Combat Mechanics? Attack Speed, Recovery

[Daemonjax]

 

 

 

My hunch is that Dex is actually going to be superior most of the time, since (again), the most Might ever provides you is 3% of your base weapon damage per point, while Dex provides you 3% more speed, which translates to all your attack multipliers (Might included) being multiplied by 1.03. Probably. The DR DPS decrease increase is still a problem for Dex though, which is why I honestly don't know..

 

 

True, except it's closer to 2.25% in practice due to the animation transition delays.

 

 

Really? That would at least bring it a bit back in line with Might... though probably unintentionally. I hope Obsidian does a once-over on the attribute system before PoE 2. Design decisions are one thing (stop complaining about Might being both physical and magical, FFS), but when one attribute is almost strictly better than another in every way, that's just a broken system that needs to be mended.

 

 

Dexterity is strictly better than might (due to it being a reduction instead of an increase), except for spellcasters that like to use damage/healing spells since there's no way to increase spell damage like you can for weapons.  

 

It also really shines for classes that have inherent damage bonuses, like rogue and cipher (to a lesser extent).

 

Might is definitely more of a (damage) caster attribute.  A str 10 rogue does nearly as much damage as a str 20 rogue, and that becomes more true at higher character and equipment levels -- e.g. a 3.0 damage multiplier (at 20 strength) is only 11% more damage than a 2.7 damage modifier (at 10 strength).  The difference is more substantial at level 1, though, and there is the fortitude save bonus to consider.

 

There's exceptions -- like with moon godlike might starts looking better if you want to maximize heal amount.

Edited April 13, 2015 by Daemonjax

[Mysh]

I have written a MATLAB code, which does auto-comparisson of many things and tries to find optimum, weapon, talents, attribute distribution, weapon mods etc.

 

Before communicating my results, could you please confirm the following formula?

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

Edited April 13, 2015 by Mysh

[blinkicide]

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

1. It's my understanding, based on other's tests, that nothing effects the attack/reload delay.  Hence the + 10 should be by itself.

 

I'd also just point out some assumptions in your equation that I am doubtful of, but really don't know myself:

 

2.  That attack speed and dexterity are multiplicative instead of additive.  Seen some people claiming attack speed bonuses are additive with each other.  Whether dexterity is considered just another attack speed bonus in this context, or it's own thing, I'm not certain.

 

3.  Same with reload speed. 

Edited April 13, 2015 by blinkicide

[iszathi]

 

 

 

 

My hunch is that Dex is actually going to be superior most of the time, since (again), the most Might ever provides you is 3% of your base weapon damage per point, while Dex provides you 3% more speed, which translates to all your attack multipliers (Might included) being multiplied by 1.03. Probably. The DR DPS decrease increase is still a problem for Dex though, which is why I honestly don't know..

 

 

True, except it's closer to 2.25% in practice due to the animation transition delays.

 

 

Really? That would at least bring it a bit back in line with Might... though probably unintentionally. I hope Obsidian does a once-over on the attribute system before PoE 2. Design decisions are one thing (stop complaining about Might being both physical and magical, FFS), but when one attribute is almost strictly better than another in every way, that's just a broken system that needs to be mended.

 

 

Dexterity is strictly better than might (due to it being a reduction instead of an increase), except for spellcasters that like to use damage/healing spells since there's no way to increase spell damage like you can for weapons.  

 

It also really shines for classes that have inherent damage bonuses, like rogue and cipher (to a lesser extent).

 

Might is definitely more of a (damage) caster attribute.  A str 10 rogue does nearly as much damage as a str 20 rogue, and that becomes more true at higher character and equipment levels -- e.g. a 3.0 damage multiplier (at 20 strength) is only 11% more damage than a 2.7 damage modifier (at 10 strength).  The difference is more substantial at level 1, though, and there is the fortitude save bonus to consider.

 

There's exceptions -- like with moon godlike might starts looking better if you want to maximize heal amount.

 

 

actually, due to the way DR works having might proabably does more damange in some cases. 

