Before 7.2 came out, I wanted to share some examples of tests I do to make sure our simulator is accurate. This is the kind of testing we will continue to do after 7.2 launches and again with Tomb of Sargeras.
Meanwhile, for Nighthold, you might have seen that we have actual NH bosses in the simulator, which takes theorycraft and gearing advice to the next level. Why? In past expansions, advice has been based on generic fights that don’t exist in the game.
Generic fight sims limit the ability to optimize rotations, compare trinkets, relics or talent choices. You’re missing a big piece of the puzzle – the actual fight. So when we built our simulator, we set out to recreate the actual bosses and have the whole simulator (rotations and bosses) match combat logs from top players.
I wanted to determine if we were able to simulate so accurately, that we matched logs.
TLDR: YES! I started with DPS and reviewed dozens of logs for each spec on Krosus and share a few examples below. There are some special notes for a handful of specs, so be sure to read the specifics for the class(es) you play. If you have questions, head over to this thread on our forums.
With simulations that match actual logs so closely, you have a lot of really useful information at your fingertips to make in-game decisions. And by providing specific boss fights, we add an extra layer of accountability to ensure the simulations are accurate, and eliminate the excuse, “Well of course the logs are different, you can’t sim a real fight.” As we’ve proven, yes you can 🙂
Since the boss-specific scripts proved to be such a success, I’m going to make sure they are at the top of my to-do list for Tomb of Sargeras. (You can also see use our Spellblade script for NH, explained here, for DPS and tanks.)
Of course – things are changing with 7.2 tomorrow! We’ll be updating the simulator and rotations to go with it. It will take a couple of weeks of combing over logs again to find and fix any bugs, but hang in there! As you can see, we are dedicated to making sure the simulator reflects reality.
What kind of Krosus logs did I look at?
For the purpose of this analysis, I wanted to find high-skilled players, but eliminate the super lucky (and unlucky) attempts. I also wanted logs of teams that weren’t over-gearing and face-rolling the boss. So I looked at heroic players with 890-892 gear, a +45-49 artifact, in the 90th-95th percentile. If a player falls outside of these rules, I note it and explain why.
- 890-892 iLevel is close to the average iLevel of people fighting Krosus without dying too often or crushing the boss.
- Logs from the 90th-95th percentile tend to be the higher skilled players, but it removes the super-lucky streaks that you find in the 99th percentile.
- I tried to find players with +45-49 artifact points. Why? Artifact levels aren’t taken into account in the percentile rankings, so a player with 6 more artifact points does 3% more damage (well, pre 7.2 changes at least).
I use 4 different version of the Krosus script on our simulator:
The default Krosus script is explained in this blog post. The wonderful thing about our simulator is that it is so customizable. I noticed a few different styles of play when reviewing logs, so I made some tweaks to the boss scripts to offer other playstyle options.
Nearly everything about the customizations for the simulator is very user-friendly, however, the boss script interface needs some love. Since it’s so raw, I’ve shared some of the different boss versions below, hoping to cover a lot of the common cases I see in logs.
- The default script TLDR: Ranged don’t lose DPS time on eyebeams, have a chance to get targeted with Orb of Destruction, and have a small movement time for Burning Pitch. Melee have a 6 second DPS gap during bridge breaks.
- The “MoveIt” script was created for players who spend time moving for the eyebeams, for 1 second.
- The “DW No Miss” script is the same as the default except that it has the option flagged that makes dual wielders never miss on Krosus. Normally dual-wielders have about a 19% chance to miss, but on Krosus that isn’t true (we assume it’s a bug).
- Then Fury Warriors get “Fury Special,” which shortens the length of bridge breaks to 4 seconds. It turns out, the top Fury Warriors are staying in to DPS longer and then Heroic Leaping out – and by doing this it keeps the Juggernaut buff stacks up, adding a massive amount of damage to Execute.
To use the special scripts, click on the link for each one. When you are logged in to the site, you will be able to add it to your script list. Then you can select it on the sim setup page (you may need to refresh the sim page).
Standard Deviation – Don’t skip this section!
This explanation is simplest with an example. Let’s say your sims did 600k DPS, and has a standard deviation of 30k.
