Chance to hit (accuracy) - does not sense in 3.1 (?)

The place to discuss balance changes for future versions of the game.
(Master releases & 3.X)
User avatar
Stratadrake
Trained
Trained
Posts: 197
Joined: 07 Sep 2008, 09:43
Location: Pacific NW
Contact:

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Stratadrake » 05 Oct 2012, 06:27

A better idea just hit me. Instead of having a default "scatter" based on arbitrary values, base these values on the actual weapon stats. So if a weapon's stats SAY it has 100% accuracy at 8 squares, this means the shot scatter at 8 squares does not exceed the diameter of a given-sized target.

And maybe instead of assuming a "cyborg"-sized target, how about a "user"-sized target? So a cyborg vs. cyborg (viper vs. viper, ptyhon vs. python, etc.) will see equal accuracy (relative to weapon specs, of course) but, say, a cyborg vs. python will show a small accuracy bonus favoring the cyborg.
Strata @dA, @FAC

User avatar
Iluvalar
Regular
Regular
Posts: 1802
Joined: 02 Oct 2010, 18:44

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Iluvalar » 05 Oct 2012, 07:33

"small" That's the problem with all your suggestions. Who said an angular scattering would cause a "small" effect ? It would be a tremendous game breaking change !

The same weapon would have an accuracy of 100% on a factory from as far as 26 tiles, but at the same time would have only half it chance to hit a cyborg as close as 4 tiles away. The distance of the target would become by far the most significant modifier of the game...
Heretic 2.3 improver and proud of it.

User avatar
Stratadrake
Trained
Trained
Posts: 197
Joined: 07 Sep 2008, 09:43
Location: Pacific NW
Contact:

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Stratadrake » 07 Oct 2012, 00:20

On the other hand, if a weapon's stats say it has 50% accuracy at short range of 4 squares (equally against any-sized target), then this means that in order "miss" a factory at 4 squares the shell must land a minimum 1.5 squares (up to 22º) off target. Whereas versus a cyborg at same range it only has to land more than 1/8 square (7º) off target.

Subsequently, if your weapon inflicts a blast radius or incendiary effect then the cyborg will nearly always get hit by it when the factory does not. What is the reason for that? Why should the same weapon, fired under the exact same conditions, show a huge shot scatter versus target A but a small shot scatter versus target B? (And by a factor of 12x?) I think that is the reason your patch has not been agreed with, because while it does fix the current behavior the fix does not make logical sense.

Bottom line, the problem is that the accuracy model and actual hit/miss are out of sync. In original WZ, they were out of sync but not biased (because the hitbox check was completely ignored). In 3.x they are out of sync and biased in favor of a "hit" - shot scatter is independent of range or target size and pretty much any weapon with at least 70% accuracy just doesn't scatter widely enough to "miss" the target ever.

And again, IMO the quickest, dirtiest fix to the current behavior is that a "miss" never inflicts direct-hit damage even if it strikes the target hitbox. It can still inflict a blast radius or incendiary (where applicable), but never a direct hit. When accuracy falls out of sync with the hitbox check this biases it in favor of a "miss", which means increased survivability in combat (but equally for all sides). Maybe that rocket didn't punch into a vital part of the factory's structure; maybe that bullet just glanced off the cyborg armor at an angle; maybe there was a manufacturing defect in that artillery warhead.
Strata @dA, @FAC

Reg312
Regular
Regular
Posts: 681
Joined: 25 Mar 2011, 18:36

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Reg312 » 26 Feb 2013, 11:26

Hello.
I still thinking about accuracy :)

We have following bug: 10% increase accuracy in some cases means 50% increase.
That happens because place of landing missed shot is directly depends on 'hitChance'

I suggest following:
1) make fixed placement of missed shots and make independent of hit chance.
(make fixed radius of circle of possible places there missed shot can land)
2) make spreading of missed shots wider on large distancec, and more narrow on close distances
- to prevent case when unit attack with 90 angle at close up distance.
3) prevent hitting direct damage to target if it was missed
- unit can wound another target with missed shot, but never wound primary targret (which missed)
- splash/periodical effects still applicable to primary target
4) do not place missed shots more far than maximum range of weapon + 0.5 tile


here is my formula for missed shots:
radius = (range*0.2)^(1/3)
there radius - radius of circle of missed shots
range - range from attacker to target
acc_formula.png
advantage of my variant is minimal changes to game. just correct current situation and do not re-make anything :)

as result we will get:
1) Working accuracy system. 70% chance to hit is real 70% chance to wound target.
2) More wider area of landing missed shots for long ranged weapons (this should look just real)

User avatar
Iluvalar
Regular
Regular
Posts: 1802
Joined: 02 Oct 2010, 18:44

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Iluvalar » 26 Feb 2013, 16:09

Same old s***... sorry.

