Sorry i suck at explaining.

hedna123b14_ESO wrote: »

Amazing work, if you ever decide to release the commercial (easy to understand for non programmers) I would probably build you a monument...

Besides the training ground that ZOS for sure will provide for us to practice our rotations in a clean fashion, we should have this monument for theocrafting.

@Destyran, yes that would be possible. I'm sort of satisfied with how it works at the moment. I should have said that my next baby step would be to model an actual rotation for templars instead of my simplified rotation and in that I'll try a few different combinations and see how it works. Even if my accuracy is slightly off I hope that the results are self consistent and it would be possible to determine which is the best.

@hedna123b14_ESO , thanks for the jokes but actually I've been thinking of a web page too just to help organize stuff sometimes I have difficulty finding the right equations from my previous post >.>

@Asayre this is beautiful it could show is the perfect rotation the procs of valkyn scathing mage. Im exicted!!

@Destryan, yup I'm excited too. Maybe I'm being too optimistic but if everything works out I can stop spending time calculating and just make the computer do it!

@dpencil, I don't think a gear bonus section is particularly difficult. There are already several spreadsheet/programmes that will happily calculate your stats for any arbitrary gear set. Similarly for CP and Mundus inputs. It's not difficult to get decent estimates of damage on any mob and I was thinking of doing that as well but I didn't think it would present too many hurdles and was interested in checking the core idea of a combat simulator first. The main problem with set bonuses will be coding the unusual set bonuses like 2p Kena or 5p Scathing. I'll rely on in-game testing to check the mechanics of this unusual set bonuses and try to implement them.

Nifty2g wrote: »

but its more fun when you calculate it rather than the computer, isn't it?

NOPE.

Magicka Sorcerer
I’m using a build practically identical to The Yolo Wizard V2 - Non Overload PvE DPS Build. Please refer all build questions to @Yolo-Wizard.

Gear and CP Distribution
While the link above will provide all build relevant build details, I’ll include a snapshot of my stats and bar for completeness.

Ability Prioritisation
Spell Power Potion > Shooting Star > Liquid Lightning > Boundless Storm > Elemental Blockade > Mages’ Wrath > Crystal Fragments > Force Pulse

Mages’ Wrath only enters the priority queue when target health is below 20%. Crystal Fragments is only casted when it procs. Only one Shooting Star is used as I haven’t coded in a section to keep track of ultimate.

Simulation Results
I tested my simulation on Slimecraw in Normal Wayrest Sewers. It has the same amount of resistance as Bloodspawn in Veteran Spindleclutch but is more readily accessible and testing can be done without the aid of a tank or healer. Due to the lack of a tank and healer, Major Breach is not used thus the resistance of Slimecraw is 18200.

Simulation verification on Slimecraw
Due to the peculiarities of testing, I am unable to use Boundless Storm on Slimecraw and Slimecraw is always too close for my light attacks to be boosted by the Maelstrom staff enchantment. For this simulation, I didn’t account for Skoria and it has been removed from the in-game parse.

Damage accuracy

The table above highlights the accuracy attainable.

DPS Breakdown

The image above shows the results of 10,000 simulations. The DPS for each ability and the sum of all abilities shown is automatically binned and shown in the corresponding histogram. The top left most histogram labelled DPS is merely the sum of all the other panels. The vertical red line in each histogram corresponds to the result of one in-game trial. In all cases, a reasonable consistency is achieved. Interestingly all the histograms do not fit a normal distribution based on a chi-square test for normality

Law of Julianos and Twice-Born Star
I know this topic has been discussed in depth and the difference between the two sets is on the order of 1% in favour of Julianos. I’ve repeated the simulation above with Twice-Born Star in place of Julianos and obtained the following DPS Breakdown.


For all histograms, note that the mean for Twice-Born Star is marginally lower than Julianos confirming previous results and as expected the standard deviation is larger in all cases consistent with the larger critical modifier.

DPS Prediction on Bloodspawn
I haven’t done any comparison of in-game DPS to simulation DPS for Bloodspawn so one could consider this a prediction of sorts. The same gear (Julianos) as shown in the beginning is used. It is assumed that a tank is present to constantly provide Major Breach. No other external buffs are assumed. I ran 4 simulations for deciding on the best monster set to use. The first simulation uses no monster set and the remaining 3 use Kena, Nerieneth and Skoria. DPS histograms for each scenario is shown below and then a table summarising the results is included. In these simulations, only one Shooting Star is used as I have yet to properly include ultimate tracking.











We see that in rough accordance with my previous calculations, Kena provides an 8.6% DPS increase compared to not having a monster set. However Nerien’eth and Skoria only provide 4.8% and 4.5% DPS increase, respectively. I’m looking towards a more comprehensive handling of ability cost so that we can compare magicka efficiency with and without Kena.

Back to the top

Magicka Nightblade
My ability rotation is based on Nos's Ezmode Magblade DPS but I’m using 5 pieces of Julianos, 2 Skoria, 3 Willpower, a Lightning Maelstrom staff and two Torug swords.

Gear and CP Distribution

Ability Prioritisation
Spell Power Potion > Shooting Star > Merciless Resolve > Elemental Blockade > Twisting Path > Impale > Assassin’s Will > Crippling Grasp > Funnel Health

Impale enters the priority queue when target health is below 25%. Assassin’s Will is casted when proc, Merciless Resolve is not recasted until it expires. Only one Shooting Star is used as I haven’t coded in a section to keep track of ultimate.

Simulation Results
I tested my simulation on Slimecraw in Normal Wayrest Sewers. It has the same amount of resistance as Bloodspawn in Veteran Spindleclutch but is more readily accessible and testing can be done without the aid of a tank or healer. Due to the lack of a tank and healer, Major Breach is not used thus the resistance of Slimecraw is 18200.

