Combat Mechanics? Attack Speed, Recovery

[tonberrian]

I've been lurking in this thread, and I thought that the current sheet wasn't doing enough to make comparisons between equipment sets easy. So I'm working on a different one. It's not finished yet, which is why it's not open to modify. Any thoughts on format? Other variables I should add in?

[Daemonjax]

I've been lurking in this thread, and I thought that the current sheet wasn't doing enough to make comparisons between equipment sets easy. So I'm working on a different one. It's not finished yet, which is why it's not open to modify. Any thoughts on format? Other variables I should add in?

 

A full blown program written in java or whatever with a gui would be best.

Edited April 5, 2015 by Daemonjax

[tonberrian]

 

I've been lurking in this thread, and I thought that the current sheet wasn't doing enough to make comparisons between equipment sets easy. So I'm working on a different one. It's not finished yet, which is why it's not open to modify. Any thoughts on format? Other variables I should add in?

 

A full blown program written in java or whatever with a gui would be best.

 

Probably, but I've never been good at GUIs myself. I can make a very pretty spreadsheet, though.

[Daemonjax]

 

 

I've been lurking in this thread, and I thought that the current sheet wasn't doing enough to make comparisons between equipment sets easy. So I'm working on a different one. It's not finished yet, which is why it's not open to modify. Any thoughts on format? Other variables I should add in?

 

A full blown program written in java or whatever with a gui would be best.

 

Probably, but I've never been good at GUIs myself. I can make a very pretty spreadsheet, though.

 

 

Yeah, it takes me longer to make a GUI for a program than the program itself.

[Myrten]

I'll make some changes to the spreadsheet - goal is to have melee weapons with different bonuses separate as they have different DPS. I'll also try to accommodate for the results made by Daemonjax.

 

One question though - do all weapons with the same declared speed in the description(like slow\fast etc...) actually have the same speed?

Edited April 6, 2015 by Myrten

[Daemonjax]

I'll make some changes to the spreadsheet - goal is to have melee weapons with different bonuses separate as they have different DPS. I'll also try to accommodate for the results made by Daemonjax.

 

One question though - do all weapons with the same declared speed in the description(like slow\fast etc...) actually have the same speed?

 

No way to know for sure until tested, unfortunately.

 

Btw, verified there was zero attack speed difference between:

 

a) daggers (dual-wielding), 10 dex, naked

b) daggers (dual-wielding), 10 dex, padded armor (-20%), talent: two weapon style

Edited April 6, 2015 by Daemonjax

[ProjectBG2Respawn]

Does spell combo exist?

For example, if you use spell combinations, damage is increased.

 

If you cast a spell that reduces opponent's fire resistance and then you cast a firestorm i think that's a good combo.

[Myrten]

 

I'll make some changes to the spreadsheet - goal is to have melee weapons with different bonuses separate as they have different DPS. I'll also try to accommodate for the results made by Daemonjax.

 

One question though - do all weapons with the same declared speed in the description(like slow\fast etc...) actually have the same speed?

 

No way to know for sure until tested, unfortunately.

 

Can you test dagger vs stiletto vs club at 10 dex naked? This should give us an answer

 

 

Btw, verified there was zero attack speed difference between:

 

a) daggers (dual-wielding), 10 dex, naked

b) daggers (dual-wielding), 10 dex, padded armor (-20%), talent: two weapon style

This is a bit weird.

 

We got 4 parts of an attack:

 

1)frame attack delay

2)attack animation length

3)frame recovery delay

4)frame recovery length

 

We need to figure out how different modifiers like dex, talents, armors etc.. affect them.

 

Taking a look at your data I got a feelingthat 2)4) are working better than dex would suggest while 1) and 3) are barelly affected

 

I'm also wandering what's the cause of all these variations, did you lock the frame rate like suggested here?

http://forums.obsidian.net/topic/72272-combat-mechanics-attack-speed-recovery/page-6?do=findComment&comment=1628773

Edited April 6, 2015 by Myrten

[Myrten]

I think I've found why we have such problems with DEX when looking at Daemonjax data - DEX effect is not linear!

