Accuracy bug layer

Discussions about AI types, units, tactics & strategy.

Accuracy bug layer

Postby Iluvalar » 27 Jul 2012, 19:17

[reserved for teaching player how to use it]
This is what might happen if the devs are too stuborn to fix the bug

What is the Accuracy Bug Layer ?
This layer is the most important thing to remember in your design screen. With an effect of about 40% of your total DPS (sometime higher sometime lower) it is a more important issue than to chose the good weapon modifier.

What happen is that depending on the exact accuracy of your weapon and the size of the target, the game engine might try to "miss" inside the hitbox of the unit causing an hit. Making the accuracy abruptly jump to 100%.

Why is that so bad ?
At first, you might think that taking into account the size of the target is a cool thing. But it's not done on a rational and intentional manner. It's just abruptly messing the game up. For exemple, you might fight an enemy correctly but than you decide to design mantis instead of python. And because mantis is 3 pixel wider than the python (exemple need fact checking), your enemy might as well jump from an accuracy of 60% to and accuracy of 100%. Making is weapon suddenly double damage on you. Same might happen after a random accuracy upgrade, or simply by getting closer to the target.

What is that topic
This topic will try to teach you exactly how and when to trigger the accuracy bug for you and how to avoid it for your enemy. This will soon become the #1 priority in all 3.1 games.
Heretic 2.3 improver and proud of it.
User avatar
Iluvalar
Regular
Regular
 
Posts: 1782
Joined: 02 Oct 2010, 18:44

Re: Accuracy bug layer

Postby JJjopando » 27 Jul 2012, 19:38

Wow, this is FASCINATING. I must see if what you say is true.
Proud former soldier of the 39th Infantry
BATTLE OF THE PLAIN OF REEDS - WE WILL NOT FORGET
User avatar
JJjopando
Rookie
Rookie
 
Posts: 26
Joined: 17 Jul 2012, 17:20
Location: Akron, OH

Re: Accuracy bug layer

Postby Andrie » 27 Jul 2012, 20:44

This is so INTERESTING :wink:
"My IRC en multiplay naam is Andrie"

Groete Andrie
User avatar
Andrie
Regular
Regular
 
Posts: 533
Joined: 20 Jun 2012, 14:11
Location: Suid Afrika

Re: Accuracy bug layer

Postby Giani » 27 Jul 2012, 21:20

I don't say that what you said isn't true, but I think you are exagerating a bit.
My maps: http://forums.wz2100.net/viewtopic.php?f=10&t=9501
User avatar
Giani
Regular
Regular
 
Posts: 804
Joined: 23 Aug 2011, 22:42
Location: Argentina

Re: Accuracy bug layer

Postby NoQ » 27 Jul 2012, 21:21

Nice, but i still don't understand one thing: doesn't every weapon have 100% accuracy most of the time due to firing at a unit mass rather than at a single unit most of the time?
User avatar
NoQ
Special
Special
 
Posts: 6157
Joined: 24 Dec 2009, 11:35
Location: /var/zone

Re: Accuracy bug layer

Postby Per » 27 Jul 2012, 21:37

Iluvalar wrote:if the devs are too stuborn to fix the bug

or not stubborn enough to implement a fix despite objections...
Per
Warzone 2100 Team Member
Warzone 2100 Team Member
 
Posts: 3745
Joined: 03 Aug 2006, 19:39

Re: Accuracy bug layer

Postby Iluvalar » 27 Jul 2012, 23:11

NoQ wrote:Nice, but i still don't understand one thing: doesn't every weapon have 100% accuracy most of the time due to firing at a unit mass rather than at a single unit most of the time?


No, first the ai chose target that it see in direct line in priority. Unless very special circumstance, that will be the first visible in the front. So at least the front 50% of the area around the unit will be empty. Second, the misses always land very close to the target. So even in a dense pack, it should fall in between units. Third, the standard strategy is to reduce the enemy number by focusing on one target at a time, having you damage falling on the wrong unit do not have the same impact at all.

Per, what objections ?
Heretic 2.3 improver and proud of it.
User avatar
Iluvalar
Regular
Regular
 
Posts: 1782
Joined: 02 Oct 2010, 18:44

Re: Accuracy bug layer

Postby Stratadrake » 29 Jul 2012, 00:05

And because mantis is 3 pixel wider than the python (exemple need fact checking)

Mantis geometry is (38W x 24H x 92L) whereas Python geometry is (39W x 21H x 102L). Now given the Mantis's rounder shape, I would wager that the Mantis looks (from the side) maybe 5% smaller than the Python, so the base chance-to-hit should be nearly the same between them.
Strata @dA, @FAC
User avatar
Stratadrake
Trained
Trained
 
Posts: 197
Joined: 07 Sep 2008, 09:43
Location: Pacific NW

Re: Accuracy bug layer

Postby Iluvalar » 29 Jul 2012, 01:16

Stratadrake wrote:
And because mantis is 3 pixel wider than the python (exemple need fact checking)

Mantis geometry is (38W x 24H x 92L) whereas Python geometry is (39W x 21H x 102L). Now given the Mantis's rounder shape, I would wager that the Mantis looks (from the side) maybe 5% smaller than the Python, so the base chance-to-hit should be nearly the same between them.