That means the majority of your attempts will be within 30k higher or lower, so 570k to 630k. As we all know, you can get somewhat lucky or unlucky. That standard deviation accounts for those high and low attempts. If you do the same fight twice in a night with the same players, you can expect to see your DPS vary between that range, without a change in your skill, just a change in your luck.
About 30% of your attempts might be 2x higher or lower – so down to 540k or up to 660k. That accounts for the super lucky or unlucky streaks. This isn’t a variance in the simulator, it’s a stats thing. We’re just following the rules statisticians have laid out.
Along the line of stats: the best players SHOULD beat simulations some of the time. If not, the simulator isn’t reflecting reality. This all comes back to the fact that simulators average the highs and lows, but for an individual attempt, you should get some that are higher if you’re playing at the top skill level.
Comparisons – alphabetical order by class
Death Knight – Frost
Charongar – 558k on an actual fight and 564k in AMR simulations for a difference of 1%. This used the “DW No Miss” version of Krosus that doesn’t let dual-wielders miss (because that’s how it works in game for this boss, an assumed bug).
Let’s take a look at how they match up:
- We don’t simulate anti-magic shell – he got 2 uses, generated 45 runic power (of which 16 was wasted). He generated a total of about 2,300 runic power over the course of the fight, so anti-magic shell accounts for about 2% of that.
- Spell usages match well (log vs sim):
- Howling Blast: 52 vs 51.3
- Obliterate: 79 vs 82.1
- Frost Strike: 59 vs 47.7 – The simulation has higher Breath of Sindragosa uptime (BoS), so more runic power was spent on BoS, which means less frost strikes.
- Breath of Sindragosa: 2 uses on the logs and sim, but 55 hits on the log and 73 on the sim. He got less procs than average during BoS, so it was shorter.
- Remorseless Winter: 11 vs 12.9
Death Knight – Unholy
Let’s see how the log and sims line up:
- The ability uses and damage look good (log vs sim):
- Virulent Plague: 169 vs 160.9
- Clawing Shadows: 74 vs 74.7
- Death Coil: 66 vs 62.7
- Sludge Belcher shows up as Ghoul on AMR. In logs the uses for Vile Gas / Melee / Cleaver are 52 / 147 / 62, and on AMR it’s 54 / 158 / 64. Damage matches well too.
- Army of the dead, uses match up, as does damage.
Demon Hunter – Havoc
Let’s take a look at how they match up:
- The ability uses look pretty close (log vs sim):
- Annihilation: 32 vs 32.8 – damage is a bit higher in the log from a lucky number of crits
- Chaos Strike: 51 vs 58.1
- Death Sweep: 14 vs 12.1
- Chaos Blades hits: 20 vs 17
- Blade Dance: 20 vs 18.9
- Eye Beam: 6 vs 5.1
- It looks like he used less Chaos Strikes and then a couple more of other abilities here and there instead.
Druid – Balance
How do they compare?
- The # of uses for the top damaging abilities match very closely – log vs sim:
- Starsurge: 48 vs 48.8
- Lunar Strike: 43 vs 47.3
- Solar Wrath: 69 vs 65.2
- New moon / half moon / full moon: 5 to 6 vs 5.7
- The simulation averages lucky and unlucky streaks and gets about 1 million more damage on starsurge and Goldrinn’s Fang than the log because of this. That adds up to about 10k DPS.
- Note: AMR logs group Goldrinn’s Fang under Starsurge (which are separate on WCL)
Log-Sim compare tip: Highly skilled players using Nature’s Balance as their level 100 talent are able to keep Moonfire and Sunfire up almost solely with Solar Wrath and Lunar Strike. If you have to cast Moonfire or Sunfire more than a couple times to keep the DoTs up, that is often a major difference from simulations to logs.
Then I tried to find a player who didn’t move perfectly, but still played really well. Zozo: 526k on an actual fight, and 525k when using the Krosus added-movement script. The default script comes in at 560k, getting 4 more Starsurge uses which adds up to 15k extra DPS. Add in the extra casts of other spells and you account for the extra DPS needed to get to 560k.
Druid – Feral
Let’s see how they compare:
- The spell usage looks similar:
- Rip: 161 hits vs 158.5
- Shred: 59 vs 57.4
- Rake: 107 vs 109 ticks and 27 vs 23.5 hits
- Melee (Auto Attack): 252 vs 253.9
- Ashamane’s Rip: 56 vs 56
- Thrash: The rotation in the simulator has a condition tied to the set bonus, to use Thrash, which you normally wouldn’t use on single target. However, when you get a clearcasting proc and have that set bonus, it’s worth using to buff damage. That’s why you see a 2+ Thrash uses in the simulator.