Except you don't take into account the accuracy of the weapon for your radius calculation ? Why not ?

Why preventing the bullets to land farther than the max range ? It's quite frequent in real life.
Why the angular missing area would narrow while the enemy go back ? (not that I care, for the game play, but i'm curious to know how you came up with that odd 1/3)
Heretic 2.3 improver and proud of it.

Reg312
Regular
Regular
Posts: 681
Joined: 25 Mar 2011, 18:36

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Reg312 » 26 Feb 2013, 16:26

Iluvalar wrote:Same old s***... sorry.
We have simple situation: accuracy working bad right now, and it lasting for years.
We having only infinite conversations.
We need change accuracy now, stop arguing!
My patch should be ok until someone genius will suggest something better and well undrerstandable.
Iluvalar wrote: Except you don't take into account the accuracy of the weapon for your radius calculation ? Why not ?
Accuracy is "chance to hit", radius of missed shots is something another than chance to hit.
see in statsdef.h we have "UDWORD longHit; ///< Chance to hit at long range"
Iluvalar wrote: Why preventing the bullets to land farther than the max range ? It's quite frequent in real life.
i think half-tile should be ok to make some visible randomness.
"real-life" is not argument because my suggestion serves to fix current problem with buggy hitChance.
some weapons like Heavy Cannon can receice noticeable damage bonuses when they miss farther than max range.
Our task is reduce randomness in balance stuff. that's why.
Iluvalar wrote: Why the angular missing area would narrow while the enemy go back ? (not that I care, for the game play, but i'm curious to know how you came up with that odd 1/3)
i get 1/3 simply playing with graphs.
close-up radius should be narrow and long-range radius should be wider. it looks quite real and quite balanceable.

Per
Warzone 2100 Team Member
Warzone 2100 Team Member
Posts: 3775
Joined: 03 Aug 2006, 19:39

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Per » 26 Feb 2013, 17:28

I think the biggest problem is that we now have two accuracy systems. One based on physics, and one based on accuracy rolls. You can massage the algorithms mixing the effects of the two as much as you want, but in the end the inputs are just too complex for the output to make any sense to the player.

This game just ain't big enough for the both of them. One night they will meet in dark alley, and only one of them will walk out alive... :ninja:

Reg312
Regular
Regular
Posts: 681
Joined: 25 Mar 2011, 18:36

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Reg312 » 26 Feb 2013, 17:43

We have lots of randomness and complex things in game, and this making harder to balance game strategies.
I'm asking to do something to improve game play for current game state, because me and other players playing warzone these days, not in further future... :roll: :lecture:


first of all, i think this is wrong formula: int missDist = 2 * (100 - resultHitChance) + minOffset;
there missDist - radius of circle where missed shot laning
resultHitChance - chance to hit
minOffset - constant, = 5

@per, i know you can suggest to remove hitChance and hit 100% all time.
i can agree with it, but i think visual effect will be not good. we used to see projectiles which with some distribution on enemy army.

so next variant: is simplify accuracy formula as possible.
e.g. set int missDist = 128 (fixed radius os missed shots and that's all)
.... again we having conversations :ninja:

Originway
Trained
Trained
Posts: 412
Joined: 08 Aug 2012, 06:22

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Originway » 26 Feb 2013, 21:16

Per wrote:I think the biggest problem is that we now have two accuracy systems. One based on physics, and one based on accuracy rolls. You can massage the algorithms mixing the effects of the two as much as you want, but in the end the inputs are just too complex for the output to make any sense to the player.

This game just ain't big enough for the both of them. One night they will meet in dark alley, and only one of them will walk out alive... :ninja:
this ticket fixes the issues http://developer.wz2100.net/ticket/3748 but we can't test since nobody is making any builds from it!!!!!! :( :( :cry:

User avatar
Iluvalar
Regular
Regular
Posts: 1802
Joined: 02 Oct 2010, 18:44

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Iluvalar » 27 Feb 2013, 00:22

Originway, This is not the same bug.

Dont get me wrong reg312. I totally agree this must be fixed.

However my patch (http://developer.wz2100.net/ticket/3559) is already written and do your points 1), 2) and 3) just as well. The only questions I ask is why you ask for these small variations of my fix. I've been told we should try to emulate an angular variation accuracy as much as possible.

