Warlock Spreadsheet (3.3.5)
I'm not going to make a huge post and clutter things up, just going to get straight to the point. This is a spreadsheet for warlock DPS that is ready for use in 3.3.0, the link is below. This is the first version release of the spreadsheet here and I'm sure it has some bugs or possibly some inaccurate formulas. Please leave feedback here.
We need feedback on any bugs / errors that appear.
We need feedback on the numbers shown to you
We need feedback on some of the formulas used, such as molten core
There is still a lot of work that needs to be done, we have a list of things to do. The Gear_Buffs page will slowly be remodeled and incorporate more pictures making it more visually appealing. We're going to be cleaning up the sheets to consolidate some of the sheets and optimize things.
Microsoft Excel 2007
9/1/2010 - Build 0076
8/31/2010 - Build 0075
8/31/2010 - Build 0073
4/15/2010 - Build 0017
4/15/2010 - Build 0006
2/15/2010 - Build 0005
1/31/2010 - Build 0004
1/26/2010 - Build 0003
1/26/2010 - Build 0002
1/25/2010 - Build 0001
Any kind of donations would be great, this is not an obligation. The majority of the donations will be going toward the servers I own that will host the DPS Web App that is be built.
DropBox - This file is being hosted out of my DropBox.
My Personal Website
I guess the payoff for working out exacting models for Molten Core is that we can determine what behaviour we should be exhibiting when we recieve one. Is it better to instantly cast your 3 Incinerates and delay other abilities that may need refreshing until afterwards? Is there any advantage to saving a MC proc at 37% to use on Soulfires?
Your simplified model makes sense for understanding the underlying concerns, however in footnote 3 you talk about a simcraft constant you found based on that model (I think, maybe you found it based on "real" simulations, either way), you can't then use that constant when expanding to a more complex example, it will greatly depend on the exact priority list given. I'm not sure how you intend to calculate the constant but I think you may find it extremely difficult and it will depend very heavily on haste and the reaction time and latency of the player.
Perhaps it would be easier to start at the other end, the Soulfire Molten Core doesn't require the player to react, doesn't change cast time, it's just 18% damage and 15% crit, one is just scaling and the other will depend on your standard crit rate, which isn't used anywhere else and could be left as an unknown in any final solution.
Given that any results of this work are likely to be folded back into Simulationcraft, I'm not convinced such a narrow topic deserves its own thread. I'm willing to take arguments for keeping it separate, so please present them. Otherwise I'll merge this into the Simcraft thread and let you guys discuss it there. It's not like it would be lost amid the lively three-post-per-week discussion.
It's easy to run a simulation since you can actually roll the dice on every scenario possible and compile the data. That doesn't help you determine which piece is better than which and by how much in a timely manner. The purpose of this is to find the formula that calculates the DPS increase from talented Molten Core. This merit's it's own discussion since Molten Core has undergone major changes since the Warlock Spreadsheet was updated here on the EJ Forums. Leulier's spreadsheet is horribly outdated and it was given an update in 3.2.0 by Marklar and others with varying degrees of success. Fearsom and I have been working on updating this spreadsheet for the 3.3.0a content. Molten Core is something that was giving us much trouble and we needed an accurate account of it's damage increase. If this doesn't merit it's own topic then by all means close it and we'll continue working on it ourselves at a slower pace. Fearsom wants to get it out for other people to use quickly, to help detect any errors in the formulas that Fearsom has derived. I'd like to start a new thread when Fearsom deems the spreadsheet ready for public use at which time this thread could be merged into that new one. Thoughts?
[Edit] Main reason I started this own topic is because I didn't want it to get buried and looked over by the pages upon pages of other information that would ultimately slow development of the new 3.3.0a spreadsheet.
If you'd like to talk with me about it Gilliam contact me on Mahida or Cerula anytime you guys aren't raiding.
I spoke with a few other people about my calculations above and for the most part, people have agreed with the method I described. The number for uptime and benefit derived from SimulationCraft will most likely need tweaking but they provide a stable value that we can use for the time being. On the 27% uptime of Molten Core, you benefit from only 82% of that 27% according to the rough numbers. This can tell use just how many times you will use Incinerate compared to Shadow Bolt in the 35%+ section and then provide the numbers of Soul Fire in execute range.
Merging with a future spreadsheet thread seems reasonable, or we may let the thread die off once it's served its purpose. I can see the benefit of attracting new readers to help with analysis. Carry on.
The problem with your approach is mostly that it's not very mathy. There are *some* things that you can not calculate, for example player reactionspeed, this "variable" is going to be different depending on player skill aswell as the complexity of the encounter they're currently faced with.
Past reaction time however you can calculate the likelyhood that a new molten core proc would overlap an old one- the problem with calculating this is that it would be an incredibly large formula thats probably in parts still based on player/encounter specifics, haste values and other assumptions.
The alternative as you suggest is taking simulationcraft's results and feeding those into your spreadsheet. Frankly while I think it's a noble cause to want to create a spreadsheet- I'm not really sure if there's a real merit to having a spreadsheet based on results provided by a simulator.
Past that I'm not really sure what you're asking with this thread, are you asking wether or not you can trust the simulationcrafted results, or are you asking how to best calculate molten core clipping, or are you asking if your method of calculating the benefit from molten core is correct assuming that the number of molten core clipping is [x]%?
