Actually, I meant "slightly incompetent" but I didn''t think that''d fit... Games like Civilization and Tropico have AI managers that, in effect, let players only worry about the details they care about. But these managers are usually not as efficient as a human could be. My question: Should AI managers be just as competent as the player, or should they be more inefficient? If they''re as good as, or even better than the player, the player will only micromanage if they find micromanagement itself fun. I don''t think this is many people, since you typically want a reward when you do something in a game. The activity in and of itself usually isn''t enough. On the flipside, though, keen AI micromanagers will let players pick their area of focus. If the player doesn''t want to worry about resources in an RTS, for example, or haggling in a trading game, the player can rest easy knowing the AI will take care of things. Inefficient AI managers will give the player a reason to manage. But players who don''t micromange may feel cheated. This would especially be a problem in a multiplayer game where one player micros and the other doesn''t. I don''t really see a solution to this... -------------------- Just waiting for the mothership...

An AI micromanager should be less efficient, in some way, than a player. If the micromanagment is just as efficient, or more efficient, there''s no point to managing that portion of the game. If the AI does it just as well or better, then the AI should just have control all the time.

Not to say that the AI micromangement should not be competent. A person should be able to turn the micromanagement on for things that they don''t really like. However, a good player should be able to beat the AI at it.

AI managers cannot be as competent as the player unless the game is badly designed and allows too little choices.

Pax Imperia : need I say more?

quote: My question: Should AI managers be just as competent as the player, or should they be more inefficient?

First, just how are you going to determain how competent the player is to model the AI ? Different person could have a vast difference in level of skills. So, even if the AI is set to be more inefficient than player A, player B could find it more efficient and to rely on the AI instead.

I dunno.. but overall in simulation games the AI should be slower than human and in RTS AI should be the opposite so player can micro as you said. Then what happens is you combine strategy and simulation ? maybe an AI that the player can set or influence in someway on how "smart" it is ?

Ouch. Nasty subject.

The problem is, micromanagement is something you either love or hate. If you have a manager which is only semi competent, then the players who dislike micromanagement will always be at a disadvantage to those who do like it. This is particularly true in turn based games.

In real time games, there is a trade off between time spent managing your units, and time spent managing everything else. The micromanaging nut may have a more efficient infrastructure, but the other player has more time to spend controlling his military units (or whatever) and organising his strategy. Everything boils down to whether these two approaches are balanced.

I''m convinced that the AI design should be designed in such a fashion that in THEORY the player could walk away from the computer for an hour, leave the game running, come back an hour later and resume his game.

In THEORY.

The AI will never be able to learn all the nuances that a player will be able to learn. Ever. They will never be able to respond just as smart as the player would, design tactics/strategies as good. Then again... if the player makes bad decisions etc, the AI WILL be able to outsmart that player.

I''ve been playing a demo game of Age Of Empires II for the last few days. I''m already thinking of how if I would design the game, I''d give the player the opportunity to design a pre-AI setting for his units. Simply because each game I play, I start out by issuing the same orders anyway. Why not just give me the opportunity to create a plan that will lead the units through these same orders, but without me having to focus on it during gameplay.

Sure, it would dramatically change this game, but with a system like this in place, you''d be able to change focus onto other topics. No longer is it just a race (as most RTS games still seem to be) of who can click the fastest, knows the best ''trees'' etc (I checked out some Age Of Empires II strategy sites, and all they talk about is how fast they can get to a certain point ''I can get Castle Age in 10 minutes. Here''s how.'' And then they describe in detail how they set each villager to a certain resource and when to create what etc). Instead, the player might actually have some breathing space to actually WATCH what''s going on and make tactical decisions. Right now, the way AOE2 plays is...
1) Follow procedure 1 to get your village built up (depending on your overall strategy, you''ll have a different procedure. For example, if you want to attack fast, you''ll build a barracks early on. If you want to attack late, you''ll build up your economy first)
2) Once you are attacked, you focus on the attack (but you still have to keep all your villagers going. They''re smart enough to do some things on their own, but still require a lot of micro-management... not a good thing during a fight)
3) Once the attack is over, focus on village again

I really didn''t feel like I got the time to make any smart decisions about what to attack, how to plan my attacks, etc. I guess that''s just not what the game is all about, but I think that that''s what it SHOULD be about. My opinion.

Anyway, I feel that the AI should be as smart as it can get. If the player wants to kick back and let the computer make all the decisions... fine.

But I think for an AI to be really smart, you have to give the player the ability to fine-tune the AI. You should let the player decide before the game even starts what the AI should focus on.

Taking Tropico as an example (a demo game I had a lot of fun playing, but can''t see myself buying the game, simply because I thought it was too limited... too much micro management for my pleasure)... You should be able to set the AI to follow a certain path.
1) Build houses for all villagers. Start with Tenements, then build Apartment complexes.
2) In the mean time create farms for the villagers to work on. Find an available location nearby town and determine what the best crop to grow would be depending on soil/weather etc.
3) Keep building houses for villagers until each villager has a place to live.
4) Build more farms
5) Build church
6) Build high school
7) Build medical center
etc

I guess as a player, this way, you can take care of the micro-management BEFORE gameplay starts. Because, don''t you usually follow the same procedure EVERY time you play the game? Don''t you establish what you think is the perfect plan? Well, if you let the AI take care of it, you won''t get bored of it, as I do.

