Target Selection

[Warlord Zsinj]

It may well be that people vote with their feet; but people are also stupid, so I don't take any solace in that fact.

And munch, isn't it all about "may the best man win"?

[AF]

not when its simply a computer game... unless your taking it too seriously

[munch]

This will be alleviated if all AI's available to the player are inbuilt into Spring. This means that useful AI's will be included, and everyone will have them, so it will never be unfair. It means that AI's that potentially remove essential gameplay skills will be placed under scrutiny before they are included in Spring. 3rd party AI's should only be accepted into Spring through this path. This is my proposed solution, in case everyone missed it the first 10 times.

This is a huge step back in development that rules out many features which havent been implemented but could be implemented using GroupAI, By putting them in the engien you make groupAI an obsolete interface....

I think you've misunderstood what he's saying. He's not talking about the interface itself, he's just talking about what's available from a player's point of view. When he says "built into Spring", he's not talking about the structure of the code, he just means it gets shipped with the standard installer and has a stamp of authorisation for use. At runtime you run some checksum or hash on the standard group AIs you are using to make sure that they're all the standard ones. You don't have to ditch the interface to achieve what they're after.

I hope this clarifies their position.

Cheers

Munch

[zwzsg]

Still a bad idea. It tought the whole point of group AI with external dll was to allow easy modding of AI without any changes to the source. If everytime a group AI maker must wait for a new spring release to test his little change to a group AI, then the AI developpment will be be uber slow.

If someone play with a custom groupAI not recognized by other players Spring, then, there should be some warning in the corner of the battleroom saying "player X use custom group AI xxx, do you want to download it from him?", and the host should be have the additionnal option of forbid all custom groupAI, all groupAI even official ones, or to allow them all.

[munch]

Sounds good to me, still I don't think you could say aGorm etc are taking an unreasonable position.

Munch

[aGorm]

Thanks munch...

It's not that we dont want people making AI's persay... well it is, but we dont want the game degenerating into a battel of who had the latest upgrade to there AI.

But this still lets some people alter the behaivor of there units in wild ways... I suspect there is a solution to this but I've no idea what it is because no-one realy feels liek compramising.

aGorm

[Gabba]

having everything toggleable sounds fun! :) micro wars with everything on hold fire etc. or strategy coolness with the AI doing every thing but building structures so you can eat some popcorn whilst deciding your overall strategy... wow you could have some serious funerage with this!

Ditto. I'm rather turned off by people who insist that a game must be played their own particular favourite way, even if that way happens to be my own favorite way too.

Personally, I think helper AIs are just fine. They'd allow players to play the parts of the game they liked the most, and delegate the parts they found boring to something else. It's not likely that the AI helpers will be better at their tasks than a human would, so I don't expect it would be a major advantage (except perhaps in that it'd free up some of the player's time for use on other tasks, letting the human focus on his strengths. This could equally well be accomplished by having several humans working together, though).

That's the most sensible opinion I have read in this thread. But I think some distinctions between different parts of gameplay are needed here:

1-"Mad-clicking": this is what I call what most people here call micromanagement. You give orders to your attack force as you go, and you try to keep building your base and prepare your next attack, and all the while you're scouting for enemy expansions and maybe fending off an enemy attack. If you like this, good for you. As can be read in this thread, not everybody likes it, and personally I don't. I do want to get better at it, because it does get handy, but it's not my definition of fun.

2-"Micro-planning": this is giving orders/directives to your units in advance, and it is also micromanagement, whether you like it or not: you're giving out the orders, not the computer, but here you're giving them in advance. You're doing this every time you give a queue of orders to a unit or a factory. TA was already one of the games that allowed you to give queues of orders most efficiently. I think this part of gameplay can be massively improved, and it's important, because Strategy lies much more here than in #1. If you can define the behavior and priority of attacks of your units for each waypoint on their attack path, you are microing. Maybe you won't have to manually set each target for your units while they are engaged in battle, but that just shows you are a good planner, and that you took the right decisions in advance. Now you're rewarded with more time for dealing with overall strategy.