If we assume simcraft can successfully model the standard 0/56/15 raid spec(I think most poeple think it models/results are accurate). Simulate the 0/56/15 spec with a large number of different gear sets(100 for a round number) with different ilvls of gear. Plot the results and try and calculate a line of best fit with stats being the variables. Then you could then remove 1 point in MC each time and redo the simcraft and line of best fit to get a formula for the dps with 2/3 mc, 1/3 mc and 0/3 mc for the standard 0/56/15 spec. The biggest issue i think is how many other talents will effect the dps again from MC (ie emberstorm from a 41/30 spec, Meta , etc...). Making 1 large formula to include how many points in MC and other talents might get kinda out of hand. What do you think?
Edit: if anyone actual tries to do something like this, i would suggest writing a script to import multiple sets of gear into simcraft and run it.
Please note that the "old" way of specifying gear levels still exists:
It is not necessary to use the slot=description parameters.
This is handy when using the CLI for quick calculations of various data points.
In attempt to make it simple, you can remove haste from the equation.
It affects how often Molten Core will procc, but just as much how fast it is consumed. Since the results from formula will be a percentage, haste wont affect it.
I also dont see a way formula can be created that would count in the immolate uptime, we need to ignore it or run simulation.
Asuming perfect play, MC can refresh after we used 2charges, or after we used all 3, not after 1. that makes couting much simplier. (for 3/3 and 2/3 only)
Uptime of 3/3 MC will be 21% (chance to procc for 3*incinerate cast time) * (1-0.12*1/3) (chance to lose 1procc)= 20,16%
(still asuming perfect usage of MC at cost of everything else)
79% of fillers will be Shadowbolt 20.16% incinerates. Thats constant independant of haste, if we use glyph of Quick Decay.
Now we need Damage for SB and Incinerate
20.16%*Incinerate+79%*SB for 3/3 and compare to 100% SB spam.
value of 3/3 MC will be the difference in DPS, counting 2/1 MC value can be simplified just count it as 15% 35% 100%
2/3 MC will have 33% less proccs, and incinerate will be 5% weaker AND 10% slower. at 1/3 incinerate wont even be worth casting over SB AND will have about 10% uptime, making it very poor talent (as its ^3 build).
Hope that helped.
fearsom informed of this situation:
assuming corruption is already on the boss
0.0 cast shadow bolt
0.5 mc procs
2.5 shadowbolt finish casting, you start casting incinerate
3.5 mc procs again
MC proced again before you even finished casting 1 incinerate. I don't believe simcraft profiles or any players for that matter would stop casting a shadowbolt because MC proced. Their might be so magical number where if you are else the x seconds(maybe <.5s) into you shadow bolt where it might be a small dps again to stop casting it but i am not sure.
Even though it can proc again, the chances that it does before you use all 3 charges is small. The average proc interval seems to be around 20-25 seconds per MC. The average time to cast 3 incinerates is less than 4.4 seconds, so getting a 2nd proc within that time is possible, but not highly likely.
In the real world you're going to clip molten core procs. In the real world it will happen atleast once or twice every fight. It has fairly little to do with player skill- and frankly the math to prove that it will happen is very simple. The math provided to show how much it happens is pretty poor.
2 formulas that at a very low level of accuracy could tell you something:
Likelyhood of molten core to proc again within [x]s with [y] haste rating
Example 0 haste, 3 seconds, 0 haste 6 seconds and 1083 haste and 8 seconds:
Technically you should account for raid haste buffs (5% and 3%) and it would look like:
Calculating the average time to consume molten core:
Average time consumed by refreshing immolate during proc: 1.5/1.[haste]/24*(1.5/1.[haste])
Average chance that you'd have to refresh curse of doom: 1.5/1.[haste]/60*(1.5/1.[haste])
Average chance that you'd have to refresh corruption: 1.5/15 (haste can be ignored since it would show up equally on both sides)
I think it's reasonable to say that you don't need to refresh glyph of lifetap, you'll typically lifetap about 20 seconds before it expires and so long as you keep your mana above 50% you also shouldn't need to lifetap for mana reasons.
Average cast time of incinerate: 7.5*0.7/1.[haste]
Average time left on current cast will be a bit trickier, but 1s is probably a fair estimate.
Reaction time (if you want to factor it in) 0.5s
So @ 600 haste: 1+0.5+(7.5*0.7/1.2=4.375)+(1.5/1.2/24*(1.5/1.2)=0.065)+(1.5/1.2/60*(1.5/1.2)=0.026)+(1.5/15*1.5=0.15)=6.116
These are ballpark formulas, they dont account for having to refresh immolate/corruption after your first incinerate, and a bunch of other stuff, but you should be able to use them to gouge roughly how likely it is that procs will overlap.
I think we agreed, that we play 'perfectly' for the calculation.
Therefore if MC proccs while we started SB cast should be either ignored, or if ppl realy want:
20.16% uptime will be reduced to 18.4% uptime.
calculating player lag and refreshing corruption/immo/CoD in formula is pointles, simulations are much better at that. unless u want longer formulas than excell can handle.
And once again, kick haste from the equations. It affect ticks same % as our cast time (eating procc), so only reason to include haste would be calculating CoD and Immolate effect. that will be VERY insignificant, in ranges of 0.05% dps+-
|All times are GMT -4. The time now is 7:17 AM.|
Forum Infrastructure by vBulletin 3.6.12 ©2000-2007, Jelsoft Enterprises Ltd.