 

[Daemonjax]

It's not good enough to compare dexterity vs might without taking into account weapon enchantments, talents, and abilities which increase that character's bonus damage.  

[iszathi]

In that test lenas is being used by a lvl 12 talented for ranged character, same with the arquebus, the hunting bow is without any talents. They behave in similar ways. 

[Matt516]

It's not good enough to compare dexterity vs might without taking into account weapon enchantments, talents, and abilities which increase that character's bonus damage.  

 

Yeah, that's the whole point. Obviously when taken completely alone, Might is better than Dex for DPS (even if Dex works best-case i.e. on all parts of the attack/recovery) because they both increase DPS by the same raw amount, but Might does it in a way that helps punch through DR while Dex does it in a way that causes more DPS loss from DR. That's not in dispute.

 

But the moment you start adding on any other damage modifiers, Dex starts to look better and better - because for a (for example) 20 base damage weapon, 20 Might will only ever add 6 damage to that weapon, no matter what enchantments you've got on it or what quality it is or what modals/damage boni you're running - Might is strictly additive with any other damage modifiers. Whereas Dex increases attack speed, which acts exactly like a multiplicative bonus to all other damage modifiers. And the point I've tried to make is that while this probably isn't a big deal in PoE 1, it'll be a huge deal in PoE 2. When everyone's running around with Superb++ weapons (+75% damage if we follow the current trend) and 20 levels' worth of abilities and talents that grant bonus damage - 15% here, 20% there, 50% somewhere else, etc... that static +30% you get from maxing out Might will become less and less important while the attack speed increase from Dex will get more and more important. That's the biggest worry at the moment. In PoE 2, Might will become piddly and nearly worthless, while Dex will continue to scale with levels. That just ain't right.

 

Same thing happens with accuracy, btw - since graze/crit is just a static +50% or -50% of base weapon damage (not a multiplier), accuracy becomes much less important when you can just stack a crapton of damage modifiers on.

 

I guess what I'm saying is - iszathi, you're right as far as it goes, but I think you kind of missed the main point, which is that although Might is better than Dex for DPS when you look at the very base case, in any real case Dex will probably be better even despite the DR problem because it acts as a multiplicative modifier to all the other attack modifiers, whereas Might is and will always be additive.

 

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

 

If you ask me? Might should be multiplicative with all damage, as should grazes and crits. And weapon enchantment damage boni - especially those since you're changing the "base" damage on the weapon info sheet anyway and it's extremely unintuitive that my (for example) +15% damage from the 2H weapon talent doesn't actually give any more damage as I upgrade the weapon - you're showing one base damage on the sheet (base X quality) and calculating from another (base), which is very confusing. 

 

Leave all the other modifiers additive, of course. The reason they became additive in the first place was to mitigate issues with characters like rogues stacking up ludicrous damage from just a few multiplicative modifiers. Nerfing that was a good change. Nerfing Might to the point of irrelevance was not - keep it multiplicative. As for grazes and crits and weapon enchantments - the current implementation isn't inherently broken necessarily, but it is highly unintuitive.

Edited April 14, 2015 by Matt516

[Daemonjax]

If you ask me? Might should be multiplicative with all damage

 

 

Yeah, that would bring might in line with dex if it was 2.25% per point of might instead of 3%.

 

You make some really good points how the existing combat system just breaks down at higher levels (where everyone has at least 100 accuracy).  They'll need to redesign a lot of stuff to make it work well.

 

At the very least there needs to be a distinction between the keywords for increased damage (additive) and more damage (multiplicative) -- like in POE (Path of Exile).

 

D&D 5th edition has a much more compressed progression (level 20 characters get a +6 to hit), which works really well in practice.

 

off-topic: I'd really prefer that just had unique attributes (some classes would have a strength stat, but others wouldn't, etc.) for each class in pillars 2.  I know, I know... it sounds crazy.

Edited April 14, 2015 by Daemonjax

[Mysh]

I have written a MATLAB code, which does auto-comparisson of many things and tries to find optimum, weapon, talents, attribute distribution, weapon mods etc.

