GH2: Combat System Testing

[Joseph Hewitt]

Here's the latest, and largest, output from Attacktest. It took somewhere around 4 hours to complete this test with 2000 trials per weapon. I've fixed a number of problems since the last post- there was an issue with powerful weapons throwing up smoke clouds, thereby interfering with future rolls. The pilots were originally standard Mecha Pilots, but this caused problems because the archetype has higher Reflexes than Perception. They've been changed to Arena Pilots who have straight 11s. The attacker is a SAN-D1 Daum. The target is an AD26c Vadel. Both pilots are at renown level 50.

The first test is DC. A LAS-10 Laser Cannon is used as the base weapon. Its DC is modified for each test. First the weapon is tested without any attributes, then it's tested with Brutal, ArmorPiercing, and Brutal+Scatter attributes.

Code: [Select]

LAS-10 DC5: Mean 36, Median 50  ($94500)
LAS-10 DC5+B: Mean 26, Median 26  ($189000)
LAS-10 DC5+AP: Mean 26, Median 25  ($189000)
LAS-10 DC5+BS: Mean 24, Median 25  ($189000)
 
LAS-10 DC10: Mean 16, Median 17  ($189000)
LAS-10 DC10+B: Mean 13, Median 12  ($378000)
LAS-10 DC10+AP: Mean 12, Median 11  ($378000)
LAS-10 DC10+BS: Mean 12, Median 12  ($378000)
 
LAS-10 DC15: Mean 10, Median 8  ($297650)
LAS-10 DC15+B: Mean 8, Median 8  ($595350)
LAS-10 DC15+AP: Mean 8, Median 7  ($595350)
LAS-10 DC15+BS: Mean 8, Median 8  ($595350)
 
LAS-10 DC20: Mean 8, Median 5  ($396900)
LAS-10 DC20+B: Mean 7, Median 6  ($793800)
LAS-10 DC20+AP: Mean 6, Median 5  ($793800)
LAS-10 DC20+BS: Mean 6, Median 6  ($793800)
 
LAS-10 DC25: Mean 6, Median 5  ($496100)
LAS-10 DC25+B: Mean 6, Median 3  ($992250)
LAS-10 DC25+AP: Mean 5, Median 4  ($992250)
LAS-10 DC25+BS: Mean 5, Median 5  ($992250)
The next test is Burst Value. Again, a LAS-10 is modified for use in the test. Various configurations with the same "total DC" are tested.

Code: [Select]

LAS-10 DC24x1: Mean 6, Median 5  ($476250)
LAS-10 DC12x2: Mean 9, Median 9  ($428550)
LAS-10 DC8x3: Mean 10, Median 9  ($393100)
LAS-10 DC6x4: Mean 11, Median 10  ($385550)
LAS-10 DC4x6: Mean 11, Median 10  ($378000)
LAS-10 DC3x8: Mean 12, Median 11  ($374200)
The next test is Attack Attributes. The base weapon is a SC-9 Shaka Cannon. Various attack attributes are given to the weapon to see how these affect performance. In addition, several accuracy modifiers are tested. The "Ultima" test gives the SC-9 a +5 accuracy bonus and the attributes ArmorPiercing, Brutal, Overload, Scatter, and Hyper.

Code: [Select]

SC-9 Shaka Cannon: Mean 26, Median 23  ($29000)
SC-9 ACC+2: Mean 20, Median 22  ($40300)
SC-9 ACC+5: Mean 15, Median 13  ($57350)
SC-9 ACC-2: Mean 34, Median 50  ($23300)
SC-9 ACC-5: Mean 47, Median 50  ($14800)
SC-9 ArmorPiercing: Mean 21, Median 18  ($57350)
SC-9 Brutal: Mean 20, Median 16  ($57350)
SC-9 Overload: Mean 16, Median 16  ($43150)
SC-9 Scatter: Mean 27, Median 27  ($29000)
SC-9 Scatter + Brutal: Mean 16, Median 16  ($57350)
SC-9 Hyper: Mean 7, Median 5  ($425900)
SC-9 Mystery: Mean 30, Median 50  ($37500)
SC-9 Ultima: Mean 2, Median 2  ($5144800)
The final test uses the standard weapons from STC_Default.txt. Note that range modifiers cause some of the weapons to perform worse than expected (the Vulcan Cannon and Superheavy Particle Cannon especially).

