90% of discrepancies between TC tools can be quickly solved provided that each tool generates detailed break-down information for each ability. The last 10% is decrypting the difference between how those abilities are combined...... which can be painful. Althor has volunteered to track it down and his blood-hound skills are second to none.
Actually this only serves as a comparison. If the same/similar error is found in both, then it will be passed over.
Formally Xyrm/Zurm, the Ret Pally. Now playing my rogue, Zyrm, more casually with RL friends.
A few differences I've noticed between Rawr and SimulationCraft's numbers so far (Where possible I was comparing the numbers between both programs and also with an untalented pre-made Dwarf Paladin on the PTR, taking care to account for the slightly changed base stats):
1) Rawr has +20 base AP over both in-game, SimulationCraft and wowhead.
2) If you click the Bloodlust option under the Buffs tab then it gives you Bloodlust for the entire length of the fight. Just using checking it under the Options tab seems to be more accurate. (One of those recent BiS .rar files had both checked and thus had a 3.6 speed Axe swinging for every 1.98 seconds for the entire fight length).
3) On a number of abilities, boss dmg reduction is being applied to abilities in Rawr (when you hover over it and check the Average Hit) even when you set the target level to 80. Hammer of Wrath is one example of this. Changing the target level between 80 and 83 has no effect.
4) SimulationCraft was missing level 80 versions of Exorcism (and Holy Wrath). I've fixed this.
5) SimulationCraft had a bug where it wasn't including a player's crit chance in the calculations to determine if Righteous Vengeance would crit (with 2pc T9). I've fixed this.
(Question: Is the crit chance calculated when Righteous Vengeance is applied, when refreshed or per tick?)
This was a big difference in DPS.
6) SimulationCraft has Righteous Vengeance coded like Ignite and Deep Wounds. i.e. when a new crit happens to proc it, it in effect cancels the previous DoT (though carrying over the correct amount of damage of course) and applies a new DoT that will tick after 2 seconds etc.
Rawr on the other hand basically adds the equivalent Righteous Vengeance damage to the spells that cause it but it assumes that when Righteous Vengeance is refreshed that it doesn't interfere with the tick timer (so more like Deadly Poison and Shadow Word: Pain).
Starfox in the post here: SimulationCraft for Paladins has some evidence that SimulationCraft's model is correct.
i.e. That you can "munch" ticks and thus you won't actually see the full benefit of the talent because if you're proccing Righteous Vengeance too often (more often than once a tick) then you risk delaying the next tick and thus you will have fewer ticks per fight.
If anyone else is willing to ascertain either way which is correct that would be appreciated.
There's a few other smaller differences too, for example, the .rar file I tested had the boss type as Undead and SimulationCraft by default uses Humanoid.
I wasn't able to find an option to set the mob type. Is the Patchwerk in SimulationCraft assumed to be race neutral? I.e. the Crusade talent does not grant its additional +3% damage bonus for Humanoid, Elemental, Undead, Demon.
Is the Patchwerk in SimulationCraft assumed to be race neutral? I.e. Crusade does not grant its additional +3% damage bonus for Humanoid, Elemental, Undead, Demon.
Yes, I ninja-edited my post to also mention that SimulationCraft's default target race is Humanoid (can be changed with target_race=undead etc).
Setting it Undead also causes Exorcism to always crit.
2) If you click the Bloodlust option under the Buffs tab then it gives you Bloodlust for the entire length of the fight. Just using checking it under the Options tab seems to be more accurate. (One of those recent BiS .rar files had both checked and thus had a 3.6 speed Axe swinging for every 1.98 seconds for the entire fight length).
3) On a number of abilities, boss dmg reduction is being applied to abilities in Rawr (when you hover over it and check the Average Hit) even when you set the target level to 80. Hammer of Wrath is one example of this. Changing the target level between 80 and 83 has no effect.
6) SimulationCraft has Righteous Vengeance coded like Ignite and Deep Wounds. i.e. when a new crit happens to proc it, it in effect cancels the previous DoT (though carrying over the correct amount of damage of course) and applies a new DoT that will tick after 2 seconds etc.
Rawr on the other hand basically adds the equivalent Righteous Vengeance damage to the spells that cause it but it assumes that when Righteous Vengeance is refreshed that it doesn't interfere with the tick timer (so more like Deadly Poison and Shadow Word: Pain).
I am fairly certain I have never uploaded a file with Bloodlust checked, 30% haste all the time is a little too good (that is the buffs bloodlust). The options Bloodlust is one bloodlust (unsure if the encounter was over 10 minutes would you get 2 lusts, like Arthas will likely be).
