Cat DPS Rotation - Page 38

Stejo · 02/17/09, 2:57 PM

Originally Posted by Yukizawa

I think I need to make an option to choose the number of SR combo points, to make it easier to test different rotations. I know most of what I've read suggest 5 point SRs, but I can't figure out why. The spreadsheets / testing I've done show me dropping rip more with 5 point SRs, or refreshing it signficantly early every time. Is that a function of crit level?

First of all the addon is excellent, I was pleasantly surprised by it and I recommend it to anyone who has not tried it yet. Now about 3vs5 cp SRs; the difference is cp generation per sec. Maintaining a 5SR/5RIP rotation on a dummy is next to impossible with current gear. In a 25 man raid however the crit and haste gains make a 5SR/5RIP rotation pretty trivial and the question becomes when is the right moment to throw in the FBs. The gain of doing 5cp SR is more space to use FB if in any given cycle you get lucky with OOC procs or back to back crits.

I believe that simply adding a second preset that can be manually selected and switches the cp>=3 for SR check to cp=5 would do the job without messing with the rest of the functionality. On a sidenote I think that this addon probably deserves a separate thread.

Allev · 02/17/09, 6:23 PM

For why to use 5-point SR, take a gander at page 7, post 161 here. Essentially, SR makes more efficient use of CPs to uptime than it uses energy to uptime.

nightcrowler · 02/17/09, 8:02 PM

	if tf < GCD and energy < 40 and berserk < 165 then
		--print "TF"
		return "Tiger's Fury"
	end
	
	if berserk < GCD and energy > 70 and sr > 0 then
		--print "berserk"
		return "Berserk"
	end
	
	if sr < 0.1 then
		if cp >= 1 then
			return "Savage Roar"
		end
	end
	
	
	if rip < 0.1 then
		if cp == 5 then
			--print "RIP"
			return "Rip"
		end
	end
	
	if mangle < 0.1 and trauma <= 0 and cp<5 then
		--print "Mangle"
	    return "Mangle - Cat"
	end


	if ooc > 0 then
	if mangle < GCD then
	    return "Mangle - Cat"
	end
		return "Shred"
	end
	
	if rake < 0.1  and cp<5 then
		--print "Rake"
		return "Rake"
	end
	
	
	if tf < 2 and energy > 30 and berserk < 165 and cp<5 then
		return "Shred"
	end
	
	
	if (energy >= 67 or (berserk >= 165 and energy >= 33 )) and cp<5 then
		--print "Shred"
		return "Shred"
	end
	
	if (energy <= 55 or (berserk >= 165 and energy <= 32) ) and sr >= 5 and rip >= 6 and cp == 5 then
		return "Ferocious Bite"
	end

	if (energy >=55 or (berserk >= 165 and energy <= 32) ) and sr >= 5 and rip >= 6 and cp == 5 then
	if mangle < GCD and trauma <= 0 then
		--print "Mangle"
	    return "Mangle - Cat"
	end
	
		return "Ferocious Bite"
	end

	if sr<5 and rip>=(7-(energy/10)) and cp == 5 then
		return "Savage Roar"
	end

	if energy>=80 and cp == 5 then
		return "Savage Roar"
	end
	
	if energy >= 62 and rip <=2 and cp < 5 then
		return "Shred"
	end


	
	--print "FFF"
	if fff < GCD+6 then
		return "Faerie Fire"
	end
	
	return ""

end

This is FaceMauler decision algoritm changed to match my simulation data (well not exactly but pretty close, and obviosly the cycle I use). I've tryied the addon and I should tanks it. It could be cool if in a future release he will make us able to choce between some different cycles from the interface

. Excelent works keep it going.

Important stats to use that cycle: 0/55/16 spec (with FB talented). More than 50% crit, hit/expertise capped (otherwise the cycle will be too difficult to mantain)

Yukizawa · 02/17/09, 11:09 PM

This is FaceMauler decision algoritm changed to match my simulation data (well not exactly but pretty close, and obviosly the cycle I use). I've tryied the addon and I should tanks it. It could be cool if in a future release he will make us able to choce between some different cycles from the interface

. Excelent works keep it going.

Important stats to use that cycle: 0/55/16 spec (with FB talented). More than 50% crit, hit/expertise capped (otherwise the cycle will be too difficult to mantain)

I made this an alternate algorithm in the most recent version, which will be on curse as soon as it is approved. Thanks for the suggestion.

I also incorporated a number of other suggestions/fixes from this thread / comments / PMs; thanks for all the advice.

