Sign in to follow this  
Followers 0
CptDan

Warning: Intense Geekery and Theorycrafting 25/10 Man resto

45 posts in this topic

I'm beginning this thread as a focused place for 25 man theory crafting. I have a lot of questions that only the lovely numbers can answer. I would also like to call out to other 25 man raiders to give their input and maybe share some logs with us.

 

Stoove, as our resident theory crafter I would like some help with what data I can best capture to help with analysis. I'm not bad at math, but I'm not sure which data I will need to answer my questions. Every raid is logged on my end, so logs should be plentiful.

 

I pulled some of the questions from other threads that didn't really get answered. The theory of what we do and the math behind it is a large source of entertainment for me.

 

Questions to Answer:

 

How to optimize our builds to increase our top performing spells? (for me, HR & HST) Would haste or crit help these spells more?

 

How to recognize our ideal haste breakpoints based on the gear we have

 

Value of Haste/Crit/Mastery as our gear levels and spell power increase.

 

 

I know stoove runs 10 man so he'd have some great input there, Let's see if we can get more data to look over for 10 man and 25 man.

 

To keep it easy, let's try to follow a standard format similar to asking for help.

 

Armory Link

WOL Link

Raid Comp

Feelings of the build and spell choice/priority.

Edited by CptDan

Share this post


Link to post
Share on other sites

I'll start:

 

Armory Gear has changed since last log, I have the cap now and new shoulders

 

Logs: 

 

Raid Comp: Healing (1 Disco, 1 Hpriest, 1 Resto Druid, 1 Resto Sham) 2 Tanks, Mix of DPS more ranged than melee.

 

Overall, heals felt stable. I feel like I could have a bit less spirit, but not much less, I will tinker with this over time. Cool down management was better coordinated and helped a lot. 

 

Top Healing Spells

Healing Rain - 30.2% of Total Healing (Entire raid night)

HST - 23.3%

Chain Heal - 11.2%

Share this post


Link to post
Share on other sites

This is a good topic, and deserves a long and studied response in terms of the Theorycraft we have.

 

I don't have time to do that today, though. :(

1 person likes this

Share this post


Link to post
Share on other sites

This is a good topic, and deserves a long and studied response in terms of the Theorycraft we have.

 

I don't have time to do that today, though. sad.png

Well, I look at this way, we're nearing the end of the expansion, may as well have some fun with some theory while we await new content. At first glance it looks like theory crafting won't be nearly as exhaustive in WoD with a lot of the math being taken off the table. We'll find something to do with the lower amounts of data I'm sure.

Share this post


Link to post
Share on other sites

Well, I look at this way, we're nearing the end of the expansion, may as well have some fun with some theory while we await new content. At first glance it looks like theory crafting won't be nearly as exhaustive in WoD with a lot of the math being taken off the table. We'll find something to do with the lower amounts of data I'm sure.

 

I can say for certain that having the theorycraft right BEFORE you gear is going to be more important than ever.

Share this post


Link to post
Share on other sites

I can say for certain that having the theorycraft right BEFORE you gear is going to be more important than ever.

Of course, and I'll be in the beta/PTR as soon as I can get in there to get some evaluation done. I do it every expac. Then it'll be a bit easier to have some weights to work with.

 

For this xpac, I was kind of tossed into 25 man heroic content without reading the owner's manual. I'm still figuring out 25 man stats and it's been a blast.

Share this post


Link to post
Share on other sites

The change to 20-man raiding will be good, too. It will mean an end to the annoying duality of analysis that Resto Shaman have had to do every tier since forever. :)

Share this post


Link to post
Share on other sites

Updating this thread:

 

I am currently still using the 9100 haste mark: (still trying to figure out if the HST is clipping or not)

 

Currently raid buffing to 38.6% crit, though I am getting a buff that puts me up to 43% and haven't identified what the buff is yet. Probably being granted by a class I'm not versed in. (any ideas?) Spirit is sitting pretty at 12.4k

 

Log from Dark Shaman progression kill

 

Mana gains were good with ~480k mana gained during the encounter from resurgence. I never felt like I was squeezing mana out of each cast and I don't believe I used mana tide.

 

Cooldowns were used as needed to prepare for a spike that we weren't ready for.