What damage reduction would happen with HoW on a boss versus an level 80 (other than the 4.8% crit reduction on bosses)? Armor is different, but most attacks are Holy. Spells for Ret and DKs (just Exo matters) have a 3% crit reduction on bosses, unsure about other classes.
RV is like Ignite. Rawr's code is close enough; I believe it was done that way to get the "true" damage of each attack.
Millions of words are written annually purporting to tell how to beat the races, whereas the best possible advice on the subject is found in the three monosyllables: 'Do not try.'
I am fairly certain I have never uploaded a file with Bloodlust checked, 30% haste all the time is a little too good (that is the buffs bloodlust). The options Bloodlust is one bloodlust (unsure if the encounter was over 10 minutes would you get 2 lusts, like Arthas will likely be).
What damage reduction would happen with HoW on a boss versus an level 80 (other than the 4.8% crit reduction on bosses)? Armor is different, but most attacks are Holy. Spells for Ret and DKs (just Exo matters) have a 3% crit reduction on bosses, unsure about other classes.
RV is like Ignite. Rawr's code is close enough; I believe it was done that way to get the "true" damage of each attack.
It wasn't your set I was using to perform my comparisons with on the mechanics. It was the one linked in Ret BiS Gear Lists: 3.2 Edition. I ended up disabling every buff etc. in both Rawr and SimulationCraft to perform testing of the basics.
As for the damage reduction, Rawr seemed to be reducing the Average Hit amount by around 3.5 or 4%, almost as if it was trying to reduce the damage for Boss level partial resists (though that would be 6%) no matter the level of the boss. Thus, this wasn't related to crit suppression.
It wasn't your set I was using to perform my comparisons with on the mechanics. It was the one linked in Ret BiS Gear Lists: 3.2 Edition. I ended up disabling every buff etc. in both Rawr and SimulationCraft to perform testing of the basics.
As for the damage reduction, Rawr seemed to be reducing the Average Hit amount by around 3.5 or 4%, almost as if it was trying to reduce the damage for Boss level partial resists (though that would be 6%) no matter the level of the boss. Thus, this wasn't related to crit suppression.
DPS in that set is 9555 if you select humanoid (that was set in the OP), mine was slightly higher.
I forgot about partial resists, that is that 4% less.
Millions of words are written annually purporting to tell how to beat the races, whereas the best possible advice on the subject is found in the three monosyllables: 'Do not try.'
DPS in that set is 9555 if you select humanoid (that was set in the OP), mine was slightly higher.
I forgot about partial resists, that is that 4% less.
Partial resists are 6% on bosses, but 0% on same level. I was seeing around a 4% drop in damage in Rawr on some spells even when I had the level set to 80. (It was correctly adjusting the avoidance rate etc.)
If I recall correctly it was hitting with the melee weapon for a fair bit less (in terms of Average Hit) in Rawr than in SimulationCraft when I had all buffs and talents off etc. (And that's assuming glancing blows too).
I'm not trying to disparage Rawr etc. I'm just pointing out the differences I noticed between Rawr and SimulationCraft in the modeling of Paladins to maybe explain some of the DPS differences.
Thanks to this auditing I discovered the bug in SimulationCraft where we weren't making use of the Player's Crit% when determining if Righteous Vengeance crit or not. I fixed it and the DPS difference between Rawr and SimulationCraft is a fair bit closer now.
I might be mistaken, but it looks like SimulationCraft has the Paladin recast Seal of Vengeance every ~120 seconds. The output file has SoV is listed as a "Dynamic" buff with 99% uptime. Seals are now 30 minute buffs that are generally cast before combat is engaged. Should it be treated as a "Constant" buff with 100% uptime similar to Blessings, etc.?
I might be mistaken, but it looks like SimulationCraft has the Paladin recast Seal of Vengeance every ~120 seconds. The output file has SoV is listed as a "Dynamic" buff with 99% uptime. Seals are now 30 minute buffs that are generally cast before combat is engaged. Should it be treated as a "Constant" buff with 100% uptime similar to Blessings, etc.?
That actually refers to the Holy Vengeance buff that is procced while you have SoV up.
The current version of SimulationCraft appears to be using the old value of 32.7 Haste Rating per Melee Haste %. The 3.2 value is 25.2 Haste Rating per Melee Haste %.
Ex: SimulationCraft states that I have 452 Haste Rating and 13.79% Melee Haste unbuffed. Armory states that I have 17.92% Melee Haste. 452 Haste Rating divided by 17.92% Melee Haste is 25.2 Haste Rating per Melee Haste %.