 

Before communicating my results, could you please confirm the following formula?

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

Anyone else can make some input?

[Mysh]

 

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

1. It's my understanding, based on other's tests, that nothing effects the attack/reload delay.  Hence the + 10 should be by itself.

 

I'd also just point out some assumptions in your equation that I am doubtful of, but really don't know myself:

 

2.  That attack speed and dexterity are multiplicative instead of additive.  Seen some people claiming attack speed bonuses are additive with each other.  Whether dexterity is considered just another attack speed bonus in this context, or it's own thing, I'm not certain.

 

3.  Same with reload speed. 

 

 

1. Would you suggest?

 

Attack_Time = ( 10+ ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod)  ) / (1 + Dexterity*0.03 ) )/30;

 

2. I have assumed that bonuses that affect attack speed are additive.

 

Thanks for spending time to help me with this. I appreciate your help.

[Daemonjax]

 

 

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

Anyone else can make some input?

 

 

It's hard to nail down the numbers for attack_frames and recovery_frames due to the animation transition delays.  At least some different attack animations (many weapons have two, one of which is randomly chosen each swing) also have different timings.

 

I would use total attack time instead (attack_frame + recovery_frames + attack_delay + recovery_delay).  If you do that, then:

 

1 + ((Dexterity - 10) * 0.0225) would be closer to what we observe in the game.

 

 

Not sure if the attack speed mod should just be added to the dexterity mod or not, but I'd be a little surprised if it didn't seem to work this way in practice:

 

1 + ((Dexterity - 10) * 0.0225) + (0.75 * attackSpeedIncrease)

 

 

I don't know if there's a reload_delay.

 

Note: 2.25 / 3 = .75

 

I've grown bored with trying to nail down the dps formula exactly.  The mechanics are too opaque and my free time is too valuable to me.

Edited April 14, 2015 by Daemonjax

[iszathi]

I remember reading one dev posting about attack speed making recovery 0 and then making the animation faster (if i remember right he also said it was not posible with the items and talents ingame). 

 

So im still going back and forth with attack speed afecting the animation or not. Other than that, it looks The delay being affected by dex is another issue. 

[Mysh]

 

 

 

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

Anyone else can make some input?

 

 

It's hard to nail down the numbers for attack_frames and recovery_frames due to the animation transition delays.  At least some different attack animations (many weapons have two, one of which is randomly chosen each swing) also have different timings.

 

I would use total attack time instead (attack_frame + recovery_frames + attack_delay + recovery_delay).  If you do that, then:

 

1 + ((Dexterity - 10) * 0.0225) would be closer to what we observe in the game.

 

 

Not sure if the attack speed mod should just be added to the dexterity mod or not, but I'd be a little surprised if it didn't seem to work this way in practice:

 

1 + ((Dexterity - 10) * 0.0225) + (0.75 * attackSpeedIncrease)

 

 

I don't know if there's a reload_delay.

 

Note: 2.25 / 3 = .75

 

I've grown bored with trying to nail down the dps formula exactly.  The mechanics are too opaque and my free time is too valuable to me.

 

 

 

- Don't worry about (1 + Dexterity*0.03). In my code 0 dexterity =10. So, I got it covered. Sorry didn't explain this.

 

- Do I get right that you suggest: (attack_frame + recovery_frames + attack_delay + recovery_delay) / (1 + ((Dexterity - 10) * 0.0225))

 

Where is 2.25/3 coming from? Isn't Dexterity 3% increase per attribute?

 

By the way can you please how many frames per second, I should assume? Would it be 30?

Edited April 14, 2015 by Mysh

[Mysh]

By the way, I have done the analysis for an one-handed Paladin build with +12 accuracy, +0.45dmg and flame weapon yesterday night.

 

1. It appears talent priority for DPS are:

Vulnerable attack, savage attack, weapon focus. There is no need for one-handed style skill. Better grab envenomed strike.

 

2. The best weapon you should go for is the sabre, followed by Stiletto.

 

BTW are these stats OK for Sabre?

