Shadow DPS Combatlog Analysis

[Worshaka]

I'm only really interested in replies from those who analyse their combat logs and have a fair indication on the specifics of their dps rotation.

Under conditions of self buffs only, how often are you getting MF ticks? ie. I get a MF tick every 1.5 sec on average and I wonder if this is about right or whether i'm having clipping issues due to Aussie ping. I know I've seen combat logs with MF ticks every 1.2 seconds and even lower but those involved high uptimes of Bloodlust/Heroism and were in a raid environment. Considering most raid encounters somewhat corrupt the 'perfect' rotation I was interested in comparisons on a training dummy with self buffs only.

[harky]

I don't play Shadow, but I have looked at a lot of parses. Most gravitate towards a tick every 1.5-1.7 seconds. I too have seen a few that are 1.5 or less, but they typically display a disparity between MB casts and MB cooldowns available, which means they're not clipping MF properly. I have noticed a few lower, but the players in question had very large amounts of Haste (650+ haste rating).

It's not too hard to figure out your absolute maximum ticks per second by checking how many possible ticks you have per MB CD. With say 15% haste you get a tick every .869 seconds, which is 6.329 ticks per CD. Then you need to look at a MB cast where you cast VT, or DP. Either takes 1.3 seconds, leaving 4.8 ticks. If you need to refresh both then you're looking at 3.337 ticks. Due to how the cooldowns work you'll never see the 'remainders', so you're looking at 6/4/3. VT will interrupt 8 times every 2 minutes, DP will interrupt 5 times every 2 minutes. There are 17.6 cycles in those two minutes. They'll most likely overlap 3 times every 2 minutes. That leaves 2 '3 tick' cycles, 10 '4 tick' cycles and 5.6 '6 tick' cycles. That leaves you at 79.6, but for sake of argument we should probably drop the .6. That leaves 79 ticks in a 120 second period. That's an average of 1.518 seconds per tick.

The largest gain in ticks would be from hitting a Haste percentage that allows for 5 ticks on a cycle where only DP, or VT are refreshed. To do that you'd need to reach 17% haste. Doing so would not give you extra ticks when you don't need to refresh a DoT, or both DoTs, but it would give 10 more ticks every 120 seconds which would increase you up to a tick every 1.34 seconds. The next jump would be at 27% haste, which would give you an additional tick when no DoT needed to be refreshed and bring you down to 1.25 seconds per tick.

Hypothetically, anyway.

[Worshaka]

Nice work, never thought to approach it that way... I generally run between 1.5 to 1.7 at absolute worst, so I seem to be getting somewhat close to the max possible MF ticks. MB cast time is usually 7.2 seconds per cast and my DOT uptimes are 95% plus... just wanted to make sure my MF was where it should be.

[harky]

I think what you're getting at is really the heart of the problem. That is, Mind Flay. Blizzard seems proud that 'priority systems' are being considered the norm now, but it's hurting the game. Classes that don't have priority systems are pulling more and more ahead because Blizzard's analysis is spot on for them. If you consider a rotation where you never had to clip Mind Flay, or delay MB, while having 90%+ DP/VT uptime, Shadow would be doing fine. That rotation doesn't exist.

Most 'priority' systems now are this: A set rotation while managing procs. Paladin and Priest instead have this: A big list of stuff to hopefully do in the right order. Even Paladin however doesn't need to forcefully interrupt its own casts to do 'good' damage. Looking at cooldown clashes they get every minute. Okay, now think about this: Shadow was one clash every 5-7 seconds unless you're at very precise haste amounts.

The best bet as far as I know is to make sure you're maximizing your MB use, while not interrupting MF just before a tick. Good luck if you have any haste on your gear though. Hopefully 4p T9 will help, but to be honest MF cast time should most likely be 2.5 by default.

[Deleted]

Given a fixed haste rating you can count the optimal average time between MF ticks. Lower is not always better in this case as too much MF ticks means you are doing something else wrong. The same goes for dot uptime and MB cast interval, 100% and 5.5sec are most likely not the optimal numbers, because it would probably require cutting MFs in bad places, the rotation is all about making the best compromises.