Is this just a cosmetic display bug or would this change (slightly) increase Paladin's DPS in SimulationCraft? Rawr appears to be using the updated conversion value, so this may be another contributing factor to the apparent DPS differences between the two tools.
The current version of SimulationCraft appears to be using the old value of 32.7 Haste Rating per Melee Haste %. The 3.2 value is 25.2 Haste Rating per Melee Haste %.
Ex: SimulationCraft states that I have 452 Haste Rating and 13.79% Melee Haste unbuffed. Armory states that I have 17.92% Melee Haste. 452 Haste Rating divided by 17.92% Melee Haste is 25.2 Haste Rating per Melee Haste %.
Is this just a cosmetic display bug or would this change (slightly) increase Paladin's DPS in SimulationCraft? Rawr appears to be using the updated conversion value, so this may be another contributing factor to the apparent DPS differences between the two tools.
I took the BiS set listed on the front page, swapped the cloak enchant from 22agi to 23haste, swapped out DMC: Greatness for the 245 Death's Choice and then swapped out 4 Str/Crit gems for Str/Haste Gems.
At each step I compared the DPS to alternatives and what I currently had. For the Crit->Haste gems I made 6 different profiles with differing numbers of gems swapped.
I took the BiS set listed on the front page, swapped the cloak enchant from 22agi to 23haste, swapped out DMC: Greatness for the 245 Death's Choice and then swapped out 4 Str/Crit gems for Str/Haste Gems.
Crit is better than Haste for T9 due to the set bonus. Perhaps in T10 Haste will be better, but not now. The reason why is haste is good to help put the SoV dot, then it is a weak stat afterwards (only affects auto, Seal hit, and GCD of Exo/Cons).
Millions of words are written annually purporting to tell how to beat the races, whereas the best possible advice on the subject is found in the three monosyllables: 'Do not try.'
The T9 BiS lists now seem to be converging in terms of gear pieces at least. I'm also suspicious of 10 haste rating providing higher DPS than 10 critical strike rating. Haste may have second order effects on the whole priority order by decresing the GCDs of two of the components. From what I can tell it's allowing substantially more Crusader Strikes in addition to more melee strikes and SoV procs to occur.
Crit is better than Haste for T9 due to the set bonus. Perhaps in T10 Haste will be better, but not now. The reason why is haste is good to help put the SoV dot, then it is a weak stat afterwards (only affects auto, Seal hit, and GCD of Exo/Cons).
The SimulationCraft BiS set still has 986 crit rating and only 476 haste rating even after the re-gemming. It seems that there are some local maxima and minima around the BiS with regards to slightly change the haste and the crit and that the slight change included in the SimulationCraft BiS set takes advantage of this.
As Glutton mentions it's probably enough haste to squeeze in some more Crusader Strikes etc.
As Glutton mentions it's probably enough haste to squeeze in some more Crusader Strikes etc.
Haste will affect the GCD of Cons and Exo. Maybe with enough haste you can get two more CSs in 5 minutes, but then you are losing Crit to get those extra CS. It isn't worth it. If simcraft says that, then there is something wrong with it.
The 40 Strength food is like 3 dps more than a Fish Feast.
Millions of words are written annually purporting to tell how to beat the races, whereas the best possible advice on the subject is found in the three monosyllables: 'Do not try.'
Haste will affect the GCD of Cons and Exo. Maybe with enough haste you can get two more CSs in 5 minutes, but then you are losing Crit to get those extra CS. It isn't worth it. If simcraft says that, then there is something wrong with it.
The 40 Strength food is like 3 dps more than a Fish Feast.
The relative value of any stat that provides any sort of scaling and isn't capped increases the further behind it falls behind the other stats. That is simple math.
In terms of Haste Rating, don't forget that it costs far less rating to increase Haste by 1% then Crit by 1% (especially so for Paladins). And while it may be likely that when stats are balanced that a Paladin would gain more from 1% crit than 1% haste due to the cooldown heavy nature of most of a Paladin's damage, when the stats are skewed there's no guarantee that still holds.
In terms of the melee damage and any on-next-swing type procs especially, 1% haste tends to give the same increase as 1% crit but you can get more of it. And auto-attack DPS is still a non-trivial percentage of a Retadins overall damage.
Here are the results from the SimCraft set and a 2*Death's Choice using version of the BiS set linked on the first page:
For melee DPS you can see that the SimulationCraft profile has more hits on average, though a lower crit rate, but also the average non-crit hit is higher. Why would the average hit be higher? Because faster swings mean more benefit time on things like Avenging Wrath and Berserking. They're more likely to be up when you can make use of them if you're hitting faster.