 

Take a look at these two charts:

 

First one shows dex vs frames, second shows dex vs relative frames (10 dex = 1).

Attack Animation and Recovery Length follow the same pattern while delays kinda stay the same or drop by a bit.

Look at frame decrease between 10-14 and 14-18 - both are 4 dex different but difference in frames is -41% vs -5%

 

This means that difference between DEX 10 and DEX 11 is much bigger then difference between DEX 19 and DEX 20

 

I think we can figure out this formula but we need results for:

 

Naked dagger with following dex values:

11

12

13

15

16

17

19

20

 

11,12,13 are most important

Edited April 6, 2015 by Myrten

[kmbogd]

I stated this several times, but I think you overlooked it: dual wielding does not use 20 24 20 24 or 30 36 30 36 at 10 Dex with no armor. It's more like 20 20 20 20 or 30 30 30 30 according to my tests.

[Omnicron]

 

 

Hey Daemonjax,

 

Noticed you mentioned the disparity between dw fast 1h weapons and just using 2h weapons. Were my rough calculations in that other thread accurate?

 

To summarize:

 

1) Attack frequency 1/44 frames for dw and 1/84 frames for 2h --> dw attacks 2x as often as 2h

2) Damage 8-12 dw (avg 10) and 20-29 2h (avg 24.5)

 

Therefore 2h does more DPS since it does 2.45x the damage and attacks half as often, yielding (2.45/2) = 1.225x the damage of dw.

I don't have enough free time to play so my game progress is limited so I don't know what effects some of the weapons   have but if you add added stats and effects 2 weapons will catch up.  Also the elemental damage and such helps with the dr   

 

 

We'll need a full-blown dps calculator with a character builder (including enchants and gear) that could run simulations within user-defined scenarios (starting with single-target would be easiest) to know for sure, but off the cuff I'd say dual-wielding fast weapons makes the most sense for when you want to maximize effects that proc on hit (like the cipher's draining whip) that aren't percentage-based (all of the rogue abilities and features are percentage-based).  

 

If you'd want to maximize interrupt chance, then you'd want to dual-wield fast weapons (not saying that focusing on making an interrupt character is optimal).  There's also items in the game with unique enchants that may make you consider dual-wielding.  

 

It's different from D&D because there's basically no flat modifiers to damage (Ranger's Marked Prey is not a flat damage increase, btw) and the damage reduction system favors heavy hits (because that IS a flat modifier to damage).  So classic archetypes (like a high dps dual-wielding stealthy rogue) are not favored by the game's mechanics.

 

 I am trying to decide my next talent for my Rogue since I am not getting nay class talents. I have been using the Estoc up to now because the percentage multiplier means greater sneak attack damage on my 1st hit, even if it is slower for subsequent sneak attacks. Now however I want to choose an offensive talent and I am unsure what is best for damage. I did tests with dual Stilletos and I got the exact same sneak attack damage output per round (36 vs 18x2) and since I have high perception I get more interrupts I guess. So I am torn between taking a Two Handed Style for harder hits or taking two weapon style for better attack speed and committing to that style when I am unsure which is better in the long run. This thread has suggested that rogues actually do better with their sneak attack with 2h weapons when encountering higher DR enemies? 

 

PS. I gave Scion of the Flame to my Priest and it did not seem to affect his Radiance or fire based spells

[Daemonjax]

This means that difference between DEX 10 and DEX 11 is much bigger then difference between DEX 19 and DEX 20

 

 

Yeah, it's not going to be perfectly linear because the two delays aren't affected (at least as much) as the animation/recovery frames by dex, but it may not really matter since using 2.25 per dex point is close enough -- and everything else only affects the recovery time.

[Daemonjax]

I'm also wandering what's the cause of all these variations, did you lock the frame rate like suggested here?

 

Yeah, of course it's locked to 30 fps during recording.  Otherwise, I'd be an idiot.

 

The variations are recorded because that's really what I observe, and it's there for anyone to interpret any way they want, but I think what's most important is nailing the time between attacks, and recovery time.

 

I'll do some armor tests today.

