Smart Combat, Unit AI - Why not?

All the way signed for the unit AI. I'd also like some advanced group AI to control armies or bases with accordance to selected protocol, like guard/patrol/harass/attack for army and expand/fortify/produce for base. It would be useful to have a "virtual subordinates" to manage your tons of bases and armies when attention is needed elsewhere. From what I know of Sorian, he is capable to implement AI planning from unit level to team level as a single framework. And after all, in real war army commander never controls movement of a single soldier, since it draws too much attention away from the whole picture.

I refuse to answer this question, as "UnitAI" might range from units that can barely find a path up to units that are able to think about strategies that are superior to everything I myself could come up with.

Also we had a quite big discussion about this already, where I've posted quite a lot:
viewtopic.php?t=37369

I am finished with this discussion. Somewhere a line needs to be drawn or at some point we will realize that an AI plays the game for us.

By now my blood runs cold when I read how people use the term "strategy" to justify their need for a totally automated and preferably turn-based game with no need for speed at all and to discredit games like SC2 to be purely "tactical" games.

I be happy to draw the line where I am watching the game and feel that I can't give any meaningful input to improve the likelihood of my side winning.
I think we got a long way left of AI-development before that will become a reality with simulation based RTSes.

I'd like to see some posts of these people then.

Nothing really wrong with AI playing parts of a game or even entire game for the player, since even bad-to-mediocre player eventually beats most advanced non-cheating simulation-based RTS AI to a pulp. There are kinds of AI that can outsmart humans in such a game, but they are too slow to actually do something real-time without a supercomputer. So, one who delegate everything to AIs will definately lose.

I am a long-standing proponent of this viewpoint. However, I don't really want RTS to be turn-based (turns are horrible immersion breaker to me), just to have pace comparable to turn-based games (estimated 60 peak APM, 10 total APM empirically), since most people cannot effectively think if the pace is faster. And yes, Starcraft is nothing to do with strategy, it's a contest of performing well-known predefined macro moves with maximum reaction speed and precision on micro level.

Thanks for answering his request for me.

Just because complex strategic decisions are hard to do within seconds does not mean that the knowledge that is applied while playing Starcraft is not based on strategical thought. The core-strategies are planned out before playing. Just like a general has a strategy before he starts a battle and applies tactics while fighting the battle.

From what I understand this is actually pretty fitting to the definition of strategy as a "general long term plan for a battle" and tactic as what you do in the battle to follow the overall strategy. FA works pretty much the same way: plan out a build with it's specific goals before the game and execute it while responding to whatever the opponent dos. So yeah, Starcraft is a strategy game, just as much as FA.

So according to the common definition of RTS the strategy-part actually happens mostly before and after a game. Your definition of RTS would include the creation of strategies as part of the active ingame-experience.

I think what happens is that people associate starcraft with micro, and thus associate it with tactics.

Which largely it is. Don't get me wrong, I played SC2 for a while and would like to get back into it at some point, but there are differences that are captured but not fully explained when calling SC2 a "RTT." The strategies are there for sure, but starcraft has a lot less to do with the strategy and a lot more to do with the execution, aka, the micro and timings. Which offputs a lot of people, because for most people (re: bronze, and silver really, if I want to include myself in this) can't realistically do all the micro while maintaining a strategy.

This is one of the big attractors of the SupCom market, I think, as it isn't solely focused on micro. The mechanics are there for it, but it has a lot more to do with how you plan out your economy and base size, and how you crank out units. Or at least I think. Take that with a grain of salt, as I never played competitively for Supcom, but people are looking for more than just 'take this build, at this time, with these units, micro, go!" They want options that let them try their own strategies and see what works.

This is what I get out of the whole RTT to RTS. It's not an accurate comparison, but what it's trying to say is that the bigger picture is what a lot of people want out of games like these, which isn't as prominent in Starcraft.

Mhm, ok. I'm unsure if I agree, but that could be because I'm unsure of what exactly you want.

But I must admit, I thought Cola_Colin was referring to my previous post here.

This is a fascinating topic. I haven't voted on it yet, because I remain as-yet undecided.