You can also take a look at the feral theorycrafting article we wrote, that took a deeper look into “pooling.”
Hunter – Beastmastery
NOTE: The Dire Frenzy talent that buffs your next Titan’s Thunder with 200% nature damage is bugged in the game. If your pet is more than about 8 yards away from you, it doesn’t benefit from the buff. The simulator assumes you always get the buff, so the simulations will likely be much higher if you have this talent. (You can customize that buff in the simulator to suit your needs, however).
How do they compare?
- The number of spell uses match up, with a few minor differences – he used a few extra in some places and a few less in others. Logs vs sims:
- Kill command: 46 vs 44.6
- Dire Beast – Stomp: 27 vs 29
- Main pet – Bite: 65 vs 73
- Main pet – Auto attack: 127 vs 127.8
- Cobra shot – 45 vs 49.9
- Auto shot – 80 vs 85
- It looks like just a few small differences in ability uses, and the damage looks right for each ability use, and overall is very close to the sims.
Let’s look at a log with Dire Frenzy to check out this in-game bug:
- Do the # of uses line up?
- Kill command: 56 vs 56.8
- Dire Frenzy: 23 vs 25.3
- The sim gets more auto attacks and cobra shots off, like the other log.
- The main difference is the Titan’s Thunder damage – due to the sim accounting for it, but the pet being out of range in-game, and the bug not buffing Titan’s Thunder.
- The logs have 4 million damage from Titan’s Thunder and the sim has about 9 million. That accounts for about 21k DPS.
- The rest of the damage difference is from the extra uses of auto attacks and cobra shots – in the log I see some gaps in his activity, perhaps due to running away for orbs, accounting for some of the loss in cobra shots and auto attacks.
Hunter – Marksmanship
Note: The simulations come up a lot higher for most characters. We’ve combed through every detail we can think of and the simulations look correct (if you see otherwise, let us know). It appears that just a few misses or cancellations of Aimed Shot is accounting for most of the difference, adding up to a lot of damage. However, when looking at the best players (who didn’t get too lucky or unlucky), the sims match. For most classes, the top ranks are a combo of skill and lucky streaks, but for MM hunters, it appears to really come down to getting in those extra aimed shots.
Because of this, I looked at a couple 99th percentile players, finding ones that weren’t there because of lucky streaks. (Other classes I looked around the 95th percentile).
How do they match up?
- The ability uses are in line, but you can see that he used more Marked Shots and less Aimed Shots.
- Aimed Shot: 59 vs 61.8, log has a higher crit rate, bringing damage up.
- Marked Shot: 27 vs 24.5
- Arcane Shot: 64 vs 76.5
- That’s about it, pretty simple.
Hunter – Survival
When comparing, you have to expand the abilities on WCL in order to see the breakdown per pet. AMR has separate lines for each pet.
- He has a few less uses of Mongoose bite in the logs and using other abilities a handful of times more, like Lacerate and Raptor Strike. Logs vs Sims:
- Mongoose Bite: 79 vs 85
- Flanking Strike: 28 vs 26.4
- Lacerate: 23 vs 20.6
- Serpent Sting ticks: 104 vs 102.9
- The rest also match pretty closely, with the log using a bit more Raptor Strikes.
- You can see the sim uses more Mongoose bites, but a few less of the others. Since he has the 4 pc bonus, get the extra Mongoose Bites in is a high priority. However, he had lucky crits for that ability, evening out the damage a little compared to the sims.
- Logs vs Sims:
- Mongoose Bite: 90 vs 90.9
- Flanking Strike: 18 vs 22.5
- Lacerate: 18 vs 16.6
- Serpent Stine ticks: 78 vs 81.8
- The rest also match pretty closely, with the log using a bit more Raptor Strikes.
- Why is the sim doing more damage if the spell usage matches (especially the top damaging ability, Mongoose Bite)? There are 2 larger things I noticed:
- He actually got 5% less crits on Mongoose Bite than the average, accounting for a lot of that damage.
