Wake of Fire Inquiries and Examinations

[Socialism]

Hi.

Wake of Fire is so damn cool, and I really like it because of its nifty animation and application to targets. Damage, unfortunately, is the core of the issue. It's really, really small. Not only that, but '4-frame nextdelay' adds insult to injury by ignoring a good chunk of the actual fire waves, so not all of them hit - in fact, generally speaking, less than half will hit.

Damn you, Blizzard developer responsible for game balance.

This thread uncovers optimal Wake of Fire trap laying speeds in an attempt to work around 'nextdelay', maximizing potential Wake of Fire damage per second (the tiny amount of it there is). Gogo Excel.

Note: Yet another 'signature' long read. Skip to the end for scientifical conclusions.

Dissected a recording of Wake of Fire I made, frame by frame (VirtualDub (http://virtualdub.com/) is superbestawesome).


Animation Analysis
The total duration of a Wake of Fire trap is either 136 frames [when graphical 'remnants' of trap become solid] or 144 frames [last frame of the dead trap's shrinking light radius].


Wake of Fire animation sequence
Frame 1: the moment the trap graphic appears after casting
- Frame 2-26: Trap Delay
Frame 27: trap pivot animation
Frame 28: Flame1 graphic appears
- Frame 29-44: Trap Delay
Frame 45: Flame2 graphic appears
- Frame 46-61: Trap delay
Frame 62: Flame3 graphic appears
- Frame 63-78: Trap Delay
Frame 79: Flame4 graphic appears
- Frame 80-95: Trap Delay
Frame 96: Flame5 graphic appears
- Frame 97-112: Trap delay
Frame 113: Trap break animation begins
Frame 136: Trap break animation ends
Frame 137: Trap light radius shrink begins
Frame 144: Trap light radius shrink ends


-----------------------


General assumptions, or making methods of experimentation scientitious
*Note: "Volley time" is calculated from the time the first trap laid appears until the last trap laid begins its break animation.


Assumptions
1. I'm laying five Wake of Fire traps
2. with maximized synergies and +12 to trap skills (arbitrary, reasonable endgame estimate)
3. without interruption (i.e. holding down the right mouse button to lay all 5 traps)
4. to isolate trap laying 'speed' (FPA = frames per action/attack) as the only variable.


Damage, for purposes of this analysis
Fully synergized skill level 32 Wake of Fire does 634-722 fire damage; ~678, rounded to 680 for convenience.

At 9FPA, the fastest trap laying speed possible, only 13 of the 25 total 'waves' are outside the nextdelay cancellation periods.

So at 9FPA, for every ~149 frames (end of break animation of the last trap in a 5-trap Wake of Fire volley) or 5.96 seconds,
A. Your average damage is 680 x 13 = 8840 damage
B. Which means you do roughly 8840 / 5.96 = 1483 damage per second

Still, each wave only does ~680 average damage; this means you have no chance of stunning anything in a 2+ player Hell game. That. Sucks. Ok. Fine.


-----------------------


Some handy Excel sheeting results
9 FPA: 13 of 25 are unaffected by nextdelay (outside of 4 frames). Total volley time: 149 frames.
Damage per second: (13 x 680) / (149/25) = ~1483 damage per second

10 FPA: 16 of 25 are unaffected by nextdelay (outside of 4 frames). Total volley time: 153 frames.
Damage per second: (16 x 680) / (153/25) = ~1778 damage per second

11 FPA: 19 of 25 are unaffected by nextdelay (outside of 4 frames). Total volley time: 157 frames.
Damage per second: (19 x 680) / (157/25) = ~2057 damage per second

12 FPA: 19 of 25 are unaffected by nextdelay (outside of 4 frames). Total volley time: 161 frames.
Damage per second: (19 x 680) / (161/25) = ~2006 damage per second

This is where it gets kinda interesting, because I'm not sure how/when nextdelay applies.

I don't know if "4 frame nextdelay" follows Scenario A or Scenario B:

Scenario A
Frame 1 Attack hits, nextdelay counter: 1
Frame 2 nextdelay counter: 2
Frame 3 nextdelay counter: 3
Frame 4 nextdelay counter: 4
Frame 5 Attack hits, nextdelay counter: 1
etc., etc.

In this scenario:
13 FPA: 23 of 25 are unaffected by nextdelay (outside of 4 frames). Total volley time: 165 frames.
Damage per second: (23 x 680) / (165/25) = ~2370 damage per second

Scenario B
Frame 1 Attack hits
Frame 2 nextdelay counter: 1
Frame 3 nextdelay counter: 2
Frame 4 nextdelay counter: 3
Frame 5 nextdelay counter: 4
Frame 6 Attack hits
Frame 7 nextdelay counter: 1
etc., etc.