 

Min Damage 13

Max Damage 19

Attack Frames 18

Recovery Frames 30

 

This weapon is way... above others.

 

3. You should prioritize to stack rending ability (+3DR), then x0.2 speed, followed closely by +0.5 crit mod. I do not think there is a sabre with rending ability for DR, so better go for weapon speed.

 

4. The optimum stat distribution for weighted DPS (see below) (assuming you can have 10 might and 10 dexterity and you don't know how to distribute points between them) is:

- 7 might and 13 dexterity

- 4 might and 16 dexterity with sworn enemy on

 

5. If you use a Foe database and you plot DR vs Deflection, you see some randomness but there is something like a trend. If you now plot DR*Deflection, it's magic all enemies distribute perfectly exponentially. This means Obsidian used the size DR x Deflection to weight enemies.

 

Therefore, I used the product of DR x Delection to weight DPS, by Foe size. Minions count less, Adra dragon more in order to optimize my DPS.

 

I have tons of graphs and charts, but now I am at the airport. Does anyone know any easy website to upload images fast? I used to use photobucket when I was younger... I guess things have changed now.

 

Don't know, when  will come back, but when I do, I will strike hard. )

Edited April 14, 2015 by Mysh

[illathid]

Imgur is good photo hosting site.

[Enexemander]

Use Imgur for fast image uploads.

[Mysh]

OK let me try this. I don't know if you can see these addresses.

 

http://imgur.com/zYzuFxm

http://imgur.com/2mwZ6f8

http://imgur.com/g3bqEjs

http://imgur.com/LS3n5p7

http://imgur.com/Lm5kGNv

http://imgur.com/HoUexDH

 

Formulas need verification. If you have MATLAB I can provide scripts. It's a very fast code.

 

DPS weighted was done as:

Weighted_DPS(Weapon_Loop,Attribute_loop,Foe_Loop)=Foe_Size(Foe_Loop)*Total_Damage/Attack_Time/Total_Foe_Size;

Foe_Size=Foe_Base_DR.*Foe_Deflection;

Edited April 14, 2015 by Mysh

[iszathi]

Great Job
 

[Mysh]

Thanks looking forward to your constructive feedback. To start with I need someone who wants to work together with me and look on the scripts.

[Mysh]

I forgot to mention that all this is based on carrying a bucket shield on the off-hand.

 

Dexterity increases when you add a quality mod. If you don't have a damage boosting quality mod, then having more might is more useful. At about 0.2 quality mod balanced might-dexterity yields the best results. Dexterity raises up when the weapon damage multiplier is 0.45.

 

Weapon multiplier is not enough for dexterity you have to stack it with savage attack and vulnerable attack to start shine.

 

Obviously this is mere theory-crafting.

 

Anyway, any volunteers to help me?

[Mysh]

Anyway, I also need to include the effect of armor delay to get a better estimate of optimum might/dexterity per weapon.

 

Do we know, what how armor affects combat speed? Can someone give me the whole formula for combat speed?

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

or as Demonjack suggested:

 

(attack_frame + recovery_frames + attack_delay + recovery_delay) / (1 + ((Dexterity - 10) * 0.0225))

 

Would armor increase recovery speed i.e. through attack speed mod?

[VaultDuke]

what was the average acc - deflection score you used in your formulas? the usefulness of the acc talents strongly depends on how good your acc is vs the average enemy's deflection.

 

also as I had a discussion about the sabre in one of my threads yesterday: on average, posters agreed, that slashing faces higher DR than other weapon types. so while the sabre does sport the highest 1 handed damage, weapons with -DR, 2 damage types or +acc may often still come out on top (thread was specifically about mace vs sabre though).

 

also, can you use your formula to map the usefulness of savage attack (overall damage difference per 100 attacks, with savage attack active or not) with high enchanted weapons and endgame talents vs the acc-deflection difference? that would be something I'd be interested about. how does this change depending on base weapon damage and enemy DR? if you stack high crit multiplier on a char, does switching on savage attack lose its usefulness earlier? if so, by how much?

all in all: I think once you bring some different chance to hit into the formula, things get wild, and stop being simple.