The way harky did it is one way to estimate the optimal number of MF ticks, but what you must take in consideration are the assumptions made in that model. Since the MB CD is used to calculate number of ticks and all the numbers are rounded down, it assumes that you are casting MB every 5.5 sec, not matter what. Even the 4.8 tick cycle is rounded to 4 ticks and not 5 ticks, which would be the optimal choice obviously. Everything that is rounded down is time spend waiting, it could be more beneficial to round number of ticks up, which means more time between MBs but less time doing nothing. It's would be more realistic too. Dot uptime is also assumed to be 100%, but I guess that's a smaller issue.

I could be more realistic to assume things like you said, 90% dot uptime and MB every 7.5s, because you probably came up with these numbers trying to make the best possible compromise with MF clipping.

So, assuming you cast VT 0.9*60s/21s = 2,57 times/min, DP 0.9*60s/24s = 2.25 times/min and MB 60s/7.5 = 8 times/min we can calculate that you should in theory have
60s-1.34s*(2.57+2.25+8) = 42,82s time left per min for casting other stuff, in this case MF. I have assumed 12% unbuffed haste here, so that makes GCD 1.34s. In 42.82s you have time for
42.82s/(1s/1.12) = 47.96 MF ticks, which is 1.25sec between ticks.

Ofc this is optimal with no MF lag, perfect clipping, etc.. But I think the major limitations in the rotation are taken into consideration better here by assuming more realistic numbers. Less MB and dot casting means more MF and that why the number is lower.

[harky]

Yes, my assumptions were for the maximum possible MF ticks while keeping MB on cooldown, not for maximum possible DPS. It doesn't account for things like delaying MB by a quarter of a second to allow an extra MF tick, etc. It certainly gives a baseline, but with normal raid haste you're looking at 20-25% and there will be a lot of portions where you're .1-.5 off on cycles. There are far too many variables to just napkin math how many ticks you should expect. GCD, tick time and cast time all need to be looked at... it's not fun. The only easy to analyze portion is GCD clashing, which makes casting MF for only one tick a poor cast. For instance at 25% haste the GCD is 1.2 seconds, but tick time is .8, so if you cast within .8 seconds of MB, you get a tick on CD, but delay cast by an extra .4 seconds, which is another .4 seconds off of getting a second tick.

Migraine inducing.

[Worshaka]

Given a fixed haste rating you can count the optimal average time between MF ticks. Lower is not always better in this case as too much MF ticks means you are doing something else wrong.

This is true, however in the greater context of analysing all my spells, if MB is being cast every 7.2 sec or lower, dots are up 95% of the time or greater, than obviously the lower the average MF tick there is the better... your statement is only true if you find out for example that MB is being cast every 8 seconds or greater.

[Worshaka]

I think what you're getting at is really the heart of the problem. That is, Mind Flay. Blizzard seems proud that 'priority systems' are being considered the norm now, but it's hurting the game. Classes that don't have priority systems are pulling more and more ahead because Blizzard's analysis is spot on for them. If you consider a rotation where you never had to clip Mind Flay, or delay MB, while having 90%+ DP/VT uptime, Shadow would be doing fine. That rotation doesn't exist.

Most 'priority' systems now are this: A set rotation while managing procs. Paladin and Priest instead have this: A big list of stuff to hopefully do in the right order. Even Paladin however doesn't need to forcefully interrupt its own casts to do 'good' damage. Looking at cooldown clashes they get every minute. Okay, now think about this: Shadow was one clash every 5-7 seconds unless you're at very precise haste amounts.

The best bet as far as I know is to make sure you're maximizing your MB use, while not interrupting MF just before a tick. Good luck if you have any haste on your gear though. Hopefully 4p T9 will help, but to be honest MF cast time should most likely be 2.5 by default.

I'm not sure I completely agree here... I think the main problem is the structure of shadow dps makes it difficult to benefit from haste. You always benefit from haste but the benefit isn't linear, it has very small returns then massive returns at specific sweet spots.