 

Due to the learning curve on mechanics, I feel that my throughput will increase as I become more relaxed and comfortable with spell selection.

Share this post


Link to post
Share on other sites

I am currently still using the 9100 haste mark: (still trying to figure out if the HST is clipping or not)

 

What the hell is the 9100 mark?

 

I see this thrown about a lot, but I can't find a good reason to use it anywhere. Checking Icy Veins' Haste Breakpoints section, there doesn't appear to be a 9100 value. Nor on Ask Mr Robot...

Share this post


Link to post
Share on other sites

What the hell is the 9100 mark?

 

I see this thrown about a lot, but I can't find a good reason to use it anywhere. Checking Icy Veins' Haste Breakpoints section, there doesn't appear to be a 9100 value. Nor on Ask Mr Robot...

The 9100 mark is a "comfortable" measure above the extra HST tick. I know you don't agree with totem break points but many do and I haven't made up my mind either way just yet.

Share this post


Link to post
Share on other sites

The 9100 mark is a "comfortable" measure above the extra HST tick. I know you don't agree with totem break points but many do and I haven't made up my mind either way just yet.

 

So that's where this comes from! Thanks!

Share this post


Link to post
Share on other sites

Currently raid buffing to 38.6% crit, though I am getting a buff that puts me up to 43% and haven't identified what the buff is yet. Probably being granted by a class I'm not versed in. (any ideas?) Spirit is sitting pretty at 12.4k

Trinket proc smile.png

The crit difference sums up to about 10.6k intelligence proc.

Share this post


Link to post
Share on other sites

Trinket proc smile.png

The crit difference sums up to about 10.6k intelligence proc.

I just didn't think that would add up to like 5% crit, but I'll take it smile.png

Edited by CptDan

Share this post


Link to post
Share on other sites

So that's where this comes from! Thanks!

Yeah it was stated by many that 9100 allowed for an average latency cushion. I am not sure if WoL is detailed enough to count the ticks total vs time the totem was up. It would be nice to put this debate to rest although the controversy is kinda fun.

Share this post


Link to post
Share on other sites

I don't know about WoL, but the new site warcraft logs should let you. You can display individual ticks from healing rain so HST should be the same.

Share this post


Link to post
Share on other sites

Yeah it was stated by many that 9100 allowed for an average latency cushion. I am not sure if WoL is detailed enough to count the ticks total vs time the totem was up. It would be nice to put this debate to rest although the controversy is kinda fun.

Breaking for HST is a tricky business. Not only because of latency, but also because of recalling.

On the one hand you want to break above 8882 to always get the extra tick. On the other hand you don't want to remain very close to the breakpoint because it doesn't leave you enough time to recall the totem after the last tick.

For instance in an ideal 0 latency world, the 9100 breakpoint leaves you a "grand total" of 60 milliseconds to recall the totem. However, 7613 is quite generous and leaves you more than 1 second.

If you don't recall it, then breaking for HST is fine. Otherwise, I'm not convinced it's a good idea.

 

I intend to write a WA that tells you when it's really safe to recall HST (after all ticks actually happened). Just need to find some time to do it biggrin.png

 

 

I don't know about WoL, but the new site warcraft logs should let you. You can display individual ticks from healing rain so HST should be the same.

It does show individual ticks for HST, though I think it's best to look at fights where you don't spec into rushing streams (sha or malk).

Edited by lynx

Share this post


Link to post
Share on other sites

One thing that always bugs me about the HST breakpoint is that it's not personal lag which causes it to not tick. It's a server-side thing, and we don't entirely understand it. On the assumption that it's the server-side lag which is the cause, your Haste might not affect that. Indeed, there was evidence presented by Binkenstein that Haste did not seem to affect the number of ticks you get.

 

However, I felt at the time (and still feel) that there isn't enough statistical power in those results to really draw a good conclusion. I've wanted to do something about it since before 5.4 hit!

 

I have some code which can take the combat log for a fight and parse it for HST placement, ticks, despawns. It even counts the n.o. ticks for each one! If you're interested in getting some empirical data on this subject, send me some (text) logs. The effects of Heroism and other Haste buffs need to be taken into account, and I'm not sure how to do that yet sad.png

 

The advantage of this method is that I can compile all the data automatically and analyse it in Matlab, which is an insanely powerful tool for this kind of thing.

