Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[timferius]

But dog-fighting in space is both fun AND looks cool! So I'm just not sure why all the rage?

[NullForceOmega]

Because missiles in space are stupid, and I hate stupidity in a way I cannot articulate other than to seize the offending (object/entity) and completely destroy it.  I'm getting better at controlling that impulse tho'.  I'm fine with dog-fighting with energy weapons or ballistic weapons, but when missiles enter the picture I lose it.

[Rez]

In the interest of not derailing further, I'm going to respond in the space thread in general discussion.

http://www.bay12forums.com/smf/index.php?topic=132929.0

also, be calm, dawg.

[Niveras]

One of the reasons I don't think Newtonian space would necessarily be super difficult is that with science fiction (hell, even modern science... hell, even 60s science) you can have maneuvers that are too complex for a person to regularly pull off be handled by the computer. For example, you could still have "space brakes", essentially thrusters that would automatically orient and fire in the opposite direction of your current velocity (linear or angular); they'd just be more like brakes on an actual car and engage when you wanted them to and take time to slow you down, instead of instantly taking effect the moment you stop accelerating.

If you're going to go so far as to have ship AI act as a compensator to pilot actions in order for space flight to be easier or make more sense, it's not a long leap then to say that space flight in general will probably just be "tell ship to go there" and there ends the pilot's control. Because in order for the AI to compensate a live pilot in a way that won't occasionally result in a flaming wreckage, the AI is going to have to understand way more about space flight in general and the various possibilities in the specific situation than the pilot himself, at which point it should just do all the flying itself.

I do agree that a lot of space-related media, particularly space-fighter-sims, tend to mimic atmospheric analogs too much (not just jet fighters but also, for example, mapping space ships to terrestrial navies). I don't mind it too much, just as I don't mind how a fantasy game gives us tools in the form of various magic but doesn't explore how a reasonably intelligent person would bend that magic in ways not seen in the game. But if we're going to nitpick that stuff, I'll end up going a hell of a lot farther than just the existence of non-newtonian motion (for example, the idea that any part of the ship at all would have direct visual access to outside, that isn't otherwise intended as a means of egress. I expect a command center to be more like a Minbari from B5 than any iteration from Star Trek or Star Wars).

[alexandertnt]

If you're making your AI dumb to fix mechanics, I'd advise you change the mechanics.

It is often necessary to make the AI dumb in many games, usually because the computer is smarter than you are, and losing over and over is not fun.
One of the biggest implications I can think of here is that raycast weapons in FPS's couldn't exist, because the AI has to have it's aiming crippled to be any fun.
The trick is to not make your AI look dumb. For example, randomising the aim on bots makes them dumber, but you don't see it as being dumb because it more closely resembles how a human would play. Making the AI not look dumb in this scenario seems very doable within this newtonian game to me.

I think most space games largely ignore the implications of missiles, mines, and drones.  Try your high-speed flyby on a big ship; unless you have miracle technology, you aren't going to survive the mines and drones that you have to fly through to intercept him.

Some games I have played don't ignore this, and make shooting down capital ships impossible, because your not supposed to shoot down capital ships.

The solution to this problem could include
A) Not have capital ships/make them unkillable and not the focus, but the background
B) While your inventing futuristic space mines, invent some futuristic space mine defences. No reason why weapons should advance and defensive technologies stagnate.

Same issue with jousting.  Maintaining range means you're trading blows hoping your armor lasts longer.  The tactical situation in most of these games is dirt simple and much less deadly than what Newtonian physics, semi-autonomous weapons, and CIWS would really dictate.

I don't think maintaining range is how jousting works, I imagine it would involve two ships flying towards each other. Regardless, why would you maintain velocity with an enemy who is facing you? That sounds like suicide, bank up or something.

I feel like using a more accurate depiction of space is a fluffy decision with important crunchy implications; a decision for the game to be more speculative sci-fi than science fantasy

I already explained how such a game would be quite different to the current atmospheric dogfighting games we have, I don't see how changing the way a game worked, and therefore the game mechanics, could be described as a "fluffy decision". It seems to be very much the opposite.