There are plenty of combat lots out there they show people casting MB at 7 seconds and maintaining nearly 100% dot uptime, that really contradicts what you're saying. I also think that 2pc T9 solves most of the conflict issues I was having, only DP and MB tend to conflict now and that isn't very often.

I also think the structure of shadow dps is highly contingent on lag, I have to put up with Aussie ping and maintaining perfect dot applications is extremely difficult.

[Deyus]

There are plenty of combat lots out there they show people casting MB at 7 seconds and maintaining nearly 100% dot uptime, that really contradicts what you're saying. I also think that 2pc T9 solves most of the conflict issues I was having, only DP and MB tend to conflict now and that isn't very often.

Are you referring to a complete fight or a segment? I think those numbers would be almost impossible to achieve over a complete fight (from pull until the boss is dead) simply because of the setup time following the pull. I routinely review the logs from my guild (where two of us are raiding shadow priests - with one offspec healer as shadow from time to time) and also from other guilds who use WoL (to check our performance) and have yet to see a parse with sub 7.2 sec MB's or +95% dot uptime.

Some numbers from Koralon last night - this was the only boss we killed who didn't involve multi-dotting or substantial target switching. The fight duration was 2m 36sec:

MF - 1.21 sec
MB - 7.42 sec
VT - 93.3%
DP - 91.0%
SWP - 91.7%

I currently have 638 haste on my gear, and for this fight we had the caster haste totem (5%) but not the 3% from a boomkin / ret. We also had a full duration bloodlust.

[harky]

I'm not sure I completely agree here... I think the main problem is the structure of shadow dps makes it difficult to benefit from haste. You always benefit from haste but the benefit isn't linear, it has very small returns then massive returns at specific sweet spots.

There are plenty of combat lots out there they show people casting MB at 7 seconds and maintaining nearly 100% dot uptime, that really contradicts what you're saying. I also think that 2pc T9 solves most of the conflict issues I was having, only DP and MB tend to conflict now and that isn't very often.

I think you missed my point there. It wasn't that ~100% uptime on dots and hitting all MB cooldowns is impossible, it's the impact that doing so has on clipping MF poorly. Unless you're at some very, very precises haste figures it's not something that's possible. It's not about the DoTs, or MB so much as it is the combination of having to understand the GCD mid-channel, the tick time and the cast time of MF. What I was referring to is MF's effective ticks per cast: Ticks / Casts. There's a huge amount of spell interruption required to be successful as Shadow.

[Worshaka]

Are you referring to a complete fight or a segment? I think those numbers would be almost impossible to achieve over a complete fight (from pull until the boss is dead) simply because of the setup time following the pull. I routinely review the logs from my guild (where two of us are raiding shadow priests - with one offspec healer as shadow from time to time) and also from other guilds who use WoL (to check our performance) and have yet to see a parse with sub 7.2 sec MB's or +95% dot uptime.

Some numbers from Koralon last night - this was the only boss we killed who didn't involve multi-dotting or substantial target switching. The fight duration was 2m 36sec:

MF - 1.21 sec
MB - 7.42 sec
VT - 93.3%
DP - 91.0%
SWP - 91.7%

I currently have 638 haste on my gear, and for this fight we had the caster haste totem (5%) but not the 3% from a boomkin / ret. We also had a full duration bloodlust.

Complete fights from the spriests that are doing the best dps seen in combat logs... ie the spriets that were pushing 5.5 to 6K on patchwerk back in Naxx25 gear... if you go to World of Logs you can basically search through the logs of the best recorded kills for every encounter.

[Deyus]

Complete fights from the spriests that are doing the best dps seen in combat logs... ie the spriets that were pushing 5.5 to 6K on patchwerk back in Naxx25 gear... if you go to World of Logs you can basically search through the logs of the best recorded kills for every encounter.

Understood on how to find the top parses for each boss. My point is that I have never seen a parse with both sub 7.2 sec MB's and +95% VT/DP uptime. I suspect it's a case of exaggeration, but if you know of a wealth of these parses please point me to them - I'd like to see what these guys are doing! In my opinion, the tradeoffs that you have to make between DoT uptime, casting MB on cool down, and not clipping MF at awkward times makes it next to impossible to achieve either.