3-Unit AI: Even fans of microing want some of this: when you tell a group of units to move somewhere, you expect them to resolve traffic problems by themselves. Or, maybe you want to play Dune, where you can only select and move one unit at a time? Even more basic, you expect units to react to enemy units in range, and fire at them - maybe even run after them. So if you really only swear by micromanagement, units in your game should be a bunch of sitting ducks, who don't know how to steer around obstacles, will not fire by themselves, won't track the location of their target. Maybe we should have doors to each factory that take three buttons to open to let newly built units out, so that the best "mad-clicker" really comes out on top?
OK, this was a rather ironic way of showing that we expect our units to have a somewhat intelligent behavior. What some people in this thread are asking is more intelligent default choices. Such as, when they are finished with some target and they don't have another one assigned, what will they attack? Right now it seems they attack the weakest nearby target. If that's the most sensible choice (or the less stupid), fine. But maybe it would work better if they attacked the biggest threat to them. Or something else. I hope we can discuss this without getting mad at each other.
Also, I believe that each player should be able to define his units' default behaviors. Such as, "take out enemy radars before anything else", and any other list of priorities after that. Of course you'd be able to assign different priorities to a group of units. Not only in-game, where you don't always have the time, but also from outside the game with some kind of configuration editor. This way your units' default behavior is more in line with your play style. Even better, you can define several sets of behaviors (let's call them "battle plans"), and then switch to another one in-game with a keypress to adjust to circumstances.

If we pay attention to these three elements of gameplay, then players will need to develop a variety of skills to come out on top. This seems to me much better than giving out the game completely to "mad-clickers" or "planners".

And yes, I'm all for options, so the server can favor one type or the other of gameplay, as long as the number of options doesn't make the code unmanageable. Still, we need to find a default way of playing that balances all these aspects. Otherwise, you'll never "learn the game" as you could in TA.

[aGorm]

I dont mad click... Infact i use very few clicks. The simple point of our side is this. If you can't be bothered to check what your units you sent into battel are doing and tell them to attack what you think they should, then its your own fault! You hardly have to realy do much as it is, and thats why currently its fun. But it lose its fun if you make the units realy smart, because there will be no point in you actully being there. All youd have to do is build up your base, build up some units, send them over to an enamy position and then thats it, you just go build some more, and do the same... WHAT WOULD BE THE POINT IN PLAYING?!?!
Or with just limited AI, you actully have to work to make your units do clever thing... and it will be those things that seperate you from your enamy. And teh game will be fun. The enjoyment in this game does not come from winning. It comes from what YOU do. When i have lost in a game with my mates, i dont go oh how borring i lost, we all get excited about the way someone pulled of some ammazing move in game, the way he suddenly switched his peewee attack and got to teh comander and took it down or something... Thats were the enjoyment lies... And having an AI that just makes the desisions for you removes that because the AI is not taken by such things. AI's stop those sort of things happening. You beat an AI because the AI has not enough units to beat you, you cant do somthing out of teh ordinary against an AI because it wll just calculat its way out of the situation. There is NO FUN.

What is this ment to be an AI showcase?? Or a Game that FUN to PLAY with OTHER PEOPLE.

aGorm

[Gabba]

But it lose its fun if you make the units realy smart, because there will be no point in you actully being there. All youd have to do is build up your base, build up some units, send them over to an enamy position and then thats it, you just go build some more, and do the same... WHAT WOULD BE THE POINT IN PLAYING?!?!

Sorry, I have to disagree with your prediction. Sure, what you describe wouldn't be fun, but it won't happen. Here's why. First of all, up to now AIs in any game have never been very smart. AI is very hard to code, and is the weakest part in games nowadays, don't you know?
Second, if you make the AI too "smart" (meaning it makes more choices by itself), it's more likely that it will make choices that you don't like, i.e. stupid things from a wider point of view. Third, for me, defining my units' objectives just before launching the attack is as fun as clicking on them while the attack is going on. And it feels much more like a real general. A general's greatest strength is in planning, once the action begins you lose much control, even though you can still give some orders on-the-fly.

[BeeDee]

If you make the AI too "smart" (meaning it makes more choices by itself), it's more likely that it will make choices that you don't like, i.e. stupid things from a wider point of view.

Exactly. For example, I might create a group of a dozen or so gunships and send them on a suicide mission into a base. How would an AI know whether I was:

*Trying to take out the radars so that my real strike force can approach the base undetected.

*Trying to take out that Krogoth gantry that's 95% finished constructing a Krogoth.

