.5n-1.5/H-(n-1)B/H

I'm sure you can recreate this easily enough, as this isn't a very good formatting to view it in, as I don't have any suitable software on this computer.

Post on WoW forums

edit: 4/14 I realize the naming of variables is a bit confusing as it's written, I'll clean it. The timings are not wrong, although I had a propogated sign error in the final result. See my next post

Indora · 04/14/08, 1:21 PM

I'm sorry I doesn't have the time to look through your whole post, but

Originally Posted by Meursault

0 ~ Initial Steady and Auto begin casting ("A0")
1.5 ~ Second Steady begins casting

seems wrong.

The GCD can start while your autoshot is casting. Thus the second steady can start earlier.

Makata · 04/14/08, 8:27 PM

Can anyone tell me how to factor in Vampiric Touch, Bloodlust, Haste pots and the Righteous Weapon Coating into the dps calculations?