Code: [Select]

MBAZ-17 Mecha Bazooka: Mean 8, Median 9  ($128550)
PAR-2 Particle Cannon: Mean 44, Median 50  ($102050)
PAR-6 Heavy Particle Cannon: Mean 25, Median 26  ($262400)
PAR-13 Superheavy Particle Cannon: Mean 20, Median 16  ($331650)
PHS-8 Phase Cannon: Mean 19, Median 15  ($270000)
PHS-25 Heavy Phase Cannon: Mean 12, Median 10  ($607500)
GR-12 Gauss Rifle: Mean 18, Median 16  ($39200)
GR-24 Heavy Gauss Rifle: Mean 16, Median 12  ($85350)
MAC-4 Machine Cannon: Mean 21, Median 21  ($60700)
RG-8 Railgun: Mean 8, Median 7  ($191800)
RG-16 Heavy Railgun: Mean 5, Median 3  ($292950)
MAC-2 Light Machine Gun: Mean 49, Median 50  ($14650)
VC-5 Vulcan Cannon: Mean 17, Median 13  ($205250)
SC-9 Shaka Cannon: Mean 25, Median 24  ($29000)
MB-7 Mecha Blunderbuss: Mean 11, Median 11  ($107550)
MRIF-5 Mech Rifle: Mean 44, Median 50  ($31250)
Finally, the skill level of both pilots is shown.

Code: [Select]

Attacker Skill = 9
Defender Skill = 9
One very interesting finding is that SCATTER by itself tends to make a weapon worse, but SCATTER+BRUTAL is somewhat more effective than BRUTAL alone. Increased accuracy doesn't count for much when the attacker is at a natural advantage (as shown by previous tests of Daum vs Daum), but become very important once the defender gains the upper hand.

I'm thinking about running the BV test with some weapon attributes. I expect that there would be differences with ArmorPiercing or Scatter.

[JohnnyDmonic]

Seems to me that BV really isn't all that out of line, given its costs.  Maybe try testing a DC16 single shot vs the RG-16, and give the single shot something like ARMORPIERCING, or HYPER.  Whichever gets it closer cost-wise.

[Joseph Hewitt]

Not really. The BV test was mostly to judge the combat effectiveness of burst fire, not so much the cost. I just ran a quick simulation with just the shaka cannon to test BV against other weapon modifications. Here are the results:

Code: [Select]

SC-9 Shaka Cannon: Mean 26, Median 23  ($29000)
SC-9 BV3: Mean 10, Median 10  ($74350)
SC-9 BV5: Mean 7, Median 6  ($119700)
SC-9 BV8: Mean 5, Median 4  ($187750)

If you compare these results to those above, you'll see that high BV weapons get better results for lower cost than any other modification.

[JohnnyDmonic]

Hmm, I see what you mean, although there's also the added ammo weight/cost issue. (and just plain running out faster in most cases.  Shaka Cannon has 15 shots, lot of other single shot weapons get 30+, most BV weapons get 10 shots it seems)  Plus doesn't BV add to weight of the weapon?  Or was that just my imagination.  Still, it does seem out of line cost-wise when you lay it out like that side by side with the others.  I'm pretty tired at the moment, but was there something about BV adding to accuracy?  And some mention of maybe removing that or mitigating it?  If so maybe there's something to that.  I like the potential for ludicrous damage that BV has, but maybe it should be less easy to deliver it.  Does the game currently roll individually per bullet, or...eh, tired.  I remember HEROS system? I think it was?  Rolled each shot individually with a stacking penalty per shot.   But that was to take into account recoil I think.  Doesn't make as much sense with beam weapons.

[Daemonward]

Doesn't make as much sense with beam weapons.

There are no burst firing beam guns.

If you hold down the trigger of a beam gun, you just get a continuous beam.

[plllizzz]

Doesn't make as much sense with beam weapons.

There are no burst firing beam guns.

If you hold down the trigger of a beam gun, you just get a continuous beam.

I care to disagree: Pulse lasers

[Daemonward]

I care to disagree: Pulse lasers

Gahh! Foiled again by reality!

But mark my words. One day we shall all have laser hoses!

[plllizzz]

And lightsabers! *cough*

But back on topic - I doubt recoil matters much for an object so massive as mecha

it would prolly only matter in case of really BFGs, but those can be justified as using magnetic/electromotoric forces [Gauss/Rail-guns]

AS for BV, any kind of accuracy penalty for bigger burst seems unrealistic to me - we havbe hi-tech targetting programs and we shoot to targets bigger than most buildings, so seriously - the more, the merrier

also, one idea that came with realization - some of the biggest Rail/Gauss/Other-big-mass-drivers should have their own power cells instead of low firing speed

[peridot]

Er, electromagnetic launchers will have recoil too, in principle just as much as chemical-driven weapons of the same caliber and projectile speed. On the other hand, one current approach for building armor-piercing weapons is to go for small-caliber high-velocity projectiles, which (for the same muzzle energy) have a lot less recoil. Chemically-powered weapons that use this idea fire "darts" much narrower than the nominal caliber, while current railgun designs are simply designed for very high muzzle velocities (in principle they avoid an upper limit set by the speed of sound in the expanding gases). That said, this strategy for piercing armor results in less damage to unarmored targets - the projectile just goes straight through, leaving a small hole, rather than expending all its energy doing damage inside the target.