- He also got 1 less use of the 4pc Mongoose Power buff (increases damage by 20% for 10 seconds), compared to the sim. 10 seconds is about 4% of the duration of the fight.
Mages – All
Mages lose significant damage when they have to move. We are often seeing more movement in fights than the boss scripts we have right now. Adding more movement to the boss scripts is an option that we are considering to make the DPS numbers match a bit closer. I’ll play around with it and provide an update.
Mage – Arcane
Arcane mages have a huge skill gap in the rankings, and playing perfectly is much harder than other classes. Expect simulations to be about 50-100k higher than your logs depending on your skill level. We’ve checked all mechanics, buffs, and everything seems correct in the simulator. We also checked SimC and have talked with Frosted over at Altered Time – SimC is showing similar results (both simulators match closely to each other for the mage specs). So the simulators are working, as best as we can all tell. But perfect play for a simulator happens to be very difficult to replicate in-game for a human (whereas other classes, it isn’t so sensitive).
- What is different in the spell usage analysis:
- Arcane Blast: 90 vs 94.1
- Arcane Missiles: 289 vs 270.7
- Mark of Aluneth: 5 vs 3.9
- Arcane Barrage: 9 vs 13.7
- Deadly Grace: 12 vs 14.7
Mage – Fire
Fire Mages also sim higher than most players, but unlike Arcane Mages, the very top players can match (or beat) the sims, though it’s very difficult.
How does the log match the sim?
- The spell usages looks pretty similar:
- Pyroblast: 66 vs 70.9 (the player got slightly more crits than average)
- Ignite: 220 hits vs 208 (Phoenix Reborn is 13 vs 17.4)
- Mirror Image: 117 hits vs 102
- Fire Blast: 27 vs 27.1
- Fire Ball: 26 vs 33.5 (the log got slightly less crits)
Mage – Frost
Frost mages are also very hard to play perfectly, there is a lot less forgiveness than other classes. Simulations often end up much higher than logs, even for top-notch players. You can even see for the top rankings for an 890-892 gear level, the top players are 100k higher than players at rank #75. Those in rank #75 are still in the 99th percentile too. If the sims are 100k higher for you, don’t be surprised. While the simulations are correct, we are looking into smart ways to account for the skill gap.
To triple check that the simulator could be matched, we also did tests on a target dummy, and since procs can really make a difference, it took about 10 tries to get an ‘average’ log that would be similar to a simulation. Notice how Frost Mages have a higher standard deviation than other classes by about 10k at the 890 iLevel.
Because of this, I look at the 99th percentile, instead of the 90-95th, in order to compare the logs to the sims.
Let’s see how the log compares to the sims:
- Spell usage looks great:
- Ice Lance: 94 vs 87.9 (with the log having a slightly lower than avg crit rate)
- Flurry: 42 vs 39.9
- Frostbolt: 104 vs 113 (when comparing total damage, expand the frostbolt section on WCL or add up the Frostbolt, Icicle and Unstable Magic lines in AMR. And compare the hits column on both for usage details)
- Ebonbolt: 6 vs 6.3
You can see the spell uses are very close, and this top player performed exceptionally, without getting overly lucky (some up, some down), which is why he’s matching the simulations closely.
Monk – Windwalker
How do they compare?
- The # of casts for the top damaging abilities match very closely: Log vs Sim
- Fists of Fury: 12 vs 12.1 (Note WCL’s damage includes crosswinds, and it’s separated out on AMR)
- Rising Sun Kick: 34 vs 32.3
- Storm, Earth, and Fire: 4 vs 4 (you can see the # of casts on the casts tab on WCL)
- Strike of the Windlord: 6 vs 6
- Blackout Kick: 38 vs 51 – He wasted 7 more chi than the simulation and one less combo breaker point, accounting for 8 of his missing Blackout Kicks.
- It looks like he also used Crackling
- Other spells also match closely
Tip: When comparing the log to AMR, if you expand the ability on WCL you can see the extra details that are grouped on AMR for things like Storm, Earth and Fire.
Paladin – Retribution
The simulator doesn’t add in extra damage from the Retribution buff (20% damage increase for 20 seconds, every time a player dies). It’s so variable – looking through 890 logs, the buff happens anywhere from 1-5 times throughout the fight, and that goes down as gear goes up. We may add it as a configurable option in the future.