So i'm asking again : Why should the angular accuracy increase when the enemy go away ?
Heretic 2.3 improver and proud of it.

User avatar
Iluvalar
Regular
Regular
Posts: 1802
Joined: 02 Oct 2010, 18:44

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Iluvalar » 27 Feb 2013, 00:38

Per wrote:I think the biggest problem is that we now have two accuracy systems. One based on physics, and one based on accuracy rolls. You can massage the algorithms mixing the effects of the two as much as you want, but in the end the inputs are just too complex for the output to make any sense to the player.

This game just ain't big enough for the both of them. One night they will meet in dark alley, and only one of them will walk out alive... :ninja:
I think I already demonstrated that a full physic accuracy system would cause sweat spots (1-2 tiles wide) where a unit could get 50-100% stronger than anywhere else. It would suddenly become more important than research, weapon modifiers and design all together. The knowledge of those spots and there usage would become by far the main (if not the only) aspect of the game. Keeping your units between 5 or 6 tiles from your opponent would become your only priority. The first player to move one tile away from is strategy would lose badly. XD

That sound like an interesting concept for a new game. It might be fun to play. But it certainly doesn't seem like warzone 2100

Addendum : Sorry, after running more simulation, the sweat spot would actually be the spot were both weapons are balanced. Closer the bad accuracy kick ass, farther the good accuracy dominate.
Heretic 2.3 improver and proud of it.

Reg312
Regular
Regular
Posts: 681
Joined: 25 Mar 2011, 18:36

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Reg312 » 27 Feb 2013, 08:04

Iluvalar wrote: So i'm asking again : Why should the angular accuracy increase when the enemy go away ?
because it looks more real,
when flamers attack enemy army - they should not fire at twice range or with big angle.
for close up fight radius of missed shots should be narrow.

another case is long ranged weapons, if your ripple rockets fired with defective rockets then this bad rockets can go in wider area, not just exactly near target.

User avatar
Iluvalar
Regular
Regular
Posts: 1802
Joined: 02 Oct 2010, 18:44

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Iluvalar » 27 Feb 2013, 22:45

I'm talking about angular deviation

Let me make a clear exemple :
I'm shooting with my HMG at a group of scavengers 3 tiles in front of me. At that point, my bullets spread at 10° angle in front of me. When they realize they are doomed, the scavengers decide to flee. With YOUR equation, my shots would reduce to 9° then 8° then 7... until they leave my maximum range while i'm shooting at 5° angle in front of me.

My question is : Why would my weapon seem more accurate when the enemy increase his distance to me ?
Heretic 2.3 improver and proud of it.

Reg312
Regular
Regular
Posts: 681
Joined: 25 Mar 2011, 18:36

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Reg312 » 27 Feb 2013, 23:51

Iluvalar wrote:I'm talking about angular deviation

Let me make a clear exemple :
I'm shooting with my HMG at a group of scavengers 3 tiles in front of me. At that point, my bullets spread at 10° angle in front of me. When they realize they are doomed, the scavengers decide to flee. With YOUR equation, my shots would reduce to 9° then 8° then 7... until they leave my maximum range while i'm shooting at 5° angle in front of me.

My question is : Why would my weapon seem more accurate when the enemy increase his distance to me ?
hmm, ehh. thanks for example
1) why we need discuss angle deviation?
farther range = less angle OR ripple rockets should miss in half-map area?
i think it is normal when ground shaker can miss in 5x5 tiles area. (circle with radius 2.5 tile)

angle = atan((x*0.2)^(1/3)/x)*180/pi * 2 (?) - not sure is it correct.

2) my formula makes lesser angle decrease with increased range
lesser than we have now(!)

3) i dont see your 9° 8° 7° 5°
angle of 9 is equal to ~20 tiles range (?)

--------------------------------------------------------------
why we need think about angular deviation?
i believe in future we can extend 'hitChance" and add more parameters to accuracy calculation


I like realism. hmm, but for such angular realism we need more than 1 formula

User avatar
Iluvalar
Regular
Regular
Posts: 1802
Joined: 02 Oct 2010, 18:44

Re: Chance to hit (accuracy) - does not sense in 3.1 (?)

Post by Iluvalar » 28 Feb 2013, 03:31

Dont you realise that your bullets are curving mid-air ?

With an accuracy of 80% + at least one upgrade and the gaussian distribution in my patch, most ripples missiles hit (90%) or fall short (7%) from the target with some quite epic exceptions once in a while. Which still seem natural for a computer guided missile.
Attachments
accuracypatch.png
Heretic 2.3 improver and proud of it.

Post Reply