Melee Cadence Mechanics [v1.2.1]

1. internal cooldown (ICD) mechanic

Cadence has an internal cooldown (ICD) mechanic. Whenever you gain a charge, you must wait for ICD to get next charge, or to consume the charges and launch the 3rd Cadence attack.

ICD for melee Cadence is 0.35 seconds now (in v1.2.1) and was 0.5 seconds before.

ICD scales with Attack Speed (AS). At high AS, character attacks faster, and ICD become shorter as well. So high AS won’t reduce charges or reduce frequency of Cadence in theory.

————————————

2. base-Cadence and WPS-Cadence

When character keeps 2 charges and starts a new attack, the game will check whether the ICD has been over. If ICD has already ended, then, the next Cadence will use “Cadence animation” (hereafter called base-Cadence).

For 1-hand or 2-hand, Cadence does have its own special animation, but it does not for dual-wield. “Cadence animation” for dual-wield actually means base attack animation, 25% for left-hand base attack, 25% for right-hand base attack and 50% for double swing.

On the contrary, when the new attack starts within ICD, “Cadence animation” is forbidden and “WPS animation” can be used (hereafter called WPS-Cadence).

So, we have Markovian-Cadence, Zolhan-Cadence, Belgothian-Cadence and many other WPS-Cadences.

————————————

3. fast charge for dual-wield

ICD works between the 1st charge and the 2nd charge, and between the 2nd charge and the 3rd Cadence attack. However, from the 3rd Cadence attack to the 1st charge of the next round, there is no ICD, meaning that you can gain one charge immediately after the 3rd Cadence attack, or even at the same time if two hits occur in the same frame.

Specifically, when the Cadence is dual-wield WPS-Cadence or double swing base-Cadence, a fast charge is generated for next round.

Base-Cadence has only 50% probability for a fast charge (50% for single hand attack and no fast charges, 50% for double swing and one fast charge), while WPS-Cadence has 100% probability, so higher probability of WPS-Cadence means more frequent Cadence, which is the key of dual-wield Cadence Builds.

In previous versions (v1.2 and before), ICD was 0.5 seconds and longer than the recovery time (interval time from Hit Frame to Start Frame of the next attack) of most WPS. Characters with 100% WPS can keep the loop of “WPS → WPS-Cadence → WPS → WPS-Cadence” for a long time (nearly 50% attacks are Cadence), much more frequent Cadence than normal loop of one Cadence per 3 attacks (33% attacks are Cadence).

In v1.2.1, the shorter ICD makes things different. Some WPS with hit frame at late time can still guarantee the next Cadence to be WPS-Cadence, while WPS with early hit frame cannot. See the following sheet.

WPS easily followed by WPS-Cadence
  • Nec Edge, Smite and Korvaak Brand (yellow backcolor), their hit frame is late, and the following Cadence has a high probability to be WPS-Cadence, thus high prob for a fast charge. Very few exceptions in hours of tests.
WPS hardly followed by WPS-Cadence
  • MA, Execution, Zolhan, AQC and Reaping (blue backcolor), their hit frame is too early and the ICD will have already ended when the next attack starts, so the next Cadence can only be base-Cadence, with only 50% probability for a fast charge.
WPS in between
  • Whirling and Bel’s Shears (green backcolor), their interval time is close to 0.35 seconds. I tested at various Attack Speed, and reached different results. Sometimes their following Cadences were mainly WPS-Cadences, but sometimes were mainly base-Cadences.

  • One possible reason is that Animation frames must be integer. When the theoretical interval frames are decimal, the actual interval will round to integer and therefore may become longer than ICD. For example, 8.5 frames round to 9 frames.

  • Another possible reason is random delay (or called random pause). When I slowed test videos frame by frame, I found that there might be short pauses between attacks. The duration for pauses were not fixed, sometimes a little longer (3-4 frames of 60 fps video), sometimes shorter (1-2 frames) or even no pauses. On average, pauses with keyboard are usually shorter than pauses with mouse (LMB or RMB).

  • Both reasons can make the actual interval longer than theoretical interval, and also longer than ICD.

  • I also found Vertical Sync having an impact. There are significant fewer WPS-Cadences when Vertical Sync is on.

