+% Attack Speed vs weapon speed

I’m working on a blademaster (as my 7th character) and with all the melee procs, it looks like attack speed will be critical. What is the relationship between +attack speed on items, and the speed of the weapons themselves? Like, which is more beneficial? Would a slow mace with +10% attack speed be better than a very fast dagger, all other things being equal? You get the idea.

IIRC Very fast weapons scale better on attack speed. Their base attack speed is higher, therefore will scale better with attack speed.

Yeah, they scale better. But if you had to choose between a slow weapon with +attack speed stat on it, or a fast weapon with no such stat, which would you pick?

The one that gives more stats I need.

I mean specifically with regard to attack speed. Like, what is the % difference in attack speed between a fast and very fast weapon? See what I’m getting at?

Varies from weapon to weapon. The ‘Very Fast’ or ‘Average’ text is arbitrary and masks the real AS multiplier. For instance, this Imperial Warhammer has -0.13 AS, but this Imperial Warmace has -0.16 AS, while each of them are listed as ‘Slow’.

As far as a % difference, it’s tough to say, because these values are multipliers whereas % Attack Speed is additive. It really depends on how much % Attack Speed you have, as a lower multiplier will get less AS per %AS than a higher multiplier would.

There’s a range so you’ll have to look at something like gracefuldusk to determine the exact multiplier.

In addition weapon types also have a hidden secondary speed penalty. For example a fast pistol has a significantly greater speed penalty than a fast sword.

1H melee weapons including daggers and scepters have a 100% gross baseline attack speed. All other weapons have a 92% gross baseline attack speed.

The weapon speed modifier (which is negative) is added to the gross baseline attack speed to yield the net baseline attack speed. Total attack speed bonus is then multiplicative with net baseline attack speed.

Example: A Scrapmetal Musket has a gross baseline attack speed of 92%. The weapon speed modifier (as listed by Graceful Dusk) is -0.2. The net baseline attack speed is therefore 92% + (-0.2) = 72%. This is verifiable in-game.

But wait, there’s more:

If you spam an attack, the delay between attacks does not remain constant. There is a random deviation in attack delay, even if the attack animation stays constant. This is readily verifiable by spamming ABB with a rifle at 74% attack speed. Sometimes you can proc ABB exactly every 3rd shot. But sometimes your 3rd shot occurs just a tiny bit before the ABB cooldown ends, which forces an additional shot to proc ABB. This problem does not occur at 73% attack speed.

As you might suspect, one of the greatest applications to studying the exact nature of attack and cast speed is to optimize interaction with cooldown timers. ABB again is a great example. Given a certain CDR, how much attack speed should I have to eliminate extraneous shots? You can apply the same thinking for spam skills with attached devotion skills with low cooldown. For example, CT with Flame Torrent or Cadence with Bull Rush.

But wait, there’s even more:

I believe there is a “button press delay” for cooldown skills such that it would be impossible to “keep the clock hand constantly rotating.” That is, the cooldown timer doesn’t start until a certain delay after you’ve pressed the button for that skill. I don’t know exactly what it is or what affects it, if anything. If you are holding down the button it still occurs. I don’t know if this issue affects devotion skill cooldown timers but intuitively it probably doesn’t.

But wait, if you can believe it, there’s even more:

Certain animations, like that of 2H melee Markovian’s Advantage, have an animation time that is noticeably longer than a normal attack’s. Therefore, if you can theoretically proc 2H melee MA 100% of the time, your effective attack speed will be lower than spamming 100% normal 2H melee attacks, even if your sheet attack speed remains constant.

So, as you can see, there are still plenty of unknowns regarding the exact nature of attack/cast speed, cooldown timers, and animations in Grim Dawn.

Dunno if it’s related, but my battlemage (and some other classes) attacks significantly faster alternating auto attacks and spells, than simply auto attacking or casting spells separately.

In the same time frame (it’s actually even faster but it becomes less noticeable as you get more attk speed and cast speed, maybe because you can’t just click fast enough) it takes me to cast 2 AA-s back to back, I get to cast AA-Calidor’s Tempest-AA, which obviously at least doubles my damage output.

Same result with my Pyromancer alternating AAs and Fireball.

If I had to guess, it’s similar to animation cancel, by casting a spell, you reset the AA animation which in return enables faster combos.

That could also be tied to the way attack speed and cast speed interact with eachother, being 2 different stats.

There was a topic about it, but it only mentioned component skills I believe, when it most likely affects all skills with no significant to no cast animation.

It’s not similar to animation canceling, it is animation canceling. It’s also an exploit.

Similarly, the attack delay that Dioarchet mentioned is a bug with m1 and m2 inputs that was introduced back in B24 I’ve been complaining about it since it was first introduced.
Funny enough the delay doesn’t occur when you attack with keyboard slots, which is how I know it’s a bug besides it being really annoying. It’s likely an issue with how skill queuing works for m1 and m2.

At this point you probably need a mouse with a third button set to do a LMB+RMB action with a customizable delay inbetween and afterwards.

But attack/cast speed alternation also works with cooldown skills too, if you can somehow find a way to make that work (I have ;)).

I put Flashbang on a keyboard hotkey, held it down, and recorded a video of the Flashbang icon.

Then I did a frame-by-frame analysis of the footage.

The button press delay, as visually represented by the Flashbang icon looking “depressed,” lasted about 200 ms. Afterward, the “cooldown clock hand” appeared very close to the 12 o’clock position.

If the button press did not affect the cooldown timer, that clock hand would have appeared close to the 3 o’clock position instead (I had about 20% CDR).

So when the hotkey is held down, the delay does occur.

But, I don’t think the delay occurs when the hotkey is pressed just once. When I do that, the button press depression appears to last longer, but the clock hand starts to appear at well past the 3 o’clock position.

I’ll do more analysis of the footage at some point. These are just preliminary findings.

It’s far easier to tell with a spammable skill. I tested it using pblades with 200% casting speed. Casting it for a minute straight would give me around 3 less casts than casting it with a key input within the same time frame. If a delay does occur with the keyboard it is far less prominent.

I was testing specifically cooldown skills that have a button depression effect; therefore I did not test spammable skills. I think you may be testing a different effect.