Prophecy of Fear [Megathread]

[Curnivore]

As a troll I'm not surprised at all. During bloodlust even any race may not need haste. Though thanks for reminding me of this, I may go mastery again there since we seem to rely too much on me for the fist archi phase and for the rest phases other classes are better anyway.

I had done something similar in manno. It was based on intuition though. I theorized that since we rely so much on killing the imps fast and since almost all dps on them is done on full mana, max mastery would be best for them.


A bit of an issue though is that I would drop way below 1400 and I don't know if that's wise, if only for the utility disadvantage of slowness.

[Rinoa]

It's not worth dropping a lot of haste to those levels unless your fight length is extremely short, which typically means well under a minute. Not that this is a thread about such things, let's not derail further. I just wanted to clarify.

[Searix]

It's not worth dropping a lot of haste to those levels unless your fight length is extremely short, which typically means well under a minute. Not that this is a thread about such things, let's not derail further. I just wanted to clarify.

This is pertinent, we're talking about how much to gear for PoF.

Mastery is also better if you ever hit multiple targets, and haste is worst stat for fire

[Searix]

Question is whether the gain to switching to mastery for PoF damage, which hits first doomfire, is enough to warrant recommending other mages regem/enchant mastery

[Searix]

You can enter commands in the Override tab in SimC in order to force it to do certain things which you cannot do with the built in drop down menus. One of these things is dictating fight length. Courtesy of Frosted, here are three commands which you can use to do so:

vary_fight_length=0

fixed_time=1

max_time=x

Cool, thanks for that. Ran it again with those settings

Single target, 40 second fight (full lust):


# Gear Summary
# gear_ilvl=743.53
# gear_stamina=7056
# gear_intellect=5310
# gear_spell_power=2666
# gear_crit_rating=1737
# gear_haste_rating=1406
# gear_mastery_rating=2000
# gear_multistrike_rating=1276
# gear_versatility_rating=391
# gear_leech_rating=144
# gear_avoidance_rating=149
# gear_armor=1210
# set_bonus=tier18_2pc=1
# set_bonus=tier18_4pc=1

Two targets (aka doomfire/add cleave), 40 second fight (full lust):


Does a second burn phase because fight duration is only 40 seconds (not thattt bad of a thing, again we're modeling 1 doomfire strat archimonde here where p1 push is all that matters for a while)

Obviously the truth for p1 Archimonde push is a mix of those two profiles, since doomfire add is required to kill but extra damage to demon add is just extra damage

[skiz]

https://www.altered-time.com/forum/view ... =620#p7014" target="_blank

Hint: there are some hastecaps, but most of them are probably shitty. You should do those 2D Plots first, find possible errors and do global stat scaling after it on non messed up profiles, or just ignore it and take 2D plot anyway. Discussing doomfire strategies on archimonde in a PoF thread seems weird. However, I recommend fire.

[Searix]

Used as many data points as i could in as few plots as possible in 250k samples

[Curnivore]

I keep getting that kind of plot frequently and I don't know what to do with it. Values seem to jump up and down every iteration to different priorities giving the impression that in a real fight there is no point taking them seriously since the setup won't be exactly the same with the simulation anyway. I end up ignoring them and using the basic scaling method until it drops DPS.

[skiz]

Helps a lot to start with this approach:
- Where are potentional hastecaps?
- Can I reach them?
- Then do a 2D reforge plot. ( Haste vs Mastery reforge.)
- edit: you did scaling.

I.e. I would expect a hastecap at roughly 9% haste. Which is the gcd during bloodlust. If that is important. There is somewhere a list of arcane haste caps on this forum. I would bet that those 9% are the only softcap you somehow want to reach. But well, just a guess. Just give it a shot near 9% haste and with your gear?

[Frosted]

Used as many data points as i could in as few plots as possible in 250k samples

Should be a way to pull out all those data points and then just fit them to a linear equation to give you an easier time looking at it. That's basically what stat weights are giving you.

[Komma]

If you're doing a 40 second fixed_time sim, 250K iterations is nothing. Your variance is still way too high to make any meaningful judgements from the output, and it's not even good data because robomage will do a end-of-fight burn.

You didn't include the distribution or sigma values in your screenshots, but the DPS-error is ~0.2%. The default target_error for reforge plots is 0.05%, which is 1/4th of your current value. Going by the square law, this means you need roughly 250K * (4^2) = 4M iterations before you will get the usual stable output - and that's assuming you made necessary changes for the APL to behave in a deterministic, non-broken manner.

Running 4 million iterations is the simple part. Knowing that all 4 million iterations are behaving properly despite the 40 second RNG swings? Good luck with that.

[Searix]

If you're doing a 40 second fixed_time sim, 250K iterations is nothing. Your variance is still way too high to make any meaningful judgements from the output, and it's not even good data because robomage will do a end-of-fight burn.

You didn't include the distribution or sigma values in your screenshots, but the DPS-error is ~0.2%. The default target_error for reforge plots is 0.05%, which is 1/4th of your current value. Going by the square law, this means you need roughly 250K * (4^2) = 4M iterations before you will get the usual stable output - and that's assuming you made necessary changes for the APL to behave in a deterministic, non-broken manner.

Running 4 million iterations is the simple part. Knowing that all 4 million iterations are behaving properly despite the 40 second RNG swings? Good luck with that.

pretty much where im at

[Curnivore]

Looking at the big picture, when an investigation needs 4,000,000 iterations instead of 250,000 iterations to give smoother results, is it really of great value to a gamer? The gamer will unlikely do even the 1/1000 of the pulls of the 250,000 iterations. Would the results of one or the other matter to him, or will they give results that are not practically reproducible usually?