Valerian · 02/17/09, 11:19 PM

Hmm even following the facemauler algorithm (both the default and the one posted by Nightcrowler) I cannot match the results in seeing in Rawr (even completely unbuffed). Rawr is giving me values of ~3100 and I'm hitting max 2900. I did just under 10 minutes so I suspect this should be enough to move out most RNG elements.

Does anyone know what exactly the algorithm used for Rawr is? I.e, what "rotation" it is simulating?

Aldhissla · 02/17/09, 11:21 PM

Originally Posted by Valerian

I did just under 10 minutes so I suspect this should be enough to move out most RNG elements.

No, it's not.

On a more general note, it would be unlikely to perform at the level of the "theoretically simulated maximum" (assuming rawr's model is accurate) anyway - due to human error, bad luck, latency, etc.

nightcrowler · 02/18/09, 12:43 AM

In a 10 minute fight you can see up to 7-8% dps difference.
Rawr is a statistical model, i'm pretty sure it undervalue dps loss for waiting time (it use a fixed SR/RIP rotation)

a civilian · 02/18/09, 1:40 AM

Rawr doesn't use a rotation; it assumes you can always put combo points into rip (with no waste) until you reach 100% uptime. That is, the number of rips over the course of the fight is set as simply

(duration of fight)/(duration of rip)
or
(number of combo points generated over the course of the fight [after subtracting for savage roar])/5,

whichever is less.

Maihes · 02/18/09, 2:42 AM

Thank you to all the info thus far in this thread, the reason I am posting though is about the trackers, quite a few people linked. Of all the ones linked that I tried I couldn't recall which one linked was just a single block of icons.

2 Hours after hitting 80 I was pulling ~3k DPS on Patch, but the timing is still off on everything I am watching. The one addon for displaying the data that I started using mid way through raid was Badkitty. But the base display seems so clunky compared to some pics I saw of others in this thread.

Also how heavily do cats weigh ArP in comparison to other stats. Not gemming wise, base gear wise.

nightcrowler · 02/18/09, 4:56 AM

Originally Posted by a civilian

Rawr doesn't use a rotation; it assumes you can always put combo points into rip (with no waste) until you reach 100% uptime. That is, the number of rips over the course of the fight is set as simply

(duration of fight)/(duration of rip)
or
(number of combo points generated over the course of the fight [after subtracting for savage roar])/5,

whichever is less.

As I said that's why Rawr numbers can't be reached. If you only use RIP in your rotation you will have tons of waiting time. So you need to use FB to avoid waiting time but FB is less DPE than RIP.

nightcrowler · 02/18/09, 4:58 AM

Originally Posted by Yukizawa

I made this an alternate algorithm in the most recent version, which will be on curse as soon as it is approved. Thanks for the suggestion.

I also incorporated a number of other suggestions/fixes from this thread / comments / PMs; thanks for all the advice.

Thanks a lot. I've seen on Curse that you added my algoritm that's cool. On a side note: I can read the fix and news about 0.9e version, but the most recent version on curse is still 0.9d.

EDIT: Damn sorry for double posting.

Stejo · 02/18/09, 7:58 AM

I just tried version 0.9e. Been testing 0.9d extensively earlier, so a couple quick pointers regarding the default algorithm before I test this one more extensively too.

FF check seems to have changed, the icon pops from the moment the timer hits 30 onwards now which is pretty annoying and frankly quite useless. An option to stop suggesting FF usage altogether would be welcome since most ferals are used to do it on every free gcd anyway.

It keeps suggesting to clip the last tick of rake all the time which didn't happen in 0.9d. Don't know if it's intented or not but it's an obvious dps loss. Same behavior for mangle, suggesting to refresh it couple sec before it expired. [edit: apparently same behavior on all abilities, considers them down before they actually expire]

At 45% crit unbuffed on a dummy, cp generation seems to be enough to maintain a 5SR cycle using the default algorithm without sacrificing rip uptime at all. FB flag still raised in very rare occasions and only during berserk. Further testing required here. [Edit: Wrong, disregard]

The alternative algorithm sacrifices a good bit of rip uptime (can't give exact percentage yet) in favor of FB. First couple million dmg done using it shows rip dmg % dropping from 20ish% of total dmg to ~17% while FB becomes ~5% of total dmg which seems to be an overall gain. Total dps output between the 2 algorithms however is very close so far, within statistical error. Will update the post later with more solid numbers.

UPDATE.