You can roughly account for the extra damage: go to the buff tab of WCL and look at the total % uptime for Retribution. Multiply that by 0.2 (for 20% damage increase) and you’ll have a rough estimate of how much extra damage you got from that buff. It isn’t perfect, because while that buff is up you might have used some big cooldowns or gotten lucky (or unlucky) crit streaks.
So I tried to find logs that had the buff 1 time to do a comparison. Zenithium had it up 7% of the time, for a total (rough) damage increase of 1.4% extra damage.
How do the logs and sims match up?
- The spell uses look accurate (log vs sim):
- Templar’s Verdict: 80 vs 79.3. Total damage is a bit higher on the log due to lucky crit rate
- Blade of Justice: 47 vs 43.6
- Zeal: 69 vs 73.4
- Judgement: 27 vs 28.5
- Legendary: Liadrin’s holy power gain: 22 vs 18.7 / 4pc bonus check: Power of the Highlord Holy power gain: 21 vs 21.6. Other power gains look similar too.
- Damage matches up as well. It looks like the simulator used Blade of Justice a few less times, and Zeal a few more.
- The spell uses look good (log vs sim):
- Templar’s Verdict: 79 vs 75.1
- Blade of Justice: 44 vs 40.9
- Crusader Strike: 77 vs 77.6
- Judgement: 27 vs 27.1
- Shield of Vengeance – the simulator gets the theoretical max, which can be hard to do in a real fight. Not to mention, you are using this to stay alive in a real fight, not to max your dps. So you may see an extra 1-2 million total damage on the simulator, when compared to a log, adding up to roughly 5,000 DPS in a 5 minute fight, give or take.
Priest – Shadow
- Let’s see how the spell usage matches up:
- Void Bolt: 86 vs 95.6 – he also got a slightly lower crit rate than average, dropping his damage by about 1.5 million (6k DPS).
- Vampiric Touch: 156 hits vs 151.5 hits
- Shadow Word: Pain: 237 hits vs 227 hits
- Mind Blast: 41 vs 40.2
- Shadow Word: Death: 15 vs 15.8
- Mind Flay: 55 vs 71.6 –
- Mind Flay is the filler spell. And since it is channeled, if you have to move, you lose Mind Flay casts. It also looks like Impiety took some time to DoT some adds, which would reduce the Mind Flay casts. Void Bolt is instant, so you cast that when you have to move, same with SW: Death.
- Buff uptimes look good. Voidform was up in the log 5% lower than the average sim.
Rogue – Assassination
How do the logs match up to the sim?
- The number of ability uses look right (log vs sim)
- Envenom: 20 vs 22 – Note that in AMR we have the Poison Bomb grouped with Envenom, but they are seperated in WCL.
- Rupture: 9 vs 8.1
- Mutilate: 51 vs 50.2
- Kingsbane: 4 vs 4 – They are grouped in AMR, but separated in WCL, so add the lines up in WCL to get total damage.
- Garrote: 10 vs 9.5
- The damage matches up as expected as well.
Rogue – Outlaw
After looking through logs, it has become clear that due to the high variability of the spec, it is not realistic to get sims to match any single boss attempt. However, this doesn’t mean we haven’t verified individual components of the spec, we definitely have. It just doesn’t lend itself well to this type of comparison. Unless they are using the Slice and Dice talent, ha! So I grabbed a log of a top ranking person who did happen to use that talent, for comparison purposes.
- The spell usage looks close:
- Run Through: 57 vs 53.7
- Melee (Auto Attack): 586 vs 593
- Saber Slash: 157 vs 170 hits – Note in WCL it’s on 2 lines (Saber Slash has a chance to hit twice, tracked separately in WCL, and tracked on 1 line in AMR).
- Recursive Strikes: 80 vs 114
- Main Gauche: 176 vs 153
- Damage matches up well too. It looks like the player got unlucky with Recursive Strikes and lucky on the Main Gauche procs, roughly averaging out in the end.