I do not know what that type of reasoning would be called in Statistics but it sounds to not be totally worthless to look at it like that.

Say for instance that for a type of boss I often do 250 pulls. Then I could see the variation of multiple 250-iteration runs.

[Komma]

By now, I think it should be clear why we haven't had any serious SimC-based deep-dive threads on discussions related to PoF.

[Curnivore]

Because I noticed I'm not the only one that gets an impression that the debuff procs more frequently on the enemy when the pull starts with a crystal right on it - I usually attributed that impression to pessimism and I didn't give it much weight - I wonder if you know if Unstable Magic somehow registers on the enemies before it registers on the crystal, because that might explain it.

[Zagann]

Im fairly sure UM registers after AB. I've had UM proc on my pre-casted AB and then have MoD proc from the UM of that AB, onto the crystal, which therefore makes it seem that it registers after the AB.

[Searix]

Because I noticed I'm not the only one that gets an impression that the debuff procs more frequently on the enemy when the pull starts with a crystal right on it - I usually attributed that impression to pessimism and I didn't give it much weight - I wonder if you know if Unstable Magic somehow registers on the enemies before it registers on the crystal, because that might explain it.

Again though the preferred strat is to avoid this altogether by dropping the crystal at -3 outside of boss range, hitting it at -1 so it's a guaranteed proc, and having tank pull it to crystal



This is a good spot on Archimonde, you can also get pretty close to it so more AMs hit at the end of crystal life, and it will normally be on doomfire spawn

That + t12 shenanigans i can do 15-20m in the first 40 seconds consistently with almost no WF pieces

[Komma]

Again though the preferred strat is to avoid this altogether by dropping the crystal at -3 outside of boss range, hitting it at -1 so it's a guaranteed proc, and having tank pull it to crystal

Napkin math says that you need to idle for ~226 seconds before you can reach a 100% proc chance on first attack. I find this impractical because I don't remember having a chance to stand idle for 4 minutes in a raid. Then there is the issue that you're wasting at least 2 additional seconds of PC, potion and AP in order to do this, and you don't even have Time Warp to speed up your casts unless you want to waste ~4 seconds of Time Warp for the entire raid. That's a lot of resources spent just to force MoD to proc on PC. Then there is the issue of needing pinpoint timing cooperation from your tank, moving the boss to the specific location right at the right second.

Where exactly is your supporting evidence for claiming that this yields a higher average return? How exactly are you justifying this as the "preferred" strat?

[Curnivore]

You know what? The "pulling a boss" bonus to RPPM trinkets might explain the "pessimism" about it. Since after a fast pull the RPPM trinket is still on a long cooldown potentially after a wipe, then it does require the actual pull itself to gain some momentum in its chance to proc, hence, that might be why the proc appears to happen very rarely on the crystal if the pull is happening really fast after a wipe with a crystal right on the boss as a first action.

edit: After all, if the mage does it early she WILL be out of combat when the first hit hits, hence there would be no bonus to the chance for certain.

It might be ideal for the mage to waste an extra second to make sure the boss is in combat before hitting the crystal, if it's right on the boss.

[Searix]

Again though the preferred strat is to avoid this altogether by dropping the crystal at -3 outside of boss range, hitting it at -1 so it's a guaranteed proc, and having tank pull it to crystal

Napkin math says that you need to idle for ~226 seconds before you can reach a 100% proc chance on first attack. I find this impractical because I don't remember having a chance to stand idle for 4 minutes in a raid.

We already ask the raid to wait 2 minutes if i'm respeccing fire for sandmans (assuming i don't do it the trash pull before the boss)

Where exactly is your supporting evidence for claiming that this yields a higher average return? How exactly are you justifying this as the "preferred" strat?

Supporting evidence is hundreds of wipes on Archimonde as the only mage, and the raid asking me to find a way to get more consistent damage. We have consistently only wanted to do the one doomfire strat (first kill 3 months ago) so being the only mage (and 0 rogues) my dps has been crucial in making the push, where we'd often be off by 0.1% and call a wipe

Then there is the issue of needing pinpoint timing cooperation from your tank, moving the boss to the specific location right at the right second.

You're right of course that it's stressful on your tanks, but we were desperate to try anything to get consistent dps.

Then there is the issue that you're wasting at least 2 additional seconds of PC, potion and AP in order to do this, and you don't even have Time Warp to speed up your casts unless you want to waste ~4 seconds of Time Warp for the entire raid. That's a lot of resources spent just to force MoD to proc on PC.

I'd dare say it's even a dps increase. Using t12 shenanigans we're often at, what, 70% mana when ring pops? Doing this method (and ring on pull) we're exchanging a 0 stack AB hit on crystal, for a 4 stack AB hit (or two) on boss with a lot of mana.

---

For reference when learning Archimonde a good benchmark was I wanted to do 15m damage during the first bloodlust. However when it procced on the boss I would often get 11-14m damage, compared to 13-20m consistently with the new prophecy of fear strategy always proccing on crystal. We went from occasionally making the push to consistently making it after we made the change.

Few notes: Was using mythic (non-warforged) staff off xhul, mythic (non-warforged) class trinket, and 744 ilvl.

Example VOD: https://www.twitch.tv/searix/v/68381770" target="_blank