Default algorithm, 3SR

Fight duration: 9:58, 3215,1 dps
White 28.5%
Shred 23.6%
Rip 20.9% (87% uptime)
Rake 19.5%
Mangle 4.9%
FB 2.6% (5 entries, 4 crit 1 hit)

Default algorithm, 5SR (done this twice for doublechecking, 2 sets of numbers)

Fight duration: 9:06, 3116,8 dps - 9:38, 3115,6 dps
White 29.3% - 29.7%
Shred 24.5% - 25.3%
Rip 19.9% (80,6% uptime) - 19% (76,1% uptime)
Rake 19% - 19.1%
Mangle 5.1% - 4.8%
FB 2.2% (5 entries, 4 crit 1 hit) - 2% (4 entries, 3 crit 1 hit)

Alternative algorithm.

Fight duration: 9:47, 3186,2 dps
White 29.3%
Shred 25.6%
Rip 18.9% (78.5% uptime)
Rake 17.4%
Mangle 4.6%
FB 4.1% (8 entries, 6 crit 2 hit)

Seems the uptime loss of Rip when doing 5SR cycles is bigger than I estimated initially and the strict limitations for FB usage on the default algorithm don't allow it to make up for the loss. The alternative algorithm basically doubles the amount of suggested FBs without affecting Rip uptime much compared to the 5SR default. The loss in rake is probably due to the clipping since it was the first test I did of the 3 and I wasn't looking out for it as much as I did later.

I don't see much point trying to analyze these results further since they have small value in a real raiding setting. I'll try and test it in some actual fights tonight and come up with more data.

Yukizawa · 02/18/09, 9:02 AM

Originally Posted by Stejo

I just tried version 0.9e. Been testing 0.9d extensively earlier, so a couple quick pointers regarding the default algorithm before I test this one more extensively too.

FF check seems to have changed, the icon pops from the moment the timer hits 30 onwards now which is pretty annoying and frankly quite useless. An option to stop suggesting FF usage altogether would be welcome since most ferals are used to do it on every free gcd anyway.

It keeps suggesting to clip the last tick of rake all the time which didn't happen in 0.9d. Don't know if it's intented or not but it's an obvious dps loss. Same behavior for mangle, suggesting to refresh it couple sec before it expired.

At 45% crit unbuffed on a dummy, cp generation seems to be enough to maintain a 5SR cycle using the default algorithm without sacrificing rip uptime at all. FB flag still raised in very rare occasions and only during berserk. Further testing required here.

The alternative algorithm sacrifices a good bit of rip uptime (can't give exact percentage yet) in favor of FB. First couple million dmg done using it shows rip dmg % dropping from 20ish% of total dmg to ~17% while FB becomes ~5% of total dmg which seems to be an overall gain. Total dps output between the 2 algorithms however is very close so far, within statistical error. Will update the post later with more solid numbers.

I think I know why it's suggesting that you clip rake. I was trying to lower the amount of "spammy" change, so it predicts up to 1 second into the future instead of right now. Obviously, this means you need to wait a little bit if your going to clip, but that's not obvious, so something needs to change. Maybe with move prediction I can turn that back down again. the FFF change was to use FFF if it isn't on cooldown, but that may be too much. I'm still trying to find a good mix for it. I'll tweak it some more tonight.

Blazefire · 02/18/09, 12:52 PM

Thats some really nice work on Facemauler, however it kinda makes me sad that tools like this take the thought process out of playing the game. This would be very similar to the powershifting macros that came out in TBC. It took the skill of powershifting as a DPS boost and turned it into a 1 button spam that any old monkey could do. Now, similar with this mod, you are now just playing simon says, pushing the buttons the mod tells you to instead of making the determination for yourself.

What ever happened to actually playing the games instead of having a mod play for you?

Yukizawa · 02/18/09, 1:11 PM

Originally Posted by Blazefire

Thats some really nice work on Facemauler, however it kinda makes me sad that tools like this take the thought process out of playing the game. This would be very similar to the powershifting macros that came out in TBC. It took the skill of powershifting as a DPS boost and turned it into a 1 button spam that any old monkey could do. Now, similar with this mod, you are now just playing simon says, pushing the buttons the mod tells you to instead of making the determination for yourself.

What ever happened to actually playing the games instead of having a mod play for you?

I do know what you mean. I wrote this because I didn't like the debuff/buff watchers, as they give you too many things to pay attention to, but I needed something to help me keep track of what needed to come next. I have found that this acts like training wheels. After you get used to the rotation, you end up knowing what will come next without even looking at the mod, and use the mod as a reminder that mangle or rake is down, rather than paying complete attention to it. That, and if you just follow the mod, you will eventually start finding places where the mod is wrong, because of some condition it isn't looking for, and will miss dps if you don't veer off course a little bit.