Why do other aspects have to be realistic, or have anything to do with speculative sci-fi? I think newtonian physics would be a neat idea to make a space shooter fun and original, independent of it's realism.

Regardless, if you really need it to be "realistic", think outside the box a bit. Actual wars may be fought with AI and computers, but that doesn't mean our newtonian game has to be set in a war. Make it a competition, a sport with spectators. That would explain why there are human poilots, non-computer-assisted flying, limited arena size etc. Many constraints can be put in place in the name of fun that would be justified by the setting.


Don't get me wrong, I don't want to say your perception is incorrect or anything, I just want to explain why I think such a game would still be workable overall. I think you are just focusing on a different aspect (realism and setting consistancy) then I am (fun pew-pew ships with a novel movement mechanic that allows for a new and different experience than "WWII dogfighting IN SPACE").

[Rez]

Why do other aspects have to be realistic, or have anything to do with speculative sci-fi?

Cause I say so and I want a game with that fluff. 

I assumed you meant make the AI too stupid to run away from the player when they could and would lose a fight.  That's a different kind of stupidity from making aimbots merely human.

I think a newtonian racing game has potential.

[Delta Foxtrot]

If you're making your AI dumb to fix mechanics, I'd advise you change the mechanics.  That's one of my gaming pet peeves.

Reminds me of that one talk by a Civ IV designer. Basically do you design an AI that tries to win, or do you design an AI that tries to be fun to play against? They're not necessarily the same thing and I know I generally prefer the second one.
Link (1 hour) for those interested.

[alexandertnt]

I assumed you meant make the AI too stupid to run away from the player when they could and would lose a fight.

Run away to where, exactly? The depth of space? Because it's so much safer out there .

If there was something like a repair platform, then yes, the AI should probably try to run away to it at some point. But the AI could only get away in specific circumstances. For example if the AI decided to run away and you were between it and the platform, there would be plenty of oppurtunity for the player to score a kill. If you were already travelling towards it at a higher speed than it was travelling away, then you would inevitably catch up, and be able to score a kill. It seems that the problem with being unable to catch an enemy would only really happen occasionally, in which case thats fine and part of the game play. You yourself could make maneuvers in such a way that would allow a good chance to escape if you became too damaged, which sounds like something fun to learn and master.

I think a newtonian racing game has potential.

I agree actually, that sounds like it could be a lot of fun.

[Nick K]

The 'Atomic Rockets' site is one I hadn't looked at for years, but they talk a lot about the hard sci-fi implications of various space weapons. They seem to think that missiles are a plausible space warfare option - unlike fighters. See under missiles here: http://www.projectrho.com/public_html/rocket/spacegunconvent.php

[Shadowlord]

They're a lot more plausible than trying to hit a very distant ship with unaimed projectiles, especially considering the energy costs to propel anything to relativistic speeds. You don't have to see something coming to avoid it, either. Just move unpredictability, and sufficiently to prevent aim prediction from working, and hope the enemy doesn't get lucky (if you've got a ton of missiles, say bomb pumped laser missiles, like in the honorverse, to swarm the enemy with at once, and they have railguns... Should go well for you as long as you don't get close and you can overwhelm their PD).

[Rez]

Derail unsuccessfully derailed.

[Rolan7]

It's interesting, artificial intelligences are vastly better at us at some tasks and much worse at others.  I'm going to discuss a favorite game of my childhood because I'm bored and [redacted].

Consider Unreal (or Unreal tournament *I GUESS*).  Only 3/10 of the weapons were hitscan, and only one of those (the sniper rifle) did massive spike damage.  All the rest fired projectiles with various gimmicks, dodgeable at moderate range.  Programming an AI to 360-noscope with a hitscan weapon is trivial... but these other weapons (potentially far more dangerous) present a computationally difficult problem. 