Rogue – Subtlety
- The ability uses match well:
- Eviscerate: 53 vs 54.7
- Nightblade: 12 vs 11
- Shadowstrike: 92 vs 93.7
- Recursive Strikes: 67 vs 92 – It looks like he got unlucky with Recursive Strikes
- Buff uptimes look good, with one note. The way we implemented Sub rogue is we ‘re-used’ Stealth for a lot of the buffs. So the Mastery of Subtlety uptime we show is only the 5 seconds after stealth, not the time while you were in stealth. In game, you get a Mastery of Subtelty buff in addition to Shadow Dance / Stealth. So while the simulator is doing the correct things and matching logs, the buff tracking isn’t displaying in a way that lines up with logs. We’ll make some changes to improve on that.
Shaman – Elemental
Even though they are super close, it’s worth taking a look at some of the differences.
- The spell usages look similar (log vs sim):
- Lava Burst: 76 vs 72
- Lightning Bolt: 72 vs 85.5
- Elemental Blast: 22 vs 21.9
- Flame Shock: 25 vs 24.8 (add up the 2 line in AMR to match the 1 line in WCL)
- Spell uses look good:
- Lava Burst: 52 vs 53.3 – The Pristine Proto-Scale Girdle ticks are on a separate line, but grouped with Lava Burst on AMR – so to compare total damage, add the Proto-Scale line in WCL to the total Lava Burst damage line.
- Lightning Bolt: 50 vs 50.3
- Elemental Blast: 16 vs 16.6
- Fire elemental: 31 vs 35.9 Fire Blasts
- Buff uptimes also look good and damage matches too.
Shaman – Enhancement
These tests use the “DW No Miss” version of the Krosus script.
How does the log and sim match up?
- Spell uses match well
- Stormstrike: 104 vs 108.2
- Lava Lash: 53 vs 54.4
- Boulderfist: 56 vs 54.7
- Unleash Doom: 221 hits vs 175 – Looks like he got lucky procs
- Flametongue: 20 vs 18.1
- Flametongue, Boulderfist and Landslide buff uptimes are near 100% uptimes in the log and sims. Looks like he got lucky Stormbringer procs for about 10% more uptime. Since Stormbringer has a % chance, not an RPPM, the can vary a lot from fight to fight.
And let’s see how these compare:
- Spell uses match pretty well here too
- Stormstrike: 105 vs 108.6
- Lava Lash: 50 vs 54.2
- Boulderfist 48 vs 52.5
- Windfury Attack: 228 hits vs 226.5
- Unleash Doom: 163 hits vs 172
- Buff uptimes also look good here.
Warlock – Affliction
A note about our default rotation: The rotation in the simulator puts slightly less emphasis on Unstable Affliction damage than a lot of the top logs. In logs you will often see average Unstable Affliction damage higher, but the average damage on the other DoTs lower. Overall, the average damage is very similar.
We have experimented with many rotations for Affliction warlocks and will continue to do so. Creating a rotation that matches exactly what some of the top logs are doing has resulted in slightly lower damage on average, which is why we are using the current default. The play style that focuses on boosting UA damage above all else might have a higher top end potential, but it makes it much more difficult to achieve 100% uptime on Corruption and Siphon Life. The rotation can be customized, here’s a 5-ish minute video that gives a quick overview.
How do they compare?
- The # of casts are close:
- Unstable Affliction: 44 vs 45.1 – Compounding Horror is added in to the line on WCL, and separated out on AMR
- Agony: 16 vs 17.8
- Corruption: 19 vs 19.8 – Harvester of Souls is added in to the total damage on WCL, and separated out on AMR
- Pet’s Doom Bolt: 100 vs 102, damage looks similar too
- Siphon Life: 17 vs 19.8
- Drain Soul: 127 hits vs 156
Warlock – Demonology
Alaguster did 598k DPS in an actual fight and 611k in AMR simulations with the added-movement script, for a difference of 2.2%. (The sim does 639k dps with the regular default Krosus script). He also got targeted with an Orb, and remember the simulations even out the times you get targeted and don’t get targeted.
How do they compare? I will use the added-movement script.
- The # of casts are fairly similar:
- Wild Imp: 1918 casts vs 1849 (In WCL add up the wild imp Fel Firebolt line and the one under Call Dreadstalkers to get the total).
- Call Dreadstalkers: 14 vs 15.9
- Summon Felguard: 122 hits from the Felguard vs 132.2 – it looks like he could have used it at the end of the fight again for 8-10 seconds, so that accounts for the lower # of hits in the log. Also if the warlock switches targets to an add, the pet will start running to it and then run back to the boss. We don’t have that modeled in the simulator right now.