Where did you found those precise data ? So here is a factual exemple :
The radius of the mantis is 49 tick and the radius of the python is 54 tick. Therefore, the accuracy bug trigger at 75.5% of accuracy for the mantis but at only 73% for the python.

This mean that when a mg (any type) enter in close range with a python, that python suddenly lose 33% more damage. While the mantis is immune to that effect.
Heretic 2.3 improver and proud of it.
User avatar
Iluvalar
Regular
Regular
 
Posts: 1782
Joined: 02 Oct 2010, 18:44

Re: Accuracy bug layer

Postby Stratadrake » 29 Jul 2012, 06:08

Iluvalar wrote:Where did you found those precise data ?

Drhbod11.pie (python) and drhbod12.pie (mantis).

So here is a factual exemple :
The radius of the mantis is 49 tick and the radius of the python is 54 tick. Therefore, the accuracy bug trigger at 75.5% of accuracy for the mantis but at only 73% for the python.


?
Strata @dA, @FAC
User avatar
Stratadrake
Trained
Trained
 
Posts: 197
Joined: 07 Sep 2008, 09:43
Location: Pacific NW

Re: Accuracy bug layer

Postby Iluvalar » 29 Jul 2012, 07:58

The machine guns at short range try to miss at 50 ticks from the center of the target since the python is 54 tick wide even their "misses" hit and it's a 100% accuracy.

Because of that bug, the python is as good against machine gun at short range than cyborgs.
Heretic 2.3 improver and proud of it.
User avatar
Iluvalar
Regular
Regular
 
Posts: 1782
Joined: 02 Oct 2010, 18:44

Re: Accuracy bug layer

Postby Iluvalar » 07 Aug 2012, 05:38

Could someone that can play 3.1 could fire with an single machine gun on short range against a tiger, a mantis and a python simultaneously. It MUST be enemies units, because misses don't hit allied unit.

I need the time it take to destroy each of them please.
Heretic 2.3 improver and proud of it.
User avatar
Iluvalar
Regular
Regular
 
Posts: 1782
Joined: 02 Oct 2010, 18:44

Re: Accuracy bug layer

Postby Stratadrake » 07 Aug 2012, 19:42

The machine guns at short range try to miss at 50 ticks from the center of the target since the python is 54 tick wide

Are we talking radius or diameter? If the Python has a total width of 54 and the MG has a shot spread of 50 (in any direction from center), that is actually 54 / (50 * 2) = 54% accuracy, not 100%. Which means it's still a bug, but not as major an effect as you're thinking.

As for testing, sure we can fire up some Pythons or Tigers and send them out against enemy MG hardpoints (the type you so commonly see in Full T1 Bases maps). Obviously doesn't have to be MG's, they're just best for high ROF.
Strata @dA, @FAC
User avatar
Stratadrake
Trained
Trained
 
Posts: 197
Joined: 07 Sep 2008, 09:43
Location: Pacific NW

Re: Accuracy bug layer

Postby Iluvalar » 07 Aug 2012, 20:43

Radius.

I don't know who talked about spread in the actual game, but it's not me. When I say the MG miss fall 50 away from the center. It's really 50 away, not a spread between 0-100 away. So it's 100% accuracy if the target is larger than 50 of radius.
Heretic 2.3 improver and proud of it.
User avatar
Iluvalar
Regular
Regular
 
Posts: 1782
Joined: 02 Oct 2010, 18:44

Re: Accuracy bug layer

Postby Stratadrake » 10 Aug 2012, 00:42

Anyway, I tested a weapons mod with the following values:
- HMG altered for zero firepower, salvos of 10 shots with 2 sec salvo reload time, reclassified as Command weapon type (no accuracy upgrades exist for that weapon type). Left accuracy as is (75% short, 50% long)
- Zeroed all durability upgrades
- Zeroed all chassis HP; set Truck turret to exactly 101 HP

Tested using the Sk-Rush map, where it's only a few tiles away to the nearest enemy HMG hardpoints.

Initial tests show no appreciable difference whatsoever between Python and Mantis chassis, which are each taking damage about 75% of the time. Compared to rated weapon accuracy, this translates to about 50% of misses, comparable to given calculations in previous post.

At a rated accuracy of 75%, Mantis and Python took damage 100% of the time (a.k.a. 100% of misses), again no difference between them.

At a rated accuracy of 70%, Mantis and Python still showed no visible differences between them - both took damage about 100% of the time. Whereas under the same conditions, a Viper took damage about 90% of the time.

So yes this is a bug and the ramifications are just as serious as you say - but I don't see a sudden jump to 100% miss accuracy.
Strata @dA, @FAC
User avatar
Stratadrake
Trained
Trained
 
Posts: 197
Joined: 07 Sep 2008, 09:43
Location: Pacific NW

Next

Return to Skirmish / Multiplay