In this scenario:
13 FPA: 15 of 25 are unaffected by nextdelay (outside of 4 frames). Total volley time: 165 frames.
Damage per second: (15 x 680) / (165/25) = ~1546 damage per second

14FPA: 15 of 25 over 169 blah blah I don't care, not optimal.

15FPA: 15 of 25 over 173 blah blah I don't care, not optimal.

16FPA: 11 of 25 over 177 blah blah I don't care, not optimal.

17FPA: if you're here, don't bother casting more than one Wake of Fire at a time with the 'no interruption' method; 4 of your 5 traps are rendered completely useless.


---


This 13-frame-optimization 'discovery' begs the question: when does nextdelay actually start?


Conclusions
Here you go. This is for you, dear reader, who so desperately wants to use Wake of Fire, but always gets put off by the ostensibly low damage output. In any event, perhaps this will help optimize the "Nova" of our trap skills.

If '4 frame nextdelay' Scenario A applies, you want to be laying traps at 13FPA against monsters.

If '4 frame nextdelay' Scenario B applies, 11FPA is your most damaging trap lying speed.




Cheerio,

Socialism
-achieving Rockstar status, one day at a time.

[Socialism]

Read Ruvanal's theory on the LurkerLounge that stipulates scenario A is the case.

Logic for that, in sum:

Attacks like Nova and War Cry aren't 'really' circular - the game graphic displays at that, but the damage mechanic is actually very many tiny missiles that shoot out at a gradually expanding range (i.e. think of a multishot fired very close to the Amazon). Ruvanal's argument is that if nextdelay didn't apply in the frame of initial contact, enemies very close to a Sorceress using Nova would be hit by multiple Nova-missiles, doing several times the listed Nova damage -- same goes for War Cry.

Interestingly enough, that's also apparently how Battle Command/Orders/Shout works. Since the Barbarian graphically at the source of the expanding 'Nova' of missiles, he'd be getting the effect of the skill applied for each one...


Anyway. Took a video of 13FPA to see if enemy life is reduced in 23 increments.

I am correct.

Frame by frame, an enemy's life went down 22x (I'm pretty sure I missed the first one mousing over to the name... I was getting kicked around by wee Fallen) from my 5 casts of WoF. Nifty. Scenario A is the case; 13 FPA is Wake of Fire's most optimal "no interrupt" trap laying speed.

[wizAdept]

Very interesting read. Congrats on the research I like this type of stuff.:thumbup:

Now for a my questions:

This was taken from TienJe's trapper guide:
Trap Analysis

Lightning Sentry
Initial Delay – 33 frames
Firing Speed – 30 fpa
NextDelay – None
Shots – 10

Wake of Fire
Initial Delay – 40 frames
Firing Speed – 17 fpa
NextDelay – 4 frames
Shots – 5


He told me himself he never got a real confirmation on those numbers although they have been cited before elsewhere, these numbers on wof do not agree with your fraps footage.

I also must ask, which video recording device do you use? fraps or hypercam or something else? some of these recording programs may skip frames while recording which would mess tests up.


Another question I have, does trap setting speed have actionframe breakpoints? What is the actual action frame in the trap setting animation when the trap is laid down?

Perhaps a combination of trap setting (+action frame delay + trap delay) with 9 or 10 frame mindblasts could force nextdelay on certain frames improving wof damage per second with 9 or 10fpa trap casting?

Although this sounds nice in theory another thing to take into consideration is nextframe afaik does not start on (or after) the frame the trap shoots, but the frame (or after) that it hits the opponent. Im thinking more pvp applications here than vs monsters, so players that walk out of wof will inevitably change the time at which the wof waves strike them by a frame or a few, which can mess up attempts at forcing nextdelay timing.

[wizAdept]

Just tried it vs some players, had a high hp barb stand still while I set traps under him, tested 9fpa vs 11fpa and 12fpa setups and saw no difference in damage from 5 consecutive traps set with any of the different trap setting speeds.

Anyways there are too many other variables such as moving opponents to really make use of forced nextdelay timing. That and even with opponents standing still as my in game testing showed me no difference in dps between fast trap setting speeds and slower ones. The barb seemed to take the same damage from the traps regardless of which trap speed I used to lay them, give or take an additional wave here or there which really isnt enough to make use of this.

[dainbramage]

Another question I have, does trap setting speed have actionframe breakpoints? What is the actual action frame in the trap setting animation when the trap is laid down?