I wrote a post on the Starcraft 2 forums just yesterday in which I stated that mastering finicky, repetitive tasks is a poor way to separate the wheat from the chaff. Here, I was referring to things like squinting at the minimap, automated build queues, worker management, and compensating for units that lack the ability to fire on the move. These are all issues that Planetary Annihilation happily sidesteps, which is one of the reasons I'm excited for it.

But I'm not sure that taking control of units on the battlefield constitutes a finicky, repetitive task -- it is a task that involves rapid assessment of the battlefield, concentration, and multitasking. I once read a Dawn of War 2 discussion thread in which one player suggested the following exercise for improving multitasking:

• Start a game against the AI
• Build a low-level squad
• Move it to a power node (roughly equivalent to a geothermal vent in Total Annihilation)
• Tell the squad to run in circles around the node without queuing up move orders
• Play out the game normally; endeavor to keep this squad running in circles around the power node throughout the game without queuing up move orders
• When you can successfully do this throughout a game and win, increase the number of squads running around in circles to two

This is effectively Training from Hell, and anyone who mastered it became a formidable opponent in Dawn of War 2.

Consequently, I find myself in agreement that similar ability to multitask, concentrate, and maintain attention on multiple fronts should be a hallmark of great players.

The problem is, I'm not sure how to apply this sort of thing to Planetary Annihilation. This is a game in which skirmishes could potentially be happening on dozens of planets at any given time -- and even the ability to micromanage three battles more-or-less simultaneously (using the above technique) would only yield three victories out of a dozen. Win the battle, lose the war. I feel that Planetary Annihilation is a game in which the player who can consistently macro up across multiple fronts in an efficient manner will triumph.

Furthermore, there is precedent for automated combat: a game known as Sins of a Solar Empire (which is basically a real-time combination of Homeworld, Civilization, and Warcraft III, the latter present in the way capital ships in Sins gain experience and ability ranks) demonstrated extensive automation to compensate for the fact that battles could be taking place across dozens of gravity wells simultaneously:

• All abilities could be set to autocast, and most of them had very intelligent casting conditions
• Fleets had multiple cohesion and aggression settings
• Orbital structures could be set to automatically be placed within a gravity well, which permitted players to effectively build bases at maximum zoom (the game had a strategic zoom)

None of these detracted from the game; in fact, Sins was widely praised for its UI and for its automation. Though players certainly gained an advantage in micromanaging their fleets and timing abilities during major battles, there was an extent to which battles could be ignored, allowing players to wage war on multiple fronts.

I feel that a degree of combat automation could and should be present in Planetary Annihilation; to be honest, though, I feel the most important question is, "how much?"

Recall that Total Annihilation had multiple attack and maneuver settings; perhaps units in Planetary Annihilation could benefit from other settings that might influence the manner in which they approach a battle.

I invite you to take a trip back in time to Homeworld. You could choose to have squadrons of fighters use Aggressive, Evasive, or Balanced tactics. Evasive tactics caused the fighters to make shorter attack passes, but fly more erratically, avoiding fire -- good for a holding action. Fighters using Aggressive tactics completely eschewed maneuvering and flew on straight (and often suicidal) trajectories, firing constantly. This was good for times when targets needed to die quickly. Balanced tactics were a combination of the two.

Once again: I feel the most pressing question is not whether or not combat in Planetary Annihilation should be automated, but how much it should be automated.

Perhaps just as important is the matter of implementation. Any automation present must be streamlined and easy to engage (or disengage, to permit players manual control), to avoid provoking an equally frustrating response in players: Stop helping me!

To be honest, I feel that the ability to control certain behavior settings in units would help immensely.

For example, imagine what would happen if players could pick between two movement modes, "Normal" and "Swerve." Units set to Normal movement mode go straight from Point A to Point B; units set to "Swerve" take an S-shaped path -- thereby confusing any projectile with a flight time (but taking three times as long to get anywhere): in order to properly lead the target, the projectile would need to be aimed ahead of units, but if units are taking an erratic path, then it would be very difficult to lead the targets, thus promoting a more conservative unit composition (with faster-flying but less-devastating projectiles).

