Sweep

"
Silty wrote:
I love using this skill, but I hate how it suddenly becomes useless in Chaos, since I used it as my main attack through the whole game up till then.

The problem is that Sweep demands a somewhat tanky build in order to gather a cluster of monsters and take their blows as you pummel them. But tanky builds become impossible without a shield in Chaos, so it kind of becomes useless in Chaos because it doesn't work no matter what build youhave. Maybe quality on this gem could cast a 2-3 second self-buff that gives some armor.


Use molten armor + warlord's mark and you'll be just fine.
"Life is a tragedy when seen in close-up, but a comedy in long-shot." - Charlie Chaplin
"
bkboggy wrote:
"
Silty wrote:
I love using this skill, but I hate how it suddenly becomes useless in Chaos, since I used it as my main attack through the whole game up till then.

The problem is that Sweep demands a somewhat tanky build in order to gather a cluster of monsters and take their blows as you pummel them. But tanky builds become impossible without a shield in Chaos, so it kind of becomes useless in Chaos because it doesn't work no matter what build youhave. Maybe quality on this gem could cast a 2-3 second self-buff that gives some armor.


Use molten armor + warlord's mark and you'll be just fine.

Yeah, I've gotten used to cycling endurance charges, immortal call, and warlord's mark - it's a bit more usable that way. However, I also need a lot more life leech gear, because of the 0.9.9 life leech nerf. You also need Kaom's on top of all of those things for it to be usable! Crazy.

Molten shell doesn't work at all, though, since it only lasts about a second before blowing up. I end up having to keep recasting it more than I sweep, at which point you either have to give it up or make it your main attack.
In my opinion, sweep should do less damage and more knockback distance.
You get what you deserve...
This skill is interesting because it completely obliterates everything from normal to ruthless, and then suddenly just stops working. I guess it's because fast weapons and elemental damage are just more viable lategame options than slow physical damage weapons. Of course you can add elemental damage to sweep too but it's much less effective given the low attack speed.

It's a shame - I really wanted to build a tank with massive life leech who could stand in a throng of chaos mobs and swing his mace like a madman, but in the end he ended up kiting just like everyone else.
It's ok in merciless.

I have a physical damage melee witch who uses it fine in fellshrine. I'd need way better gear and to move some points into defensive masteries to hazard chaos though :(
I would like to confirm with the developers: sweep's speed is independent of base weapon speed, correct? According to some quick tests, it looks to me like this is the case.

Assuming this is true, I suppose that damage is based on weapon damage, rather than damage per second, correct?

Sweep's speed, however, is affected by ias on items. Assuming it is also properly affected by skill ias, its base speed appears to be something like 0.87 attacks/sec.

Could I suggest that on skills that do not use weapon seed (sweep, and I believe ground slam?), the skill descriptions reflect this? Say, "Sweep attacks 0.87 times per second, independent of base weapon speed but affected by increased attack speed."

Could I also ask what the design decision is for making sweep and other skills have their own (hidden) attack rate?

Finally, I believe attacks may be ~1% slower than they appear on the character screen. More careful tests may confirm this.
Last edited by FrederickHermanJonesJr on Aug 11, 2012, 4:01:59 AM
"
I would like to confirm with the developers: sweep's speed is independent of base weapon speed, correct?
Correct

"
Assuming this is true, I suppose that damage is based on weapon damage, rather than damage per second, correct?
Assuming I understand what you mean, then yes. All attack damage is based on weapon damage. Nothing is ever based on DPS, DPS is just calculated form damage/attack speed/hit chance/crit chance to be displayed as an average to the user. That value isn't (and couldn't reasonably) be actually used for anything. It's also wrong for sweep because it doesn't know about this behavior, so will be using weapon base attack time.

"
Sweep's speed, however, is affected by ias on items. Assuming it is also properly affected by skill ias, its base speed appears to be something like 0.87 attacks/sec.
1.15 seconds base attack time, which works out at 0.869ish attacks per second.

"
Could I suggest that on skills that do not use weapon seed (sweep, and I believe ground slam?), the skill descriptions reflect this? Say, "Sweep attacks 0.87 times per second, independent of base weapon speed but affected by increased attack speed."
This may be added in future, I'm not sure. We will fix the displayed attack time on the popup in future to account for this behavior.

"
Could I also ask what the design decision is for making sweep and other skills have their own (hidden) attack rate?
Leap slam is the most obvious example. There's no way to make an animation of a guy quickly crouching, leaping into the air, coming down again and crashing his weapon into the ground, which looks even vaguely decent if that entire animation takes the same time as it takes said guy to make a single quick strike with his weapon on something in front of him.
The same is true of swinging a two-handed weapon in a full circle, to a lesser extent. I don't know why it's that way on ground slam, that was before my time - I suspect it's the same consideration, because ground slam has it's own animation, but I think (but haven't tested) we could make ground slam use the base weapon attack time without it looking too silly.

There are also balance reason, especially with leap slam (if the entire leap took the time it takes to make a melee attack, it would be completely broken and a great way to get around).
"
Finally, I believe attacks may be ~1% slower than they appear on the character screen. More careful tests may confirm this.
Could you please provide the reasoning and evidence that leads you to this conclusion?
Thanks for the rapid and detailed response! My replies are below.