Perhaps a combination of trap setting (+action frame delay + trap delay) with 9 or 10 frame mindblasts could force nextdelay on certain frames improving wof damage per second with 9 or 10fpa trap casting? Action frame breakpoints would exist, though I have no idea what they are. Either way, the position of the action frame won't change the results (at least under the conditions in the OP) - there will still be the same gap between two traps being laid.

Although this sounds nice in theory another thing to take into consideration is nextframe afaik does not start on (or after) the frame the trap shoots, but the frame (or after) that it hits the opponent. Im thinking more pvp applications here than vs monsters, so players that walk out of wof will inevitably change the time at which the wof waves strike them by a frame or a few, which can mess up attempts at forcing nextdelay timing.
Yeah, the counter starts the frame it hits the target, so moving will screw it up. The OP was directed at pvm uses though, even if you find a use, I doubt halving your trap-laying speed would pay off at any rate.

[Socialism]

Very interesting read. Congrats on the research I like this type of stuff.:thumbup:

Now for a my questions:

This was taken from TienJe's trapper guide:


He told me himself he never got a real confirmation on those numbers although they have been cited before elsewhere, these numbers on wof do not agree with your fraps footage.

I also must ask, which video recording device do you use? fraps or hypercam or something else? some of these recording programs may skip frames while recording which would mess tests up.


Another question I have, does trap setting speed have actionframe breakpoints? What is the actual action frame in the trap setting animation when the trap is laid down?

Perhaps a combination of trap setting (+action frame delay + trap delay) with 9 or 10 frame mindblasts could force nextdelay on certain frames improving wof damage per second with 9 or 10fpa trap casting?

Although this sounds nice in theory another thing to take into consideration is nextframe afaik does not start on (or after) the frame the trap shoots, but the frame (or after) that it hits the opponent. Im thinking more pvp applications here than vs monsters, so players that walk out of wof will inevitably change the time at which the wof waves strike them by a frame or a few, which can mess up attempts at forcing nextdelay timing.

Environment:
FRAPS, skip-free-lag-free; socket 775 3.3Ghz dual core running perspective-free Direct3D on a pair of ATI Rad DDR3 graphics cards; separate HDs for OS, pagefile, and applications/storage; XP SP3, matching 4GB (2x2) PC2-6400 sticks, ergo synchronous up-down. Timings like 5/5/5/15 or something.

Test locations:
Single Player
Battle.net USEast Hardcore Ladder server .162


I have no idea who TienJe is, or what guide you're referring to (so lazy, can't be bothered to click stickies right at this moment).

Frame-by-frame observation lined up perfectly with formula predictions and held true over half a dozen sets of observation -- no deviations -- which was enough for me to discount any assumed margins of error in the recording output. I think that is what you're hinting at by posting that someone else's claims-on-frames don't match my observations - do correct me if I'm wrong.

I'm not sure what you're referring to with actionframe breakpoints. I assume you're asking about IAS% breakpoints affecting changes in # of frames per action, so yes - all animation speed related things on D2 rely on 'breakpoints' to have any impact on # of frame reduction. As to the actual action frame -- it is the frame immediately after the listed FPA. So at a 13FPA trap laying speed, the trap displays on the ground at frame 14.

Running sets of 9fpa - 13fpa setups, frame-by-frame watching enemy life bars matched the numbers of assumed hits in the DPS calculations in my first post, though recent back-and-forth posts with AmazonBasin folk in the assassin forum there have led me to believe it's not for the reasons I think it's for (having to do with the # of missiles actually firing out per wave, their length/duration etc.).

You misunderstand why I'm lining up nextdelay assertions with wave launch animations; it's not because I believe that nextdelay counters start when the traps start shooting, but because I have demonstrable evidence that the waves travel at constant speed, i.e. if 2 waves launch 6 frames apart, they will hit an enemy (regardless of how closely they're standing to the trap) anywhere in the line of fire 6 frames apart.

With PVP, the problem is that the overlap of traps aren't optimized for avoiding nextdelay -- so waves that 'cancel' and fall outside of the FHR frame rates of your enemies. When stunned, it's my understanding that mobiles will continue to be stunned if they take any damage while in the hit recovery animation; is this accurate?

With 9FPA waves, there are a few gaps of 8-11 frames, more than enough time for someone with 6fhr to break from FHR animation -- and assuming each of your waves doesn't do 1/8th their life total, stay out of FHR animation without the aid of mindblast.

I'm really freaking sleepy. Pardon if some of this is scattered.

And painhaemorrhage is right - your trap laying speeds don't filter into how fast the waves themselves travel. IAS tweaks are only geared towards optimization of Wake of Fire's overlapping fixed (17frame delay-to-launch) wave trajectories to minimize dissolution by nextdelay.