Depending on the variation in projectile speeds from unit to unit, this would be an effortless anti-frustration feature (if it could be implemented easily -- I'm not sure how well it would mesh with flowfields and cost fields) that would automate a substantial portion of combat without feeling like the AI was fighting the battle for the player. It removes the finicky stuff, nothing more.

Battlefield management and Unit AI are not mutually exclusive. If anything the latter will emphasize the former.

This is not personally directed at anyone but it seems necessary to say. The difference between these two is an order of magnitute - employing feigns, retreats, formations, delay tactics, skirmishing, flanking, pitching, etc. are a strategic order higher than focus firing, dodging, strafing, kiting, using abilities, etc. We are dealing with how units engage, not how battles are won.

http://en.wikipedia.org/wiki/Military_doctrine

And while I am sure some disagree, it is my opinion that unit management is integral to the RTS genre, much as resource accumulation and unit production. The purpose of smart engagement is not to diminish unit management but to shift the focus from engagement management to battlefield management. It is probable that if you have played a modern RTS you have already experienced such automation - Return Fire, Patrol, Guard, Assist, Auto-Repair, these were recently manual commands.

http://en.wikipedia.org/wiki/Macromanagement

When I say Unit AI I am referring to how units engage, some units will make use of their agility, some will make use of their superior range - this already occurs to a degree in many games - notably StarCraft as a result of special damages. I do not think anyone wants campaigns, major operations or battles automated. In fact I am purely interested in managing campaigns, major operations and battles. The RTS genre is more than micromanaging units. And an RTS of Planetary Annihilation's scale should be as well.

In Zero-K this Unit AI takes the form of a "Fight" command, analogous to an attack command. You can play the game without using Fight, indeed there are important times when you won't. But managing your army should not be easy, I agree. I like StarCraft, but the difficulty of Planetary Annihilation should not stem from high level execution, it should stem from your opponent denying you resources, thwarting your strategies and employing strategies of their own.

++

A good examples of unit behavior i saw in Warlords Battlecry 3 and Distant Worlds, having units to be spawned with pre-set behaviour like guard, patrol, attack and fallback is a feature PA must have, clickfest phreaks should go play starcraft.

If Ubers shall implement an AI like in Distant Worlds, i will actually buy this game instead of pirating it as usual.

Another good example to look at is the Warcraft 3 map Castle Fight.

I watched a lot of Supcom replays and everywhere i see the same: players cant handle the management, units idle for minutes, get slaughtered by running into smth unexpected while player is busy elsewhere.
The only tactic that is used is "spam a mob of chaff, send it to the slaughter".

The role of human in large scale rts should be giving general directives, as "build land base there", "patrol there", and so on. All the micro should be handled by AI.

The only RTS game i play these days is the League of Legends, where each unit is controlled by a separate human, thus eliminating extremely irritating mismanagement.

In Supreme Commander there are SACU's which are basically extra phat engineers. If in PA we could have a drop-in feature, where you when getting overwhelmed by scale can invite another human to control a SACU, that would be sausome.

Another excellent idea would be to merge player and AI control, where your game gets played by AI and you can override it to improve adaptiveness.

as pointed out before without defining what smart unit AI contains in this context, a yes no poll doesn't make sense.
smart AI can have a range from smartphone to skynet.

Same problem with micro/macro tactics/strategy discussions, everyone has a different perception/expectation and therefore those discussions tend to go on and on without much progress.

my personal definition of macro shall be from no on:
If I can give my order verbally or textually in triplicate, go for lunch and have the order executed on my return, then it's macro and everything else is micro.

I would like to see more unit AI stances in PA too, but it seems Uber are against unit behaviour programming and metaphysical entities that's not directly visible in game.

To donate my $0.02

I played SC2 on the NA ladder as Protoss. I reached platinum only to realize that I would have to more than double my efforts just to reach masters, and that ratio of effort/progress was too big for my taste, so I quit.