Here is a video comparing 0.35 seconds ICD and 0.5 seconds ICD. Dual-wield Blade Master with 20% Zolhan, 20% Shears, 20% AQC, 20% Whirling and 20% Execution.

GT link:Blademaster, Level 32 (GD 1.2.1.2) - Grim Dawn Build Calculator

In left part, ICD is 0.5 seconds (v1.2 and before). The number of WPS-Cadence is high and about 46% attacks are Cadences.

In right part, ICD is 0.35 seconds (v1.2.1). Only a few WPS-Cadences, and about 39% attacks are Cadences. That is the gap between two different ICDs. (p.s. In this video, only Whirling leads a following WPS-Cadence, and Shears does not. But in some of my other tests, Shears could do.)

Assume that base-Cadence deals 550%+100%=650% Weapon Damage, normal WPS is average 150% per hand and 300% for twice hit, and WPS-Cadence is 550%+150%=700%. Average Weapon Damage of dual-wield Cadence is about 480% for 0.5 seconds ICD and 440% for 0.35 seconds ICD.

In conclusion, shorter ICD reduces the number of WPS-Cadences, and thus reduces fast charges and frequency of Cadence. It is a nerf for dual-wield Cadence.

Cadence icon bug
  • By the way, here is an icon bug. When double hits of WPS-Cadence are in the same frame, triggering Cadence and fast charging, the Cadence icon won’t disappear, though charges have been indeed consumed and charging for next round has started. Attack Jump WPS (MA and Execution), Triple Hit WPS (AQC) and double swing attack have hit frames at different time, avoiding this bug. Other types of WPS all hit twice in the same frame, causing this icon bug. This is just a display issue, no actual impact on Cadence.

————————————

4. failed charge and failed Cadence

When a late-hit attack followed by an early-hit attack, like “Nec Edge → Reaping” in v1.2, the internal between two hit frames is shorter than ICD, meaning that the second attack will fail to charge, and one more attack is needed to trigger Cadence. If the first attack is the second charge, then the second attack will fail to trigger Cadence, and Cadence will be delayed until the third attack.

In v1.2.1, the ICD is shorter so failed charge problem becomes much rarer. In previous versions, when ICD is 0.5 seconds, the problem was more common, especilly for Death Knight (like Leviathan Cadence DK and dual-wield Warborn DK).

a rare situation of failed charge in 1.2.1 (ICD 0.35s)
  • In v1.2.1, failed charge problem can occur in specific situation, “double-swing base Cadence → Reaping or Markov or right-hand base attack (if not 100% WPS)”. Double-swing base Cadence will consume charges and trigger Cadence at the first hit and generate a fast charge at the second hit. However, the hit frame of its second hit is really late, and only 0.1 seconds are left before the next attack. Additionally, Reaping, Markov and right-hand base attack have an early hit frame. The theoretical interval is 0.333 seconds, shorter than ICD, thus failing to charge.

  • This situation is not stable. As mentioned above, some reasons can make actual interval longer than theoretical interval, and then successful in charging. But I did observe failed charge several times.

  • On the plus side, only dual-wield Cadence may meet failed charge problem now, no longer for 1-hand and 2-hand Cadence to meet it.

Here is another comparing video. Leviathan Death Knight with 25% Markov, 25% Zolhan, 25% Reaping and 25% Nec Edge.

GT link: Death Knight, Level 100 (GD 1.2.1.2) - Grim Dawn Build Calculator

Comparing left part and right part, the character does 2 or 3 less Cadences every 100 hits at 0.5 seconds ICD than at 0.35 seconds ICD

As above, assume that Cadence deals 550% Weapon Damage and WPS is average 150%. Average Weapon Damage of 2-hand Cadence is about 270% for 0.5 seconds ICD and 280% for 0.35 seconds ICD.

In conclusion, shorter ICD reduces failed charge and failed Cadence, especially for 1-hand and 2-hand Cadence. It is a buff for 1-hand and 2-hand Cadence.

————————————

5. better ICD

1-hand and 2-hand Cadence like shorter ICD, while dual-wield Cadence prefers longer ICD. There is no perfect ICD for both sides. 0.43 to 0.46 seconds might be a better choice with the smallest losses for both sides.