Consider one of the more typical weapons of those 7, the Eightball Cannon.  It's basically a rocket launcher.  Except that it has 8 (6?) chambers, which load over the course of several seconds.  You can fire with only a few cylinders loaded though.  So you need to anticipate your enemy about 2-5 seconds before lining up the shot.  Also, you can fire the rockets as bouncing grenades which explode on contact with flesh somehow.  You have to choose grenades vs rockets when you start loading the chambers.  Also, if you hold both fire buttons, it fires a special concentrated ring of rockets instead of the normal spread.

Okay, maybe that's not the *most* typical of the non-hitscans.  There's also the Razorjack (number 7~) which fires linear projectiles at a decent projectile speed.  Even here with have the problem of "boxing in" an enemy so they can't just dodge.  I honestly can't remember whether the AI tried, I think it just fired on the center of mass.  Which was great.  The projectiles moved fast, but could be dodged with fast reactions - an ideal challenge for players.  FURTHERMORE, there was the alternate fire mode where the projectiles can be "steered" by looking in different directions.  This isn't the target-designation of the Half Life RPG, no.  This triggered on the user looking to the left-right-up-down, and the projectile blades vectored to match.  Imagine coding an AI for that.  Unsurprisingly, the AI seemed to use that feature - but at random.  Whereas a masterful human player could bend projectiles around a corner.

Oh wow, I momentarily forgot that the projectiles for that weapon *bounced off walls*.  It was a weapon designed to punish enemies hiding around corners or cover.  Almost as much as...

The Flak Cannon.  The scrubbiest, yet satisfying weapon of the set.  Possibly the easiest for an AI to use.  The primary fire simply fired a shotgun-esque blast of shards with high spread.  At close range, no dodging was possible and the damage was extreme.  At moderate range dodging was unlikely but also unnecessary, since only a few shards would hit (and the fire rate was quite low).  So at moderate range the mortar-mode, which fired a canister which exploded on impact with the ground, was preferable.  To return to the topic, an AI could actually handle the secondary mode.  The projectile moves quickly, and all it has to do is aim the canister's Newtonian arc at the enemy's feet.  Either it hits the enemy for high damage, or it explodes on the floor close by for moderate damage.

Ah, the ASMD.  Or as known by cultureless rubes, the shock rifle.  The weapon of a master and a newbie at the same time.  I never mastered it.  The primary fire was basically hitscan (a graphical effect implied it was non-hitscan, but sources say it was).  So a novice like me could use it that way.  But the true power was the alt-fire: a moderately-fast energy ball which, when hit with the primary energy lance, made a large damaging explosion.

Anyway my point is that a good AI for Unreal would be incredibly complex and even then would probably fail against experienced players, particularly with team dynamics.  Because the mechanics are computationally intense.  With a newtonian space game, I don't know...  Even with slow-moving projectiles, an AI programmed to win may be superior to a reasonably skilled human.

[Shadowlord]

The shock rifle was spectacularly effective at long range in UT, especially if you could predict the enemy's future location well enough to get the blue ball (second fire mode) near an enemy and then light it up with the primary fire mode (for the massive blast damage that resulted).

[itisnotlogical]

Reminds me of a glitch in Perfect Dark, the FPS of my childhood: the AI doesn't know how to use remote-trigger mines. If you create a multiplayer match where the only weapons are remote mines, even the most difficult bots will be helpless. They also don't understand the implications of some weapons; i.e. they don't intentionally use tranquilizers to blur your vision, and if they're hit with tranquilizers their aim won't be impaired.

[UXLZ]

The best way to farm kills in Perfect Dark was to set up an infinite match, or a match with a enooormous kill limit, put all the AI on a team against yourself, make the only weapon the Laser Watch then climb to the top of one of those towers in A51, equip the Laser Watch, set it to Beam Mode , tape the Z trigger down and point it at where the enemy's heads will go when they go up the ladder.

Leave it on overnight.

Wake back up, 800 kills. Blam.