- Demonbolt: 90 vs 101 – he got targeted with an orb, and party way through the fight he ran for 3-4 seconds from melee back toward the ranged group. Those extra movements explain some of the loss of Demonbolts.
Warlock – Destruction
Nerwick did 506k in an actual fight and 505k with the MoveIt script (or 517k with the default Krosus script). Based on the replay it looks like he moved for 1 second during eyebeams, so we’ll use the MoveIt script.
Let’s see how the log compares to the sim:
- The spell usage details:
- Chaos Bolt: 48 vs 52.3
- Immolate: 24 vs 17.7 (he used immolate to dot the adds)
- Incinerate: 68 vs 94.4 (I see a lot of players missing incinerate casts, or replacing them with something else)
- Conflagrate: 47 vs 49.9
- Imp’s Firebolt: 132 hits vs 130.2 hits
- And so on.
- Lucky mastery values on your Chaos Bolts (or unlucky ones) can swing your damage a lot, so keep an eye on that when comparing to logs.
Warrior – Arms
Besides Combat Rogues, this is one of the hardest specs to have a log match a simulation. This is due to the high number of actions per minute. The buff, Focused Rage, is off the GCD and gets used quite often.
Even though they are that close, let’s see what ends up different:
- Spell usages looks to match up:
- Mortal Strike: 45 vs 42.2
- Execute: 25 vs 28.3
- Colossus Smash: 41 vs 41.1
- Slam: 66 vs 67.1
- Battle Cry (Corrupted Blood of Zakajz): 42 vs 44 hits
- Melee (Auto Attacks): 79 vs 80
- Damage from each ability and buff uptimes look to be in lines as well.
How does this one compare?
- The spell usage still looks good:
- Mortal Strike: 42 vs 41.3
- Fel-Crazed Rage: 39 hits vs 36.6
- Execute: 36 vs 32.8
- Colossus Smash: 39 vs 41.3
- Slam: 48 vs 53.8
- While those are all pretty close, the log is lower from some unlucky crit rates on the top damaging abilities.
- His avg crit rate for Mortal strike is about 61%, but he only got 55%. If just 3 more of his Mortal strikes crit, that would be 1.8-ish million more damage (about 7k DPS).
- His avg crit rate for Execute is also about 61%, but he got 48%. That adds up to just over 2 million damage, which is about 8-9k DPS.
- As you can see, even a small difference in crit rate can really add up.
Warrior – Fury
When Krosus breaks the bridge, there is a 6 second duration where most melee specs don’t DPS the boss. But Warriors can Heroic Leap out at the last minute, and good players have about a 4 second gap instead. This allows them to keep the Juggernaut buff stack up in-game, whereas the Krosus script forces a 6 second bridge-break on you (that can be customized). Because of this, Fury Warrios will beat the default script more often, as having the Juggernaut buff up during that last phase adds a lot of extra damage.
So I made a version of the boss script that has a 4 second bridge-break gap instead, called “Fury Special.”
Here’s an example of a player using Heroic Leap and keeping the Juggernaut buff up:
Tooyoo did 610k in an actual fight and 609 in AMR simulations using the “Fury Special” script that lets him DPS for 2 more seconds on the bridge break. When using the “DW No Miss” script, which has a 6 second gap during the bridge break, allowing Juggernaut stacks to fall off, he loses 20k DPS, for 589k.
How does the log compare to the “Fury Special” simulations?
- The spell uses match up well:
- Raging Blow: 68 vs 73.6
- Execute: 30 vs 30.3
- Melee (Auto Attack): 360 vs 370
- Rampage: 26 vs 28.2
- Bloodthirst: 74 vs 77.9
Execute: You can see that at 4 minutes and 32 seconds, he Heroic Leaps away for the bridge break. The break starts at 4:30, which is when most people run away. The fight lasts 23 more seconds after the bridge breaks, and Tooyoo was able to keep his Juggernaut stacks up really high, getting massive execute damage in. On AMR’s simulation, expand the “Log” view below the spell table, and type “Juggernaut” into the spell box. You can see that he maintains the buff while the bridge is breaking, which happens between 270 – 276 seconds. To compare, you can see in the “DW No Miss” script sim, which has the full 6 sec bridge break, Juggernaut falls off at 274.