0.43-0.46 seconds ICD for 1-hand and 2-hand Cadence
  • For 1-hand and 2-hangd Cadence, the interval of “Nec Edge (the latest hit frame) → Reaping / MA / most base attack (the earliest hit frame)” is about 0.47 seconds, longer than 0.46 seconds, so no failed charge problem. Only two rare exceptions (1-hand, “Nec Edge → 30% of base attack” or “Smite → 30% of base attack”) can cause the failed charge problem. If the total chance of WPS reaches 100%, there is no base attack so that no failed charge problem.
0.43-0.46 seconds ICD for dual-wield Cadence
  • For dual-wield Cadence, 0.43 seconds is long enough for most WPS to guarantee a following WPS-Cadence. MA, Execution (and other Attack Jump WPS) and Zolhan can guarantee their next Cadence to be WPS-Cadence. Only Reaping and AQC (and other Triple Hit WPS) are exceptions.

  • Furthermore, higher probability of WPS-Cadence also reduces the occurrence of such situations where double-swing base-Cadence is followed by an early-hit attack, causing failed charge problem, as base-Cadence itself is less common.

  • I admit that performances of Zolhan were not stable in my several tests of 0.46 seconds ICD,. When using keyboard, it seemed that Zolhan could ensure its following Cadence to be WPS-Cadence, but when using mouse, it was sometimes WPS-Cadence and sometimes base-Cadence. Sadly, I can’t find a better ICD, as 0.47 seconds or more may lead failed charges into 2-hand Cadence.

In theory, 0.43 to 0.46 should be similar. Take integer animation frame, random pause between attacks and mouse lag problems in consideration, I suggest to use the maximum value, 0.46 seconds.

Here is one of my test videos of 0.46 seconds ICD.

In this video, dual-wield Blade Master uses more WPS-Cadences, and Death Knight meets neither failed charges nor failed Cadences. It’s acceptable for both sides.

0.46 seconds ICD can be a more balanced value for all of 1-hand Cadence, 2-hand Cadence as well as dual-wield Cadence.

————————————

Summary:

  1. Longer ICD is good for dual-wield Cadence to trigger more WPS-Cadence and to generate more fast charges.
  2. Shorter ICD is preferred by 1-hand and 2-hand Cadence to aviod failed charge and failed Cadence problem.
  3. 0.46 seconds may be a better ICD with the smallest losses for both sides.
13 Likes

“it depends on your chosen wps”
when i tested it, using wps at 100% each, shears, smite, zolhan, whirling, and aqc? (i think) all benefitted
the problem starts to occur when mixing wps or more specifically mixing in specific wps; in particular Execution/Markovian, (iirc Zantai said it was because they have a delayed hit time on the other hand thus not always counting as a double hit)
i think i checked down to 0.3 ICD and on 100% wps it was still beneficial with some wps and the contrast of execution type grew (iirc) aqc benefitting even more as i recall

i think your DW findings are possibly too skewed/weighted based on your chosen wps and wps%, and the benefit of slower ICD wont translate as uniform into real world builds with different wps weights or different chosen wps combos
*i’m thinking that based on how each wps interacts with cadence if 100% at low ICD, and how it maintains that if certain wps gets simply left out in combos (particularly jump attack types) + the feedback during PTR after shorter icd got implemented

not saying it’s not worth a try, but with current ICD we already got reported positive feedback and dps gains over the 0.5, meaning 0.35 is already better in real play on DW

1 Like

Also tested a lot, on various weapon pools, but 0.35 also got so many problem, too.

I’m curious, why don’t we make cadence only use animation for cadence, instead of using other weapon animation, and slove this cadence problem Once for all. :laughing:

Cadence should be simple, not being a bad behavior in builds, a pain in player’s ass, right ?
:stuck_out_tongue_winking_eye:

2 Likes

wouldn’t cadence only animation for DW be a dps loss? :thinking:
*since you get less dual hand cadence hits, but also no bonus dmg scaling on dual hand cadence discharge “offhand”
**aside from no fast charging

We also did 100% single WPS tests. Different WPS has different animation data. I listed in the first sheet. WPS with late hit frame like Nec Edge can easily lead WPS-Cadence in both 0.35 and 0.5, as its interval (0.236) in even shorter. With 100% Nec Edge, loop of “Nec Edge → Nec-Cadence → Nec Edge → Nec-Cadence” can continue indefinitely.
But WPS with interval bewteen 0.35 and 0.5 is different. With 0.35 ICD, Execution, Markovian, Zolhan can hardly lead a WPS-Cadence, but only base-Cadence. Also tested with single 100%.