This is what stood out to me during my experience. For a long time, 4 gate was the dominant build order in a PvP (Protoss vs Protoss) setting. As soon as the game loaded both players knew exactly what the other was going to attempt and it got to the point that some people yielded PvP matches entirely, leaving as soon as the game loaded.
I REALIZE I'M WALKING INTO BAD DESIGN, NOT JUST MICRO VS STRATEGY, BUT THESE ARE MY 2¢
What happened and what put me off the most was that the outcome of the ENTIRE game came down to who could juggle production and micro their zealots better.
Similarly, a lot of build orders relied entirely on forcefield placement on your ramp when the Zerg came with his units. If you miss, a few units get through and kill your sentries. While it's very easy to argue that this is what separated the good players from the great players, it didn't seem like strategy to me, but that's just my stance.
I'm not one to use fluff or semantics when trying to prove a point, but If I want to play some strategy, I don't think, oh boy, time to place some precision forcefields; I'd like to make decisions that computers can't.

I hold the view that games are sandboxes with arbitrary rules that we chose to follow when we decide to play these games. Starcraft just has different rules that appeal to those people.

Well i wouldnt be playing this game then, for same reasons i do not play Starcraft: i am quite clumsy. And i do not quite get what you mean by "metaphysical entities", elaborate?

Actually, I think this sums the argument up rather well, and divides the question of unit AI (to any degree, really) into two different questions:

Do we want games to be won because of micro?

Or do we want games to be won because of macro?


I say macro, and not strategy, because there should be many, many times where strategy should be based on preference. Perhaps going naval is just as viable as going air, and you have an equal chance of winning based on strategy alone. What we need to decide on what it should come down to, though, is either micro (forcefield placement, fungal placement/timing), or macro, which is base management, economy management, base placement, unit composition, maneuvers (flanking, taking enemy by surprise), early base raiding. Everything that requires some abstract thought, and not just how fast you can click, or if you time an ability correctly.

Ideally, the answer would be a gray area that falls between the two. But which way will we favor? I favor macro, because I think it opens up possibilities, where starcraft limits possibilities down to you might as well just make a spreadsheet and follow it the whole game.

Will there be some micro in base management? Yah, it's hard to escape that. But I want my battles decided on if my overarching plan was executed properly. My 2 cents, of course. Heh.

Your point is m00t, because no one forbids a player from overriding unit ai and micro them himself. Thing is, no one ever microes in SupCom, i watched tons of replays in FAF, and even highest rated players do not retreat their units, not speaking of maneuvers. I sincerelly advise you to go torrent Distant Worlds and see how unit ai changes battles. Unless one side have massive design advantage, the battles look like constant retreats back and forth, with very low number of casualties on both sides. The winner is the one who pushes others back, not annihilates everything.

In terms of League of Legends, the current SupCom tactic is a "feeder" - attack, die, repeat. And "feeder" is synonimous to "loser" in LoL, because anyone able to properly retreat thus survive inevitably wins over feeders due to stacking resource advantage. As master Yoda said in s01e01, "Long the war is; only by surviving it will you prevail".

Heck, go read history of rl wars - in WW2, Soviet Union won because it retreated its industry to Ural, and had the series of fighting retreats till the stand in 43. The all-out slaughter is gamey, stupid, and should be totally annihilated from the games.

Counterpoint:

Robots slaughtering each other to the last with plenty of explosions is actually entertaining and visually stimulating.

I don't actually disagree with you in principal, but mimicking real war should not really be the goal here. War isn't actually any fun. Instead, I would like units to act intelligently, but still remain maliciously fearless. They should act out your orders to the best of their abilities, even if it would result in their destruction. I just, y'know, don't want them to be stupid about it.

Something like unit AI states, they can't be visible in the game's world without being "triggered", a player could only recognize them through UI in the most time, so Uber think they make the game experience unintuitive and inconsistent.
But I don't think minimizing them are that important, stuffs like resource and building queues are "metaphysical entities" too, and a RTS player have to deal with them all the time, it seems they are inherent for the RTS genre (and a lot of other genres too), a little more of them won't make the game experience much different.