Using Patchwerk as an example:

The top parse on WoL has 92% VT uptime and 7.72 sec MB's, the 3rd place parse has 97.2% VT uptime (97% on WoL - but the number of ticks only corresponds to 90% - he was clipping the last tick of his DoT) but only 9.95 sec MB's, and the 7th place parse has 93% VT uptime and 8.4 sec MB's.

Fights as short as Patchwerk can be skewed by BL (you might cast MF with a higher priority), or might end with MB at a bad place in the cycle, but longer fights also have more chances for error. If you are aware of any fights where exceptional priests are achieving 95%+ DoT uptime and sub 7.2 sec MB's please show us!

[Worshaka]

I didn't bother to get an archive of them, but I have spent time in the past studying combat logs of others, particularly those where spriests were achieving considerable dps figures. If you goto WoL and check some of the more tank n spank type encounters and look through the fastest kills, you'll find plenty of spriets maintaining 7sec MBs with close to 100% dot uptimes.

There is also logs posted on EJs, mind you most of those are by bad players but sometimes people post some good logs as examples.

[Deyus]

Ahh.. the old "it's there because I say so" claim. I disagree, and despite your many contributions to this community, would ask that you try to remain truthful and avoid exaggerating in your posts! I'll be happy to retract that statement if, as stated before, you can produce even one of the many parses that you claim exist.

I'm sure you're a good priest, and I'm also sure that the many who produce top parses are also exceptional. However, it is not possible to achieve 7 sec MB's and dot uptime greater than 95%. 'Tank-and-Spank' type fights might include things like Patchwerk, Kolagarn and possibly XT / Ignis. If you search the logs for those fights you will find that there are no parses that have >95% dot uptime (where the player is not putting DoT's on adds) and 7 sec MB's. The very best players you will be able to find are hovering around 7.2 sec MB's and around 93% DoT uptime.

What's more, I would argue that ensuring you have 7 sec MB's / >95% uptime is not desirable from the perspective of maximizing your DPS. If your DoT uptime is greater than 95% chances are you are occasionally clipping the final tick of the spell (the ONLY parses I've ever seen with >95% have either been multi-dotting or clipping), and if your MB is at 7 sec you are likely waiting for MB to come off it's cool down in order to time things perfectly (also not desirable).

[Worshaka]

Ahh.. the old "it's there because I say so" claim. I disagree, and despite your many contributions to this community, would ask that you try to remain truthful and avoid exaggerating in your posts! I'll be happy to retract that statement if, as stated before, you can produce even one of the many parses that you claim exist.

I'm sure you're a good priest, and I'm also sure that the many who produce top parses are also exceptional. However, it is not possible to achieve 7 sec MB's and dot uptime greater than 95%. 'Tank-and-Spank' type fights might include things like Patchwerk, Kolagarn and possibly XT / Ignis. If you search the logs for those fights you will find that there are no parses that have >95% dot uptime (where the player is not putting DoT's on adds) and 7 sec MB's. The very best players you will be able to find are hovering around 7.2 sec MB's and around 93% DoT uptime.

What's more, I would argue that ensuring you have 7 sec MB's / >95% uptime is not desirable from the perspective of maximizing your DPS. If your DoT uptime is greater than 95% chances are you are occasionally clipping the final tick of the spell (the ONLY parses I've ever seen with >95% have either been multi-dotting or clipping), and if your MB is at 7 sec you are likely waiting for MB to come off it's cool down in order to time things perfectly (also not desirable).

Its not eggageration, I created a spreadsheet to analyse many of them and I did so a long time ago... when I have time i'll post 1 or 2 from WoL... or you can go there and check for yourself.

And just to turn your argument on you, just because you say it doesn't exist doesnt mean it doesn't.

[Deyus]

As stated before, I have taken the time to check and can't find any which show the MB/DoT uptime numbers you claim amongst the top parses. Eagerly awaiting your post with links!

Please don't disappoint!