0.46 seconds is the longest time which allows as more types of WPS as possible for dual-wield to lead a following WPS-Cadence (like Execution and Markovian), meanwhile avoids any failed charges or failed Cadence in 1-hand and 2-hand.

We don’t think delay is the real reason.In some of my tests, I slowed video frame by frame, and found that Execution/Markovian sometimes failed to charge at first hit, but successed in second hit. In another part of tests, both hit failed. As my Part 4 and second sheet said, their falied charge is due to early hit. Even their 2nd hit are too early comparing with other WPS. Hit frame is Frame 8 and 1.15 animation speed, 0.23 seconds in real time.

Things for AQC is more complex. I deliberately did not elaborate on it. Sometimes, charge fails in first hit but successes in second hit, or even fails in second and success in third. The core mechanic is the same, checking the ICD at each hit. If ICD has been over at first hit, then charge or trigger Cadence at the first hit. Otherwise checking at second hit and third hit.

Yes. I just used a fair weight, 20% for each. And I deliberately didn’t bring Markovian, because it uses exactly the same animation with Execution. In my first sheet, the more WPS in yellow and green backcolor you bring, the more WPS-Cadence you hit.

No, I tested with two real DW BM builds, they become worse than before. But one real 2-hand DK build becomes better.

1 Like

i think it’s part of the reason if nothing else, because with execution at 100%; you still get delays
where as shears at 100% always hits/charges and always fast discharges
this then played out somewhat evenly among the wps types that had guaranteed instant dual hit, and the jump attack/ones with delayed hand strikes always fell off even on their own/without being mixed in
*and then messed up others if mixed in

it’s because each strike does count as potential charge, so it’s still reliant on ICD, that’s actually the reason why we don’t have super low ICD; AQC would insta charge+discharge cadence
*but overall on my 5-4-3 tests AQC still was less of an issue than Execution type
**when at 100% or not mixed in with execution types

During PTR, both banana and lee returned positive feedback after ICD change, i know Lee used Warborn DW Cadence DK, unsure which else was tested (if any)
so that’s/PTR-post after update is what i’m referring to

*just going to clarify i’m not saying we shouldn’t test this change/try 0.46,
just saying after 0.35 change there was positive DW feedback already, so there is a DW feedback clash there right now

edit, actually kinda funny, because looking at it, according to ICD/WPS sheet, the warborn dk is only using the “bad”/blue wps Death Knight, Level 100 (GD 1.2.1.0) - Grim Dawn Build Calculator :sweat_smile:

yes, and it is the same thing for jump attacks, a very shot time delay between their two hit. Shorter ICD truly rudeces its missing charge (as their both hits are too eatly), but also ruduce their following WPS-Cadence (as long interval to next hit).

Ah. But there are many changes for Cadence builds, higher Weapon Damage (500% →550%), buffs on Warborn Set, and perhaps lower monster’s Physical damage. Maybe all reasons together make DW Cadence be better.
But only for the aspect of more fast charges and more frequent Cadence, it becomes bad.

1 Like

that is very true, good point :+1:
kinda just makes it more interesting, if the buffs was so good it counteracted wps Cadence discharges getting collaterally nerfed :smile:

:smiley: Hah, of course it would be~

Then we can boost the skill’s damage, to make them behave as well as before. In the end, animation problem solved, build damage problem sloved……nobody worry about "The Mysterious Cadence” anymore. Everyone gets a happy ending. :slightly_smiling_face:

1 Like

I played a new test using this DW DK link you posted. Same 30% Reaping + 25% Triple + 23% Zolhan + 22% Jump. It can show how WPS-Cadence works to fasten charging.

In left part, ICD 0.35, no WPS-Cadence, only base-Cadence. “Bad” WPS doesn’t make WPS-Cadence.
In right part, ICD 0.46, Jump and Zolhan are allowed to make following WPS-Cadence, so about half Cadences are WPS-Cadence. In a total of 59 hits, 4 more Cadences were dealed (26/59 vs 22/59).