*Trying to destroy a fusion reactor or targetting system to cripple the assault force he's currently got pounding on my gates.

*Simply harrass him a bit so that he'll waste time building more anti-air defences in the future when my real assault is going to be from the amphibious tanks I've been secretly mustering on the ocean floor. Having them focus on attacking flakkers would be best here, even if they've got no chance of taking out enough to survive the encounter.

*etc.

If the group AI tries to be clever enough to make the decision on what to target on its own, it's almost certainly going to pick the wrong targeting priority. The default priorities that Spring currently uses would also get it wrong in most cases. True, I could keep the group selected and the moment the right target comes into LOS manually click on it to order the attack on it, but why should I have to do that? These units are supposedly being run by sentient pilots, be they Arm clones or Core patterns, why can't they be told what to aim for without requiring a floating cross-hair to be manually placed before them by a remote commander? When enemy units are closely bunched together I often click on the wrong one in the split-second between them coming into LOS and the planes passing them by, an error that is pointlessly frustrating and makes little sense.[/i]

[smokingwreckage]

***It would be nice to be able to preset sophisticated targetting priority on-the-fly but I think it might be just another barrier to learning the game, like having to bind a heap of scripts before you could achieve anything in Team Fortress for example.

***In TA there was a limited ability to manipulate target order with the +shootall command, without which the units would shoot only other units, not factories or res buildings.

***In Dawn of War a unit can be told to stand ground, stand perfectly still, attack freely (and pursue without limit) OR prefer to shoot buildings.

***I submit that an expanded set of standard stances, perhaps "prefer buildings" and/or "prefer AA" might be very useful because the unit doesnt really think and assign targets but can be preset for its specific role so that it doesn't do stupid things....

like a MERL trying to shoot peewees or Maulers switching focus to the next ground unit instead of the AA that's ripping them a new one. Not a replacement for skill but just making units more responsive.

***And a lot of what people want in the group Ai has already been done in other games in the form of stances, formations, and individual unit behaviour: Medics in Starcraft actively target wounded marines; Age of Kings allowed formations that protected weak units automatically.

Some stuff is going to be hell to implement at a stance level because of the diversity of units and the new units that pour in. Group AIs could be set up to handle that.

We should see where a few small solid ideas take the game. Sophisticated group ai's are not going to happen anytime forseeable, but certainly games should be able to designate which nonstandard scripts and ais will be allowed.

[aGorm]

In dawn of war there are simple presets for genral desion making, IE doyou run about screemin or stand and shoot. That would be fine. I dont mind that.

But tampering with unit AI so they hunt down enamy flack cannos for instance would be borring for the other players involved. You migt be grinning manicaly as you tell some units to seek and destroy them... but your opponent will be going what the hell, where are all these coming from. How come all my flack cannons are gone!?! Remeber how the crap starcraft AI would annoy you? THAT IS WHAT WILL HAPPEN.

aGorm

[Warlord Zsinj]

To the people who have recently arrived and contributed to the discussion:
Have you read this entire thread?

Because it seems like you have not. Many people are suggesting things, or discrediting things which have come up already in the discussion.

I have responded to a number of the points raised in my previous posts. This is not to say that your opinions are therefore discounted, but rather that you are simply repeating what I have already argued against. If you read what I wrote, please respond to the points I raised, rather than reiterating the original point that I attacked. I don't feel like repeating myself.

----

And for the people who are lumping the "micromanagers":
Stop it. Express your opinions, and don't stereotype mine, or anyone who agrees with my stance. And definitely don't put words into my mouth. Firstly, our opinions far too varied to be lumped together. Secondly, don't even think about simplifying my argument to "I am a fan of fast clicking", because it I am not. If you think proponents of micromanaging are simply advocating fast clicking, then you have completely missed the point of the argument.
And for anyone who tries to reduce it to "well, the computer does pathfinding for you, why don't we remove that"; your argument is a reductionist one that has no point. It wastes everyone's time. Obviously noone wants to tell their units how to tie their shoelaces. This is not the argument at hand.

Please, read my previous posts. I responded to almost all the points raised since my last post. If you still disagree with me, draw me up on the points I raised in that post, rather than forcing me to repeat myself.

[zwzsg]

If you make the AI too "smart" (meaning it makes more choices by itself), it's more likely that it will make choices that you don't like, i.e. stupid things from a wider point of view.