"
Mark_GGG wrote:
... All attack damage is based on weapon damage. Nothing is ever based on DPS, DPS is just calculated form damage/attack speed/hit chance/crit chance to be displayed as an average to the user. That value isn't (and couldn't reasonably) be actually used for anything. It's also wrong for sweep because it doesn't know about this behavior, so will be using weapon base attack time.

Thanks for clearing that up!


"
Mark_GGG wrote:
"
Sweep's speed, however, is affected by ias on items. Assuming it is also properly affected by skill ias, its base speed appears to be something like 0.87 attacks/sec.
1.15 seconds base attack time, which works out at 0.869ish attacks per second.

Glad to see my estimate lined up well (more on the method below).

"
Mark_GGG wrote:
"
Could I suggest that on skills that do not use weapon seed (sweep, and I believe ground slam?), the skill descriptions reflect this? Say, "Sweep attacks 0.87 times per second, independent of base weapon speed but affected by increased attack speed."
This may be added in future, I'm not sure. We will fix the displayed attack time on the popup in future to account for this behavior.

Sounds good.

"
Mark_GGG wrote:
"
Could I also ask what the design decision is for making sweep and other skills have their own (hidden) attack rate?
Leap slam is the most obvious example. There's no way to make an animation of a guy quickly crouching, leaping into the air, coming down again and crashing his weapon into the ground, which looks even vaguely decent if that entire animation takes the same time as it takes said guy to make a single quick strike with his weapon on something in front of him.
The same is true of swinging a two-handed weapon in a full circle, to a lesser extent. I don't know why it's that way on ground slam, that was before my time - I suspect it's the same consideration, because ground slam has it's own animation, but I think (but haven't tested) we could make ground slam use the base weapon attack time without it looking too silly.

There are also balance reason, especially with leap slam (if the entire leap took the time it takes to make a melee attack, it would be completely broken and a great way to get around).

Ah, I see. I agree that ignoring weapon speed is certainly reasonable for a skill like leap slam. However, to me at least, for skills like ground slam or sweep, it seems more intuitive if base weapon speed does affect the attack speed. That is certainly what I originally expected. Is it possible within the game code for the base attack speed for the skill be related to the base weapon speed? It could be reported in the skill description as something like: "The base speed for sweep is 20% slower than your base weapon speed."

"
Mark_GGG wrote:
"
Finally, I believe attacks may be ~1% slower than they appear on the character screen. More careful tests may confirm this.
Could you please provide the reasoning and evidence that leads you to this conclusion?

Sure. In brief, I wanted to see what the base speed for sweep was and how it changed as I swapped weapons. To do so, I recorded the sound of a character attacking (shift+click) for one minute for several skills: normal attack, heavy strike, lightning strike, and sweep. I then took a fast fourier transform of each recording, and I found that the peaks did not line up with what was on the character screen. They were all ~0.01-0.02 Hz too slow.

After I saw you wrote this response, I repeated the test. This time, I just used normal attack, and I recorded for 10 minutes to improve the frequency resolution. The character I tested with had no ias, no allocated skill points, etc., and for clarity, I used a two-handed weapon with exactly 1.00 (nominal) attack per second. I also tested with an online metronome to make sure there wasn't a problem with how I performed the FFT.

Similar to my previous tests, the attack speed I measured from the FFT was ~0.987 Hz, rather than 1.000 Hz. This value seems suspiciously close to 1/(attack time + time per frame), but that might just be coincidence. Could this somehow be related to latency?

If it would be helpful, I am happy to share the quick code to do the fft, as well as plots and sound files. However, I'm guessing that these might be better to email.
Last edited by FrederickHermanJonesJr on Aug 13, 2012, 3:10:25 AM
As a followup, I thought it might be handy to illustrate with a couple plots. More info is in the spoiler.

Spoiler

The Fast Fourier Transform (FFT) is a quick way to find important frequencies in time-series data. To illustrate, here is an demo train of "attacks":


Fullsize

Each attack is spaced 1 second apart, and the attack duration is 70 ms. This is just a small fraction: the series continues after 9 seconds.

Taking the FFT of this series shows a clear peak at the attack speed of 1 attack per second (1 Hz = 1/s):


Fullsize

The additional, evenly spaced peaks are due to the shape of the "attack" pulse: our attack is not simply a sine wave of frequency 1 Hz.

A recorded sound file is not as simple as the demonstration "attack" train above, but it's still not hard to pick out the attack speed from the FFT. After tossing one of the stereo channels (and taking an absolute value or two), here is the result:


Fullsize

The blue trace is the FFT of a 10 minute recording of (nominally) 1.00 attack per second. The green trace is the FFT of a 10 minute recording of a metronome at 60 beats per minute (=1 beat per second). The FFT of the metronome nicely shows 1 beat per second, but the attack speed is a little slower than expected: 0.987 Hz.

Again, I am happy to send more information, code, etc. if anyone's interested. This is probably already too much for a post!


Thanks again for your rapid responses and your the obvious attention you give to the community in the forums.
Last edited by FrederickHermanJonesJr on Aug 14, 2012, 12:19:17 AM
I'm not going to pretend to understand everything you've done there, but the discrepancy looks small enough that I suspect it's not really a problem, and is probably and artifact of how timing works in the game engine. There's always going to be a little bit of processing happening before an attack actually starts playing an animation, etc.
Thank you for putting so much effort into this.

Report Forum Post

Report Account:

Report Type

Additional Info