Edited April 14, 2015 by Gruftlord

[Mysh]

Things are not simple, for sure. I continually get different peaks for different conditions. I am currently confident though that dexterity builds get the edge on late game, while might builds yield higher DPS early. Let's keep some doubts though, till we verify the formulas.

 

Average_Foe_Base_Deflection=43.61
Average_Foe_Base_DR =8.85

 

They are also shown here: http://imgur.com/LS3n5p7

They are information for each foe are taken from prima guide.

 

However, I did not use the average in my calculations, foes were weighted by their size (delfection * DR). This means that the DPS against Adra dragon will count a lot more than a Xaurip, when optimising. Again neither the average size was used, but the cumulative weighted DPS against all of them.

 

The average foe size is highlighted by the red box in this image: http://imgur.com/Lm5kGNv

 

It is indeed correct that slash will face the highest overall DR, this though is not a very significant effect. Slash will face on average +10% DR from pierce, and crush -10% from pierce. Crush is by far the best. The best combined DR is crush/pierce.

 

The Sabre is pulling from the rest due to the impact on crits.

 

I do not think you mean per 100 attacks, since number of attacks is depending in dexterity. I think DPS would be better. What you've asked was shown here: http://imgur.com/HoUexDH

 

If you want something more specific, please give me specific input. However, if you want my advise for late game, do not think twice about savage attack + vulnerable attack + weapon focus. They are a must.

 

If you want to optimize your build towards something like the Adra Dragon the one-handed talent is handy. However, this is the only case it's useful. I would prefer to go with envenomed strike as a counter for bosses, as it brings a lot more damage.

 

I am happy to share my scripts, and desperately need a reviewer! ::wink::wink::

 

Thanks for your response.

Edited April 14, 2015 by Mysh

[peddroelm]

my current approach is going totally around the need for the various attack speed formulas/delays..

 

Chose 2 setup's parameters (weapons, skills , ) and then measure (in game) the number of swings each setup gets during 10 minutes of uninterrupted auto-attack .. This gives you an equivalent attack speed for weapon ..  (ex 300 swings during 600 seconds  ..equivalates to 0.5 attacks per second  .. Fast 1 handed weapon on 10 dex char, normal game speed.)

 

With this knowledge you can accurately calculate and compare maximal DPS ( all swings crit , weapon range set to average damage) between the two setups .. Maximal average damage inflicted in 10 minutes for setup 1 vs Maximal average damage inflicted in 10 minutes for setup 2..

 

fair and accurate ...

 

Counting 10 minutes worth of attacks from video recording is a huge chore and prone to human error (cannot copy paste combat log) - I'm still trying to find a way to get that combat log automatically into a file ..

Edited April 14, 2015 by peddroelm

[iszathi]

Anyway, I also need to include the effect of armor delay to get a better estimate of optimum might/dexterity per weapon.

 

Do we know, what how armor affects combat speed? Can someone give me the whole formula for combat speed?

 

Attack_Time = ( ( Attack_Frames(Weapon_Loop) + Recovery_Frames(Weapon_Loop) * ( 1 - Attack_Speed_Mod ) + Reload_Frames(Weapon_Loop) * (1 - Reload_Speed_Mod) + 10 ) / (1 + Dexterity*0.03 ) )/30;

 

or as Demonjack suggested:

 

(attack_frame + recovery_frames + attack_delay + recovery_delay) / (1 + ((Dexterity - 10) * 0.0225))

 

Would armor increase recovery speed i.e. through attack speed mod?

 

Thats what i understood for some testings, was trying recovery_frames*attack_speed_mod*armordelay.

 

//

with 0.2 attack_speed_mod you get

1-attack_speed_mod = 0.8. 

Shldnt it be

1/attack_speed_mod = 0.833?

 

Or are you accounting for that in the modifier? (its easy to check that the game does 1/asmod)

 

//

 

i havent used mathlab in a while but cld look at your scripts.