The more haste you have in general decreases the time between being able to re-use a higher priority ability once it comes off cooldown, which is why in the SimulationCraft profile you see more high priority Crusader Strikes and fewer lower priority Consecrates.
Now it's true that Rawr has a "wait for higher priority ability if it has X seconds left on it's cooldown" and SimulationCraft doesn't have a generic version of this yet (Dedmon is working on one) but even then, more Haste means that the value of such a variable can be lowered.
If you can see an actual problem with the two parses I've linked then please, let me know. Purely that it differs from what Rawr says isn't necessarily useful without some reasoning why that would show why one is correct and one is incorrect.
BTW, I haven't yet audited SimulationCraft's exact mechanics for stacking talents etc. (i.e. which stack additively and which multiply together) so it's possible that some of these could be out, but that shouldn't have much effect on a haste vs crit discussion.
I can't quite understand the simcraft so this may be a stupid question. When you adjusted the haste rating return for melee haste did you leave the casting haste % per point the same. The gcd for exo and cons would still be governed by the old haste conversion.
I can't quite understand the simcraft so this may be a stupid question. When you adjusted the haste rating return for melee haste did you leave the casting haste % per point the same. The gcd for exo and cons would still be governed by the old haste conversion.
The change I made was to overload the init_rating() method for Paladins:
rating.attack_haste is only used for abilities that are children of attack_t (as opposed to spell_t which use rating.spell_haste unless a class or ability overloads this too of course).
Exorcism is a child of paladin_spell_t (which in turn is a child of spell_t):
struct exorcism_t : public paladin_spell_t
{
...
}
So is Consecration.
Things like Crusader Strike derive from paladin_attack_t
struct crusader_strike_t : public paladin_attack_t
{
...
}
So to answer your question, the haste scaling of Exorcism and Consecrate were untouched by the change I made.
SimulationCraft has a proc-based implementation of the Tier 10 2pc set bonus (and 4pc support).
I've run the BiS sets (both SimulationCraft's and the Rawr with 2*Death's Choice) through the simulator turning off 4pc T9 and turning on 2pc T10.
I added a special variable: tier10_2pc_procs_from_strikes=0 or 1 which if set to 1 (default: 0) enables Crusader Strike and Divine Storm to also have a chance to trigger the set bonus.
1) With 2pc T10 + 2pc T9 then the differences between the "SimulationCraft" BiS set and the "Rawr" BiS set (albeit with 2* Death's Choice) is basically negligible.
2) If only auto-attack swings can proc it then the DPS gain (compared to the SimulationCraft set) is 348 DPS by losing 4pc T9 and gaining 2pc T10 (just going off the worth of the set bonuses obviously).
3) If auto-attack swings, crusader strike and divine storm can all proc it then the DPS difference is 476 DPS.
Assuming only auto-attacks can proc it then the average time between procs is 5.76 seconds (this is of course assuming there's no ICD). If CS and DS can proc it then the average time between procs is 3.02 seconds.
Naturally, it's possible that there might be further DPS gains if the priority list was changed around to suit the set bonus more, but these numbers are using the default SimulationCraft action list.
Using the files Althor provided I attempted to investigate the haste maxima we're observing in this gear set. The gear set has 476 haste rating, or 46.93% melee haste % when raid buffed. I proceded to add and subtract haste to the gear set in increments of 10 without taking away other stats.
Of course adding haste rating without taking away any other stats is going to result in a DPS increase. This occured predictably at 486 (+10), 496 (+20), and 506 (+30) haste rating. It appears that 506 (+30) haste rating is near some sort of local maxima. At 516 (+40) haste rating I observed a moderate drop of twenty nine DPS from 506 haste rating.
This increase in haste between 516 > 526 was correlated to a drop from 66 Crusader Strikes to 65 Crusader Strikes. Another major change in the number of Crusader Strikes occurs around 440 haste rating where the number increases sharply from 58 to 66; this region is also correlated with a sharp increase in the DPS per additional haste.
Red squares: number of crusader strikes per 300 seconds
Blue line: DPS
OP lists HoW>Judgement>CS>DS>Cons>Exo rotation for the BiS sets, has this changed to Judgement>CS>DS>Cons>HoW>Exo?
Even using Rawr's BiS list with Rawr 2.2.23 (the Blood Elf BiS list with all items available and using DMC: Greatness) then it shows Judgement>CS>DS>Cons>HoW>Exo doing more DPS than HoW>Judgement>CS>DS>Cons>Exo.