[Worshaka]

A lot of top Patchwerk logs are pretty corrupt now with bad people in superior gear... however here is a log of a spriest who in their 118 seconds of activity managed an average MB every 6.94 seconds and had a perfect 39 VT ticks and was 38 ticks with SW:P & DP and also managed a MF tick every 1.45 seconds...

When LK first shipped and Patch was the benchmark for combat log analysis I did countless research on some of the top spriests going around and there were lots of logs like this. You can either believe me or not, it doesn't really matter, if you dig through the logs at WoL deep enough they are there.

[Worshaka]

This isn't quite as good but they manged a MB every 7.25 sec 37 VT dots out of a max 39, 39 SW:P & 36 out of 39 DP.

Again I was talking about the absolute best logs going around at the start of LK on tank n spank encounters (ie patchwerk)... it is possible, perhaps leave room for the fact there are better spriets out there than yourself?

[Deyus]

A lot of top Patchwerk logs are pretty corrupt now with bad people in superior gear... however here is a log of a spriest who in their 118 seconds of activity managed an average MB every 6.94 seconds and had a perfect 39 VT ticks and was 38 ticks with SW:P & DP and also managed a MF tick every 1.45 seconds...

Thanks!

The first log you linked is obviously an exceptional priest, who hasn't clipped any of his DoT's and has really pushed the limits. However you may want to review your math (unless of course my calculator is broken...):

Fight duration: 2:02 = 122 sec
122/3 = 40 possible DoT ticks
MB = 17 (122 sec/ 17 = 7.18)
VT = 96.9% by WoL's debuff tracker (39/40 = 97.5% of possible ticks)
DP = 94.7% by WoL's debuff tracker (38/40 = 95% of possible ticks)
SWP = 94.1% by WoL's debuff tracker (38/40 = 95% of possible ticks)
MF = 81 (122 sec / 81 = 1.51)

You'll notice this is essentially the 7.2 sec / 95% practical limit I suggested yesterday (this priest is 7.18 / 95.2%). Regarding the fight duration of 122 sec, you very clearly stated that you were referring to the complete duration from the pull until the boss dies. As I noted yesterday, it may be possible to exceed those numbers if you are only looking at a fight segment, but this is not possible once you include the setup time on short fights (the 4 sec this guy is missing count as setup time), or the higher probability of making a mistake on fights longer than 5 minutes.

Like you note the second parse isn't quite as good, though is still clearly top notch. The guy gets a little eager and clips each of his DoT's once towards the end of the fight:

Duration = 2:00 = 120 sec
120/3 = 40 possible DoT ticks
MB = 16 (120/16 = 7.5)
VT = 95.7% by WoL's debuff tracker (37/40 = 92.5% of possible ticks)
DP = 93.8% by WoL's debuff tracker (36/40 = 90% of possible ticks)
SWP = 97.4% by WoL's debuff tracker (97.5% of possible ticks - though it appears to have gone up with at most 2 stacks of Shadow Weaving!!)
MF = 78 (120/78 = 1.54)

This guy is at 7.5 sec / ~95.6% (though that doesn't account for his clipped DoT's).



My point through this entire discussion has been that the "7 second Mind Blast's and DoT uptime of nearly 100%" target that is continually repeated within the community is wrong and misleading. It is not obtainable in a real situation (even a tank and spank situation) when you have to consider trade-offs and management of our various spells. The logs you have linked clearly show exceptional priests in action and serve to reinforce my point. A more realistic target for those of us who are merely mortal is around 7.2 sec between MB's and 95% DoT uptime while trying to maximize your number of MF ticks. If you're exceeding this by a substantial margin you should immediately go apply to Premo / Ensidia.

Regarding my personal performance (not sure why you bring it up?) - I have never once suggested that there are not better shadow priests out there. I generally consider my DoT uptime to be slightly sub-par and have been concentrating on trying to improve this.