Simulation verification on Slimecraw
Like the simulation verification on Slimecraw for a Magicka Sorcerer, I did not account for Skoria and its DPS contribution has been removed from the in-game parse.
Damage accuracy


DPS Breakdown


The image above shows the result of 10,000 simulations. The DPS for each ability and the sum of all abilities shown is automatically binned and shown in the corresponding histogram. The top left most histogram labelled DPS is merely the sum of all the other panels. The vertical red line in each histogram corresponds to the result of one in-game trial. In all cases except for LA, a reasonable consistency is achieved. This indicates that my LA weaving is very poor on a magicka nightblade.

DPS Prediction on Bloodspawn
I haven’t done any comparison of in-game DPS to simulation DPS for Bloodspawn so one could consider this a prediction of sorts. The same gear (Julianos) as shown in the beginning is used. It is assumed that a tank is present to constantly provide Major Breach. No other external buffs are assumed. Similar to my simulation for magicka sorcerers, I’ve included simulation with no monster set, Kena, Nerien’eth and Skoria. A table summary the simulated DPS is included at the end.




Similar to simulations for a magicka sorcerer, Kena provides the greatest increase in DPS. The DPS increase is larger for a nightblade as spell damage for a nightblade is lower than a sorcerer due to the lack of the Expert Mage passive thus the percentage contribution of Kena is larger also it appears that the uptime of Kena for a nightblade is higher (further investigation is probably needed to explain this). The DPS increase of Nerien’eth is lower as there are less non-DoT hits as Funnel Health is being used instead of Force Pulse. I’m unsure of why Skoria contribution is greater for a nightblade than a sorcerer.

Funnel Health and Force Pulse
If we ran the same simulation as above but substituting Funnel Health with Force Pulse, we will attain the following table. For brevity, I’ve decided to exclude all the histograms. This is in good agreement with @Nosferatuzod’s suggestion in Nos's Ezmode Magblade DPS. The DPS for no monster set, Kena and Skoria remain comparable when running Force Pulse or Funnel Health but significantly improvements are observed for Nerien’eth as Force Pulse gives 2 extra chances to proc a lich crystal.


When to refresh Merciless Resolve?
In the above simulations, I’ve been refreshing Merciless Resolve every 20 seconds or when it expires. This means Assassin’s Will is only cast once per 20 seconds. The thought process is that Merciless Resolve should be kept active as often as possible for the Minor Berserk buff and additional cast of Merciless Resolve within 20 seconds to proc more Assassin’s Will is a loss as the Minor Berserk buff is merely refreshed. But is this thought process valid?
A clear alternative is to refresh Merciless Resolve after Assassin’s Will is cast so that another Assassin’s Will can be cast. In this approach, we can expect to cast 3-4 Assassin’s Will every 20 seconds. But these additional Assassin’s Will comes at the cost of casting Merciless Resolve that provides no real additional benefit.
The choice between casting Merciless Resolve every 20 seconds or after every Assassin’s Will can be answered either through simulation or looking at skill coefficients. I’ll do both but begin with the latter.
In general, the wasted time casting Merciless Resolve can be used to cast another Funnel Health and in lieu of Assassin’s Will another Funnel Health would be used. This leads to the following question “Which does more damage: two Funnel Health or Merciless Resolve and Assassin’s Will ?”
The first skill coefficient for Funnel Health is 0.095 while the first skill coefficient for 0.206. Thus this method of analysis suggest that refreshing Merciless Resolve after Assassin’s Will is preferable to refreshing Merciless Resolve every 20 seconds. However, this does not account for the slight timing differences between the two options. My rough estimates suggest that casting two Funnel Health will take ~2.2 seconds (1.1 seconds each) but casting Merciless Resolve and Assassin’s Will takes ~2.4 seconds (1.2 seconds each). If we normalise the first skill coefficient to the time taken, we obtain 0.086 for the Funnel Health option and 0.085 for the Merciless Resolve and Assassin’s Will solution. This suggest that casting Merciless Resolve every 20 seconds is slightly preferred.
I also ran this through my simulator to see whether comparable results will be obtained in the DPS histograms. The simulated DPS for casting Merciless Resolve every 20 seconds performs better than casting Merciless Resolve after Assassin’s Will though the discrepancy is slight. From this histograms, I noted that the DPS of DoTs was lower when Merciless Resolve was


From the histograms above, I noted that the DPS of DoTs was lower when Merciless Resolve was casted every 20 seconds. Initially, I thought that this is due to the downtime in Minor Berserk and tried re-casting Merciless Resolve every 19 seconds. But the discrepancy was still present. At present, I have yet to figure out the difference in DPS for DoTs between the two different approaches to casting Merciless Resolve.

Back to the top

Update summary

Magicka Nightblade

1. Re-ran DPS simulations on Bloodspawn at a lower number of simulations (1000) for no monster helm, Kena, Skoria and Nerieneth.
2. Ran simulations with using Force Pulse in place of Funnel Health
3. Ran simulations with casting Merciless Resolve every 20 seconds and casting Merciless Resolve after Assassin’s Will

Protip - OP is awesome! I would kill for a breakdown for every DPS class, magicka and stamina, but that is obviously a lot of work. What you have here is extremely helpful and keep up the good work!

@zerosingularity
That's the rough plan. I'm doing classes that I have first so that I can check that everything is working as expected.

@Xvorg

My light attacks usually account for around 10-15% so you'll notice an increase of about 1% with a flame staff and being a Dunmer. I'm not sure if the 12% extra damage for heavy attacks with a flame staff is additive or multiplicative with the Dunmer bonus but it should be around 19%.