 

Oooh wait, even better! If you feel like really helping me, do this:

 

  1. Clear your combat log.
  2. Note your Haste % and latency in the Shrine and then start combat logging.
  3. Drop HST lots of times - minimum of 10 - and stay in range of it for all the ticks.
  4. Once you're done, stop combat logging and make a copy of the log. Send it to me with the Haste % and latency information. Also note if you have Rushing Streams specced! This is important!
  5. Repeat with different Haste % levels!

The more people who do this, the better information we get. I should be able to do some interesting plots if I get, say, 10 different Haste % levels from three different people!

Share this post


Link to post
Share on other sites

Oh crap, you guys have got me thinking about this code again. This was such a devil to write and now you've got me trying to work out how to make it better.

 

Damn you, Icy Veins!   =_=

 

(love you really)

Share this post


Link to post
Share on other sites

My rough estimate is that you can model the tick interval with normal distribution, or something which is closely related (like T). The mean should be the computed interval, and the variance will probably depend on latency.

Share this post


Link to post
Share on other sites

But it's then not clear about whether that's affected by Haste on the server side.

Share this post


Link to post
Share on other sites

You mean, whether it's a snapshot mechanic, or dynamically updated?

My money is on dynamic update, because it's a totem, so it's kind of a pet. But I could be wrong here. Anyway, it shouldn't be hard to test it.

Share this post


Link to post
Share on other sites

I will find somewhere all alone and test the totem interval. He should we send the text?

Share this post


Link to post
Share on other sites

One thing that always bugs me about the HST breakpoint is that it's not personal lag which causes it to not tick. It's a server-side thing, and we don't entirely understand it. On the assumption that it's the server-side lag which is the cause, your Haste might not affect that. Indeed, there was evidence presented by Binkenstein that Haste did not seem to affect the number of ticks you get.

 

I think I didn't read this right the first time tongue.png

The HST ticks are definitely affected by haste, and the interval varies around the computed interval (sometimes the ticks are closer and sometimes are farther apart).

 

I just ran a few tests on some older logs and this is what I found:

1. Tick interval scales dynamically with haste (it doesn't snapshot)

2. Normal distribution is somewhat ok to model the tick interval. The mean is usually in the vicinity of 50ms of the computed value, which is nothing spectacular but works, and the standard deviation seems to be ~= latency.

3. Sometimes there's a delay of up to 500ms (!!) between totem cast and the first tick.

 

Overall, the 9100 breakpoint leaves very little room for error, and if your latency is above 100ms then expect last tick to be cut sometimes.

Edited by lynx
1 person likes this

Share this post


Link to post
Share on other sites

I just ran a few tests on some older logs and this is what I found:

1. Tick interval scales dynamically with haste (it doesn't snapshot)

2. Normal distribution is somewhat ok to model the tick interval. The mean is usually in the vicinity of 50ms of the computed value, which is nothing spectacular but works, and the standard deviation seems to be ~= latency.

3. Sometimes there's a delay of up to 500ms (!!) between totem cast and the first tick.

 

No doubt that the ticks are affected by Haste, that was well established in Beta. The problem is this delay you talk about in 3. This supposedly applies equally well to the entire duration, which is meant to cause the problem. Read Binkenstein's commentary here.

Share this post


Link to post
Share on other sites

Thanks for the reference.

Hmm, I see what you mean. I'll gather some more information later on, and check whether this delay applies at the end as well.

UPDATE:

I checked a few more logs, and this is what I found:

1. Intervals between ticks don't vary as much as the time until first tick. With ~150ms latency I didn't see deviation of more than 200ms from the computed tick interval. Some intervals are shorter and some are lengthier than the computed interval, with the average not deviating by more than 50ms.

2. The totem is extended in the end. The biggest extend I saw was 300ms, but most of them were no more than 200ms. This somewhat compensates for the delay of the initial tick.

This data is a bit raw, and I'm still not sure what to do with it. Probably can make some confidence intervals for breakpoints with latencies in mind.

Edited by lynx

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now
Sign in to follow this  
Followers 0

  • Recently Browsing   0 members

    No registered users viewing this page.