To me that's the criteria that should determine if a decision can be automated or not.

Everything that could result in having the player units suddenly surprise the player and go against the player will mustn't be implemented.

But note that if it goes against the player will only very rarely, and can be turned off, and doesn't have that much impact, and that the player can clearly understand why the unit decided to do so, it's okay. For instance having my units automatically fire at enemies in los can in a few rare circonstance annoys me, like when a Merl try to fire at a fast moving weasel and so waste a few seconds reload time. But then I know it was going to do that, and I can put it on hold fire if I don't want it to.

[Min3mat]

:) my units don''t have to tie their shoelaces! they use velcro!!! but yeah i love micro and i think that the game AIs shouldn't go much further (unless they are for the AI only!!!) the MM AI is teh sex though, and if someone could improve it so it rinses your energy for metal until you are down to 10% of max then lets it recover... hehe! :) it would own! besides that i think that the only other AI that should be improved are the Formation AI

[Warlord Zsinj]

See, I think that the line shouldn't just be drawn at "this AI does things you don't want".
I rather think that the AI should be considered "too much" if it does something which is not a mundane task, but is an important part of game playing skill, as I have argued with bombers selecting their own targets during a bombing mission.

[Gabba]

For me this thread sounds too much like a few players have acquired a kind of supremacy over TA multiplayer gameplay because it fits their skills, and they're afraid of losing that supremacy.

If I was you guys, I wouldn't take much pride in something the computer can do as well as me. On the other hand, if the unit or group AI can't match your micromanagement, what are you afraid of?

See, I think that the line shouldn't just be drawn at "this AI does things you don't want".
I rather think that the AI should be considered "too much" if it does something which is not a mundane task, but is an important part of game playing skill, as I have argued with bombers selecting their own targets during a bombing mission.

If you can tell your bombers in advance which kind of targets to choose, I'm all for bombers choosing their own targets.

Maybe we should rather be discussing what are the important parts of game playing skill: why would manually clicking the bombers' target be so important, except for the fact that you are good at it, and that the game being that way gives you an edge over other players?

Besides, with better unit AI, you can focus on more things at the same time, build your base better while making bigger coordinated attacks. Just be a little bit open-minded: don't you think that gameplay can evolve from TA? Personally, I've been waiting for a long time for more strategy and less tactics in real-time STRATEGY games.

If you still think I'm talking nonsense, I suggest you buy a copy of PC Gamer and read about the philosophy that Total Annihilation's creator is gonna apply for his new RTS game. That guy revolutionized RTS gaming a first time, and is gonna revolutionize it a second time, instead of getting stuck at the level of TA.

[AF]

For me this thread sounds too much like a few players have acquired a kind of supremacy over TA multiplayer gameplay because it fits their skills, and they're afraid of losing that supremacy.

Gabba I see your portrayal of old guarde versus reformists and I salute you, and I second it, it's everywhere, from which features should be implemented, who wins arguements, what the arguements are about and why something should or shouldnt be done, it evend efines the way the engine moves forward and what units and maps look like and their design. Most of the old guard are TA Purists or people who believe they are pillars of the TA community, people who are seen as in high places within the community before spring was in existence, or closed minded individuals. There are also a large majority who are afraid of change who follow them.

micromanagement versus tactical management

There is micro and tactical. Everyone has bungled them together and labelled them as micro. Micromanagement is what warlord zsinj calls mundane tasks. Tactical management is what people are trying to defend, those clicks that can eb automated but have tactical or strategic meaning, and arent there to correct an annoyance or repetitive task, such as targetting specific types of units.

I say get rid of micromanagement, who wants to set waypoints to make sure a ship gets through a odd shaped pass, or untis dont get stuck, or metal makers dont use all your energy.

I say encourage tactical management. AI is unable to use tactical management effectively wihtout being hardcoded to do so with heuristics or weights and limits. It requires context in an abstract form incompatible with the AI and is the ultimate reason why an army of 5 cand efeat an army of 20 of equal unit stats.

I think you've misunderstood what he's saying. He's not talking about the interface itself, he's just talking about what's available from a player's point of view. When he says "built into Spring", he's not talking about the structure of the code, he just means it gets shipped with the standard installer and has a stamp of authorisation for use. At runtime you run some checksum or hash on the standard group AIs you are using to make sure that they're all the standard ones. You don't have to ditch the interface to achieve what they're after.