I also played on ICD 0.5. Though some failed charges for “double swing base-Cadence → Reaping / Markovian”, WPS-Cadences also helped to fasten charging. The two effects almost cancel each other out, 23 Cadences in 59 hits.
Or we can say, from 0.5 to 0.35, charging speed becomed neither faster nor slower for this DK.

1 Like

that might explain it then, there was no actual net gain on the cadence hit/fast charging side, but the warborn and cadence buffs then carried it :thinking:

Tested and validated the conclusion of MI023.

When ICD=0.35, there are only 31 total cadence in 80 total attack and only two WPS-Cadence.

When ICD=0.46, there are 38 total cadence in 80 total attack and almost all Cadence are WPS-Cadence.


(ICD=0.35)


(ICD=0.46)

If the original intention of reducing ICD in 1.2.1 was to enhance Cadence, then “reducing ICD from 0.5 to 0.35” seemed backfired, resulting in a significantly lower proportion of Cadence in continuous attacks.

If want to enhance Cadence, then 0.46 seconds might be a best ICD, while longer ICD may lead to failed charge and failed Cadence.

1 Like

any of you observe this btw ?

i remembered noticing that being a thing back when i messed around with all this

Strange. I know Shears is not stable, as its theoretical interval is close to 0.35, so I put it in green backcolor.Based on my testing experience, keyboard should be better than LMB and RMB.

I tested it just now.
At 200% attack speed, both keyboard and LMB could keep loop.
At 182% AS, LMB could keep loop, but keyboard not (broken by base-Cadence).
At 177% AS, keyboard could, LMB not.
At 165% AS, both could again.

I think that is what I called random pause. I don’t fully understand it, but pauses between animations are not fixed, sometimes longer, sometimes shorter, various with AS.

Yea, it is highly likely to be this reason. I tested the same thing before, but didn’t draw a good conclusion.

1 Like

i think this was a 0.4icd mod if i remember correct
it’s more that cadence was so noticeable slower and also inconsistent that it’s even audible
made me think of that old speed scaling bug that’s been reported, where some skills were slower/faster depending on using mouse or keyboard slots :sweat_smile:

Here is another test I did in v1.2.1, to see actual speed of skills. The horizontal axis is Cast Speed in Character Sheet, the vertical axis is the actual speed calculated from the total time of 60 attacks.
DEE, Doombolt and Phantasmal Blades have the same base animation length, 19 frames, and the same animation speed, 1,0.
Doombolt and Blades drew very similar curves, but DEE’s was completely different. At first, I thought it was a difference in the skills themselves. But later on, I realized that the real reason was my habit of putting DEE on LMB but Doombolt and Blades on keyboard :sweat_smile:. When I turned DEE on keyboard and tested again, it could reach similar high value of 3.7-3.8 at 200% CS.

2024-07-28_033247

2024-07-28_033301

2024-07-28_033313

inb4 speed scaling is also affected by running the game in x64 vs x86 aside from having skill on mouse or keyboard :sweat_smile:
:grin:

Just curious.
Did any of you test 2H Cadence on a build that hits 200%atk in a consistent way where it also only has Soldier WPS Markovian and Zolhan?
In this case its a WB using Occultist for the atk speed, phys res, flat vit dmg converted,etc.
I am noticing weird cadence trigger behaviour (seems missing) whenever its base 198% is then taken to 200% either by Rylok Crest medal, Chausses of Barbaros.
Animation increases but consistent Cadence trigger seems to go awol.
Its noticable enough that I now spec the char to not hit 200% atk speed.
Cheers

Sounds strange. With only Soldier WPS, all WPS and base attacks have early hit frames, so the interval between attacks must be longer than ICD and no reason for missing.
I test at fixed 200% speed, and found no missing. It works fine and Cadence every 3 hits.
I’m not sure if attack speed is changing, like you said, from 198% to 200%. I do all tests at fixed speed. Maybe things go wrong at changing speed.

2 Likes

Thanks mi023,
it could be a factor of things as well, possibly all interracting in my case - this is with the 2H Voldrak weapon.
When it happens lifesteal tanks and even the low phys res Nemesis in SR80+ take a chunk longer time to kill lol.