[Myrten]

 

This means that difference between DEX 10 and DEX 11 is much bigger then difference between DEX 19 and DEX 20

 

 

Yeah, it's not going to be perfectly linear because the two delays aren't affected (at least as much) as the animation/recovery frames by dex, but it may not really matter since using 2.25 per dex point is close enough -- and everything else only affects the recovery time.

 

It's not linear at all and this difference is huge, no matter what number you assume it will be wrong. With 2.25% at 14 dex it should be 9% faster than dex 10 so it should go down from 84 frames to 77 frames, but instead it's around 57 frames.

 

I can figure out this formula but I need more data.

Edited April 6, 2015 by Myrten

[Daemonjax]

 

 

This means that difference between DEX 10 and DEX 11 is much bigger then difference between DEX 19 and DEX 20

 

 

Yeah, it's not going to be perfectly linear because the two delays aren't affected (at least as much) as the animation/recovery frames by dex, but it may not really matter since using 2.25 per dex point is close enough -- and everything else only affects the recovery time.

 

It's not linear at all, no matter what number you assume it will be wrong. With 2.25% at 14 dex it should be 9% faster so it should go down from 84 frames to 77 frames, but instead it's around 57 frames. I can figure out this formula but I need more data.

 

 

1h dagger (not dual-wielding, naked):

 

50 dex: 50 - 10 = 40 ==> 40 * 2.25 = 90 ==> 1 + 90 / 100 = 1.90 ==> 58.5 / 1.90 = 30.79 (29.8 observed)

18 dex: 18 - 10 =  8 ==>    8 * 2.25 = 18 ==> 1 + 18 / 100 = 1.18 ==> 58.5 / 1.18 = 49.58 (49.71 observed)

14 dex: 14 - 10 =  4 ==>    4 * 2.25 =   9 ==> 1 +   9 / 100 = 1.09 ==> 58.5 / 1.09 = 53.67 (53 observed)

 

 

To get more accurate frame times for the four parts of the attack cycle, we could probably record at 60 fps and set combat to slow motion and use math (if necessary) to normalize.  I don't think we really need to go that far, though.

Edited April 6, 2015 by Daemonjax

[Myrten]

 

 

 

This means that difference between DEX 10 and DEX 11 is much bigger then difference between DEX 19 and DEX 20

 

 

Yeah, it's not going to be perfectly linear because the two delays aren't affected (at least as much) as the animation/recovery frames by dex, but it may not really matter since using 2.25 per dex point is close enough -- and everything else only affects the recovery time.

 

It's not linear at all, no matter what number you assume it will be wrong. With 2.25% at 14 dex it should be 9% faster so it should go down from 84 frames to 77 frames, but instead it's around 57 frames. I can figure out this formula but I need more data.

 

 

1h dagger (not dual-wielding, naked):

 

50 dex: 50 - 10 = 40 ==> 40 * 2.25 = 90 ==> 1 + 90 / 100 = 1.90 ==> 58.5 / 1.90 = 30.79 (29.8 observed)

18 dex: 18 - 10 =  8 ==>    8 * 2.25 = 18 ==> 1 + 18 / 100 = 1.18 ==> 58.5 / 1.18 = 49.58 (49.71 observed)

14 dex: 14 - 10 =  4 ==>    4 * 2.25 =   9 ==> 1 +   9 / 100 = 1.09 ==> 58.5 / 1.09 = 53.67 (53 observed)

 

 

To get more accurate frame times for the four parts of the attack cycle, we could probably record at 60 fps and set combat to slow motion and use math (if necessary) to normalize.

 

My bad - I looked at 1h sword at 10 dex not 1h dagger at 10 dex. Sorry. Give me some time to re-analyze this

[Myrten]

OK - it's linear, if you ignore delays the actual % value per dex is around 3%

[Daemonjax]

OK - it's linear, if you ignore delays the actual % value per dex is around 3%

 

I figured as much, but it's hard to determine a fixed number to use for both delays.  

 

I started solving the simultaneous equations and just stopped when I came up with the 2.25 and saw that it worked so well.