The player should be the one in control of the AI though. While the AI takes care of the plan that the player has laid out for it, the player watches what is going on (he has to look and see what the ''random'' factor does with the actions of the AI). While the AI is building houses and farms, the player can check the statistics of individual citizens and determine what needs to be done. While the AI is building farms, the player might find out that the citizens want better quality jobs. The player can now tell the AI ''hey, stop building farms and concentrate on creating a higher grade economy. Build some factories.''

Tropico was actually pretty limited in my opinion, since it lacks the ''battle'' aspect (well, the citizens/soldiers can create an uprise, but there''s no ''invading forces'' to battle, as there are in Age Of Empires and Civilization). A battle aspect would completely shift the focus of the game though, so I guess that''s just not what they wanted.

But, if you DO include a battle aspect, it becomes even more important to take the micro-management out of the players hands, so he can focus instead on the bigger picture. If you as the player, still have to tell each individual villager what to do, you''re not going to have time to really come up with a good battle plan against your opponent''s attacks. You''ll have a group of battle units that you''ll just send towards the enemy, and you''ll hope for the best. That works for a peon-pumping RTS, but it doesn''t work to satisfy the ''S'' in my RTS needs.

Create an AI able to take care of the game on its own. Let the player guide the AI and take care of the big decisions. BUT, give the player the ABILITY to micro-manage as much as he wants (since a smart player will probably forever be smarter than even the smartest AI, taking micro-management into your own hands WHEN YOU CAN can give you an advantage over your opponents).

I''d rather look at the big picture, but when there''s nothing going on in the big picture, I''d like to focus a little bit on the small picture and see if I can make any minor improvements. The AI should allow me to do so and should go along with whatever I am doing.

AI AS SMART AS CAN BE

Ahh, the wonderful art of delegation

I think it depends on the size of the game and the amount of micromanagement required. For instance manual MM in Civ 2 is fine when you''ve got 4-5 cities but when you have 50 spread all over the world (most of which you can''t even remember the names or locations of) then MM becomes VERY tedious. And if Civ 2 was real time then it would just be impossible.

I think the AI needs to be as good as possible otherwise there''s no point in having it, the player wont want to trust it when they know they could do a much better job.

But then again it opens up possibilities for added depth to the game in that you could hire managers/generals/governors (depending on the context of the game) that you could use for micro management. Some would be very good and some would be totally useless. If your not happy with your current governor fire them and hire a new one. This could be extended to interviewing prospective employies and even further depending on how far you wanted to go.

This could also add balance to multiplayer games in that better auto MM agents would cost more to hire. The players that chose to MM themselves would have more resources to play with but would have a lot more work to do will those that chose to use auto MM would have more time for planning strategies but less resources.

- Kaijin

"If you find a job that you love you''ll never have to work again."

Nah... I think they can be perfectly competent. For people who like micromanaging, such as myself, the benefits are in giving yourself more direct control over all aspects of your ''machine'', which gives you a wider range of outputs. You don''t have to be ''better'' than the AI, you just need to have a wider scope to work with.

If that''s too vague, consider this: the AI manager could perhaps manage your civilization perfectly under ''reasonable'' situations. It would work towards the best balance between numerous aspects, also bearing in mind any setting you chose (such as Warlike, Peaceful, Economic or whatever) that influences its decisions. But there may come a turn when you want something radical: such as asking all the farmers to make guns for a turn. With micromanagement, you get this option: the option to take action that might seem ''bad'' on purely numerical grounds, but ''good'' to the human mind that can evaluate the merits of such a gamble.

ai-manager is average-competent. for newbies users ai-manager is better, but for advanced users ai-manager is very bad.

Of course ai-manager must be some "Pavlov'' dog", in fact, if he made a good work, then you must give a bonus ($$$), if he made a bullshix work, then you must give a kick-ass. In fact there not are ai-manager, is more that ai-assistant (or ai-juniors).




-eng3d.softhome.net-

I think this can refer to AI command as well. When you think about human processes of management, the head personin charge really just gives directives, and it''s up to the subordinates to follow through on the directives. Replace subordinates with AI manager, and head person with player, and this is the situtation game designers (and players) must deal with.

I think the trick is to be able to give directives (i.e. "orders") well. Patton once remarked, "don''t give great orders, give orders that can be understood", is a saying that applies very well here. Give orders and directives to the AI that it can handle.

How to do this? Give lots of variables for the "leader" to set, and give certain modifiers that can affect those variables. For example, you could send a directive that you want a priority on developing troop transports and building training centers to increase troop quality. And you could set modifiers in a sort of if, then, else scenario, or possibly a "switch" style condition. Again for example, you could say, "Focus on developing defensive technologies and equipment until 2/3rd''s of all units are based on ''X'', then start focusing on building offensive capabilities".

Now the trick here is that it does require some micromanagement at the beginning, but then it''s out of the player''s hair, so in a way it''s a compromise between the two. The AI has to be smart enough to take certain variables into equation, and have the cognizance to recognize certain "triggers".