You thought wrong, I myself suggested that many pages ago and nobody noticed. Take note fo zwzsg, he understands perfectly and nobody is listening save those labelled as the reformists. This si most definately a step back and ym suggestion as further spelled out by zwzsg and ellaborated on is somehtign you should listen to, it is the best solution yet provided.

I dont mad click... Infact i use very few clicks. The simple point of our side is this. If you can't be bothered to check what your units you sent into battel are doing and tell them to attack what you think they should, then its your own fault! You hardly have to realy do much as it is, and thats why currently its fun. But it lose its fun if you make the units realy smart, because there will be no point in you actully being there. All youd have to do is build up your base, build up some units, send them over to an enamy position and then thats it, you just go build some more, and do the same... WHAT WOULD BE THE POINT IN PLAYING?!?!

scaremongerer.

What you speak of is impossible. I give you 5 years and whatever resources you need and you will nto be able to create a groupAI capable of generating that scenario each time without the human being powerless to act.

You beat an AI because the AI has not enough units to beat you, you cant do somthing out of teh ordinary against an AI because it wll just calculat its way out of the situation. There is NO FUN.

Calculation usually means something like a gametree. Chess uses this, and an RTS is a heck lot more complex than chess, especially TA which is one of the most varied and strategically viable games in existence. To be able to sue strategy the computer must be able to read the entire game tree which requires an infinite amount of system resources and time. It si a geometric curve. Perhaps fi you programmed the groupAI
to repsond to every knwon strategem in existence, it would still be crippled by a simple strategem that it had never encountered, not to mention the lack fo system resources to run such a groupAI.

Also such a groupAI would also need GlobalAI interfaces for the necessary information, leaving th eplayer unable to control, making ti a skirmish game and a skirmish AI.

If I was you guys, I wouldn't take much pride in something the computer can do as well as me. On the other hand, if the unit or group AI can't match your micromanagement, what are you afraid of?

That will not make a difference, that is a style of arguement that fits not with their pespective and thus has no relevance to them. However it has a meaning that is very profound to what you put forth.

Warlord and aGorm, you both fear something happening that is unfeasible when implemented in the way you describe. Computers today and the state of research means AI is incapable of reaching those proportions.

However warlord you point out again that it neednt necessarily eb as aGorm portrays, and that certain automated but not mundane tasks can tip the balance. For this I think you have argued valiantly however your arguements are futile as zwzsg has pointed out, and there has been a solution to this since the 3rd page at least that works perfectly if implemented.

[Min3mat]

WOW! what a biatch of a post alantai! but u r right. AI has imperfect information and makes imperfect decisions based on this imperfect data, unless you get the AI to guess what the radar dots are (YES i am talking about a non-cheating AI) and i think that however good the AI at a PC game a PRO will ALWAYS be able to beat it. :p i love micro though! :) although i disagree with your tactical clicks/micro clicks. EVERY click is technically micro as strategy and macro is deciding what to build, micro is clicking fast to get that shit done and scouring the map for idle builders (lazy bastards :p)

[aGorm]

Ok, look i'll state my boudries...

I'm fine (well, I would not hjave reson to complane) with the following

Being able to tell units befor you attact to priorites AA buildings over other building. Them not shooting somthing thats pointless to shoot at (like non-attacking buildings) unless asked. That construction units dont just sit around waiting but reclaim nearby stuff.
And all the ones that are in already.

Now the things i would not be happy with.

Telling your units to go and seek and destroy the enamy AA where evere it is. Ignoring enamy non-attacking buildings to the point of activly looking for somthing else to shoot. A constrution unit that goes and builds a defensive line in it spare time.

Both of thoes could be made to happen.
The first one helps people play a game without there units being stupid.
The second one plays the game for the player with teh units being to smart.

Notice how they arnt actully that far off each other, ones like a more extrem example.

The second one just removes any type of skills from the game. I mean... if you escaped with 1 construction unit and the enamy did not notice, the AI should not go activly seeking what the player does not know is there. But if you had an AI that was doin presicly that then that poor player that thought he got away would not get away atall. In the first example also as the opponent did not notice his enamy slip away hed get away with it and THAT is what makes games fun.

aGorm