Edited April 7, 2015 by Daemonjax

[Myrten]

 

OK - it's linear, if you ignore delays the actual % value per dex is around 3%

 

I figured as much, but it's hard to determine a fixed number to use for both delays.  

 

I started solving the simultaneous equations and just stopped when I came up with the 2.25 and saw that it worked so well.

 

Since we have no idea if delays are affected by anything and it seems they aren't I'd simply use their average or median. For the other two we can use 3% and it should give quite good results altogether.

[Seerdecker]

Hi guys!

Thanks so much for your collaborative spreadsheet! It was very useful to see how the different parameters interact.

I was interested in finding out whether it is true that a 2H weapon such as the estoc is optimal for a rogue. In fact, my analysis suggests that a rogue dual-wielding sabres is actually better by a significant margin. Here's my rationale. I'd be happy to be corrected if I'm wrong.

In general, DPS = (BaseWeaponDamage * DamageMod - DamageResistance) * AttackSpeed.

The base weapon damage is the number listed by the game as (base X-Y).
A sabre got base damage 13-19 (average 16).
An estoc got base damage 14-20 (average 17).
*Nothing* affects the weapon base damage (not might, not enchantments, etc). It's a well named concept.

The damage modifier (DamageMod) is a pure additive multiplier of the form (1.0 + bonus1 + bonus2 - penalty1 ...).
The base modifier is 1.0.
A graze adds -0.5 to the modifier.
A critical hit adds 0.5 to the modifier by default (some effects change this value).
Might at 18 adds 0.03*8=0.24 to the modifier.
Superb weapon quality adds 0.45 to the modifier.
Savage Attack adds 0.2 to the modifier.
Reckless Assault adds 0.2 to the modifier.
Sneak Attack adds 0.5 to the modifier.
Death Blows adds 0.5 to the modifier (*warning* that's from what I've read).
Crippling Strike adds 0.25 to the modifier.
etc.

Let's consider a normal attack under some scenarios.

Best case scenario: crit with Death Blows.
DamageMod = 1.0 + 0.5 + 0.24 + 0.45 + 0.2 + 0.2 + 0.5 + 0.5 = 3.59.

Worse case scenario (excluding a miss): graze without Sneak Attack.
DamageMod = 1.0 - 0.5 + 0.24 + 0.45 + 0.2 + 0.2 = 1.59.

Average scenario: hit with Sneak Attack.
DamageMod = 1.0 + 0.24 + 0.45 + 0.2 + 0.2 + 0.5 = 2.59.

I like to pair my rogue with a cipher to attack paralyzed/hobbled/blinded/weakened enemies, so I expect to get an average damage mod >= 3.0 for a sabre by the end of the game, accounting for special attacks like Crippling Strike, might bonuses, etc.
An estoc would get an additional bonus of 0.15 from 2H style, making the damage modifier 3.15, and a DR reduction bonus of 5.
From the spreadsheet, assuming an attack/recovery bonus of 0.2 from 2W style, I get a speed ratio of 2.14/1.35 = 1.58 for sabre vs estoc (for 1 full attack).

Assuming DR 15, for one weighted full attack, my expected damage is (16.0 * 3.0 - 15) * 1.58 = 52 for the DW sabres and 17 * 3.15 - 15 + 5 = 44 for an estoc. Lower DR values would favor the sabre more. While an estoc got a slightly higher base damage, damage modifier, and DR reduction bonus, it's not enough to offset the speed difference.

Incidentally, accuracy seems to be less important for a rogue than say a caster, since the damage modifier is fairly high in any case. A graze/crit doesn't modify the damage that much.

Summary: in theory, beware rogues wielding two sabres running around naked.












 

[Matt516]

The damage modifier (DamageMod) is a pure additive multiplier of the form (1.0 + bonus1 + bonus2 - penalty1 ...).

The base modifier is 1.0.

A graze adds -0.5 to the modifier.

A critical hit adds 0.5 to the modifier by default (some effects change this value).

 

Are you absolutely sure about this? I knew most modifiers were additive, but I've never seen anything to suggest that grazes and crits were treated this way. Are you 100% sure that this is how it works? I thought you had your DamageMod (with all the additive boni), and that was then multiplied by 1.5 or 0.5 if you grazed or critted. Then again, they changed the damage calculation completely a few times in Beta, so my info may be obsolete. Where did you get your info from?