Are these tests being done with a stationary target, or one moving at full speed? I expect the balance to change a lot as the target gets harder to hit; that's why I like SCATTER, for example.

[Daemonward]

The final test uses the standard weapons from STC_Default.txt. Note that range modifiers cause some of the weapons to perform worse than expected (the Vulcan Cannon and Superheavy Particle Cannon especially).

Code: [Select]

MBAZ-17 Mecha Bazooka: Mean 8, Median 9  ($128550)
PAR-2 Particle Cannon: Mean 44, Median 50  ($102050)
PAR-6 Heavy Particle Cannon: Mean 25, Median 26  ($262400)
PAR-13 Superheavy Particle Cannon: Mean 20, Median 16  ($331650)
PHS-8 Phase Cannon: Mean 19, Median 15  ($270000)
PHS-25 Heavy Phase Cannon: Mean 12, Median 10  ($607500)
GR-12 Gauss Rifle: Mean 18, Median 16  ($39200)
GR-24 Heavy Gauss Rifle: Mean 16, Median 12  ($85350)
MAC-4 Machine Cannon: Mean 21, Median 21  ($60700)
RG-8 Railgun: Mean 8, Median 7  ($191800)
RG-16 Heavy Railgun: Mean 5, Median 3  ($292950)
MAC-2 Light Machine Gun: Mean 49, Median 50  ($14650)
VC-5 Vulcan Cannon: Mean 17, Median 13  ($205250)
SC-9 Shaka Cannon: Mean 25, Median 24  ($29000)
MB-7 Mecha Blunderbuss: Mean 11, Median 11  ($107550)
MRIF-5 Mech Rifle: Mean 44, Median 50  ($31250)

How much do these numbers change when the pilot has Spot Weakness? Is it worth investing a skill slot and tens of thousands of XP?

[SharkD]

What does Attacktest do? Pit a number of AI controlled characters against each other in combat and iterate through a number of weapons?

Which weapons change in effectiveness depending on the terrain? Missiles can fly over hills, and beam weapons need a straight line-of-sight, correct? What about flame inducing weapons? They cause forests to catch fire, right?

I'm still thinking terrain should play a greater role. Maybe have fuel dumps positioned around the map that cause large amounts of area damage when destroyed. More objects that create partial (chest high) or complete (head high) cover and concealment would be good too. As would muddy or rocky terrain that affect walking and rolling vehicles differently, and low cloud cover or fog that inhibits flying vehicles.

[peridot]

I'm still thinking terrain should play a greater role. Maybe have fuel dumps positioned around the map that cause large amounts of area damage when destroyed. More objects that create partial (chest high) or complete (head high) cover and concealment would be good too. As would muddy or rocky terrain that affect walking and rolling vehicles differently, and low cloud cover or fog that inhibits flying vehicles.

The real problem, for GH2, is all those completely empty space maps. They mean you have to be able to win without any help from terrain. So terrain only makes a big difference if the enemies can take advantage of it to make your life substantially harder; for the most part they can't. I'd be inclined to move all, or almost all, space missions onto (or near) asteroids/comets/ring particles or SF:3 spacecraft that block sight or allow you to enter and fight indoor battles.

In GH1, with almost the same mechanics, I found I took advantage of terrain in a number of ways. The base destruction missions, I was able to win really hard ones by using a mech that could jump or fly and that had a long-range high-damage mission: pop up so I can see over the base wall, blow away the generator with a long-range salvo, land and make a run for it. Usually the enemies would only get in a round or two of shots at me, during which I'd be flying. The terrain also sometimes came in handy with the escort missions: I'd lay a big smoke cloud around the densest clump of enemies, and by the time they could see to shoot their target was usually out of range. In GH2, I find the city missions particularly brutal because I usually try for long-range sniping, and often I end up rounding a corner and winding up four hexes away from a Dragon Savin or whatever. When I win these it's by carefully using the city to separate the enemies.

As for fire, I'm still holding out for the EXTINGUISH flag - weapons that can put out fires (and have SMOKE as a side effect), so that you can have firefighting mecha and firefighting missions ("Put out the fire before it spreads to destroy the area!" or "Save the hospital from burning down!" or "Rescue people from the building before it burns") and on the other hand you have to watch out which weapons you use in inhabited areas for fear of destroying whole spinners, or just yourself, in an inferno. In game mechanics terms it would be pretty simple; an extinguish weapon would be LINE, SCATTER, or AREA, and generally (though not necessarily) do no damage, and would put out any fires in the squares it affects. This could include BURN on mecha too, optionally, so that you could put yourself or your comrades out. You might also want to beef up the burning-terrain code, so that each square had a flammability and maybe so that fires could have different intensities (i.e. likelihood to set nearby things on fire). But things like cutting firebreaks and maybe even preventative burns would become usable tactics without any additional work. And, of course, it would give some meat to the "Firefighter" profession.