Back to the topic of average time between MF ticks - the best performance possible seems to be around 1.5 sec raid buffed (though obviously this is dependent on the players haste). People below 1.5 sec, such as myself, might be delaying their DoT's and could possibly see a performance boost by adjusting slightly. I know this is not the target dummy data that you were after in your original post, but I think if you are pulling 1.5 sec between ticks on a dummy you're probably looking pretty good and can safely rule out clipping issues caused by poor latency. Given that you don't have the 8% haste from raid buffs 1.5 sec may even be a little low and potentially could indicate you're delaying your DoT's slightly (similar to my issue)?

[Worshaka]

Thanks!

The first log you linked is obviously an exceptional priest, who hasn't clipped any of his DoT's and has really pushed the limits. However you may want to review your math (unless of course my calculator is broken...):

Fight duration: 2:02 = 122 sec
122/3 = 40 possible DoT ticks
MB = 17 (122 sec/ 17 = 7.18)
VT = 96.9% by WoL's debuff tracker (39/40 = 97.5% of possible ticks)
DP = 94.7% by WoL's debuff tracker (38/40 = 95% of possible ticks)
SWP = 94.1% by WoL's debuff tracker (38/40 = 95% of possible ticks)
MF = 81 (122 sec / 81 = 1.51)

You'll notice this is essentially the 7.2 sec / 95% practical limit I suggested yesterday (this priest is 7.18 / 95.2%). Regarding the fight duration of 122 sec, you very clearly stated that you were referring to the complete duration from the pull until the boss dies. As I noted yesterday, it may be possible to exceed those numbers if you are only looking at a fight segment, but this is not possible once you include the setup time on short fights (the 4 sec this guy is missing count as setup time), or the higher probability of making a mistake on fights longer than 5 minutes.

Like you note the second parse isn't quite as good, though is still clearly top notch. The guy gets a little eager and clips each of his DoT's once towards the end of the fight:

Duration = 2:00 = 120 sec
120/3 = 40 possible DoT ticks
MB = 16 (120/16 = 7.5)
VT = 95.7% by WoL's debuff tracker (37/40 = 92.5% of possible ticks)
DP = 93.8% by WoL's debuff tracker (36/40 = 90% of possible ticks)
SWP = 97.4% by WoL's debuff tracker (97.5% of possible ticks - though it appears to have gone up with at most 2 stacks of Shadow Weaving!!)
MF = 78 (120/78 = 1.54)

This guy is at 7.5 sec / ~95.6% (though that doesn't account for his clipped DoT's).



My point through this entire discussion has been that the "7 second Mind Blast's and DoT uptime of nearly 100%" target that is continually repeated within the community is wrong and misleading. It is not obtainable in a real situation (even a tank and spank situation) when you have to consider trade-offs and management of our various spells. The logs you have linked clearly show exceptional priests in action and serve to reinforce my point. A more realistic target for those of us who are merely mortal is around 7.2 sec between MB's and 95% DoT uptime while trying to maximize your number of MF ticks. If you're exceeding this by a substantial margin you should immediately go apply to Premo / Ensidia.

Regarding my personal performance (not sure why you bring it up?) - I have never once suggested that there are not better shadow priests out there. I generally consider my DoT uptime to be slightly sub-par and have been concentrating on trying to improve this.

Back to the topic of average time between MF ticks - the best performance possible seems to be around 1.5 sec raid buffed (though obviously this is dependent on the players haste). People below 1.5 sec, such as myself, might be delaying their DoT's and could possibly see a performance boost by adjusting slightly. I know this is not the target dummy data that you were after in your original post, but I think if you are pulling 1.5 sec between ticks on a dummy you're probably looking pretty good and can safely rule out clipping issues caused by poor latency. Given that you don't have the 8% haste from raid buffs 1.5 sec may even be a little low and potentially could indicate you're delaying your DoT's slightly (similar to my issue)?

You should review the log again, the spriest in question was only active for 118 seconds. You'll find that the 4sec descrepancy is due to the fact that the tank pulls Patch while it's out of range of the casters and by the time he gets into position and the first spell lands 4 seconds has passed. With this in mind this log proves that it's possible (in tank n spank) to achieve sub 7 sec MB casts with a very close adherence to 100% dot uptime, contrary to your statement.