 

If that's really how it works, my old spreadsheet for calculating effective damage multiplier from accuracy is completely worthless now.

[Seerdecker]

 

Are you absolutely sure about this? I knew most modifiers were additive, but I've never seen anything to suggest that grazes and crits were treated this way. Are you 100% sure that this is how it works? I thought you had your DamageMod (with all the additive boni), and that was then multiplied by 1.5 or 0.5 if you grazed or critted. Then again, they changed the damage calculation completely a few times in Beta, so my info may be obsolete. Where did you get your info from?

 

 

My info comes from this post in this thread (I've seen similar information elsewhere).

https://forums.obsidian.net/topic/72272-combat-mechanics-attack-speed-recovery/?p=1627132

 

2-mins testing suggests that it does behave that way. My rogue does not have its damage multiplied by 0.5/2.0 when it grazes/crits. That said, no, I'm not absolutely certain.

[tonberrian]

Can I get a summary of what dexterity effects (and how much), what delay frames are where and how many, and frame lengths of attack animation, recovery, and reload rates as you currently understand them? It's all over the thread, and I'd like it to be in one place.

 

My spreadsheet should be finished sometime in the next 24 hours, once I get all the attack speed stuff sorted out and get all my weapon set comparisons finished.

[Matt516]

 

 

Are you absolutely sure about this? I knew most modifiers were additive, but I've never seen anything to suggest that grazes and crits were treated this way. Are you 100% sure that this is how it works? I thought you had your DamageMod (with all the additive boni), and that was then multiplied by 1.5 or 0.5 if you grazed or critted. Then again, they changed the damage calculation completely a few times in Beta, so my info may be obsolete. Where did you get your info from?

 

 

My info comes from this post in this thread (I've seen similar information elsewhere).

https://forums.obsidian.net/topic/72272-combat-mechanics-attack-speed-recovery/?p=1627132

 

2-mins testing suggests that it does behave that way. My rogue does not have its damage multiplied by 0.5/2.0 when it grazes/crits. That said, no, I'm not absolutely certain.

 

 

It would be multiplied by 0.5/1.5, not 0.5/2.0. But anyway...

 

That still seems awfully odd to me. Given that durations are multiplied by 0.5/1.5 (which means the total damage of any DoTs is multiplied by 0.5/1.5), I find it hard to believe that they'd treat normal attacks differently. If this is true, it greatly reduces the importance of accuracy. But it does look like there's been a lot of testing done. Looked at the spreadsheet and it's... interesting. Don't think they're doing accuracy modifier correctly - I can help with that over the weekend if it's still not sorted out.

 

Would be great to have dev word on how attack resolution affects damage though. All information I've seen literally everywhere except this thread suggests that it works multiplicatively with the total damage multiplier, not additively.

[peddroelm]

We need a single thing (item / ability) that allows to get a test weapon with CONSTANT damage (weapon min = weapon max).. Eliminating damage RNG roll would then make it trivial to reverse engineer the damage formula and the exact bonuses and places where each ability/talent/bonus/effect/etc.. fits in in said formula..

 

Some games don't have RNG damage roll to begin with ( Mass Effect 2/3 - they lacked the damage log  ), some provide the items required to remove it (Diablo 2/3 have ring/amulet mods that increased weapon min damage ) .

 

Only thing that would qualify that I'm aware of in this game - is a fighter talent that ups weapon min damage by 20%.. Sadly that is not enough to completly shut the RNG gap for any weapon .. ( you can get pretty narrow damage ranges on some weapons but not constant damage)

 

For accurate tesing we will need a modder support - either mod in  a weapon with min damage = max damage.. Better yet a ring with +6 to minimum damage (could test most weapon types with 2 of those rings) .. Or make that fighter talent add +100% to weapon minimum damage ( will only be able to test fighter abilities in that case) ...

 

No-one familiar enough with the unity engine capable of such simple modification ?

Edited April 8, 2015 by peddroelm