Question

[rich12545]

Wondering if you'll be removing the ai cheat that allows them to see all units in the game.

[Lt Belenko]

Why should it be removed? Not meaning to be confrontational, but seriously can you explain :confused:

So what, if AI sees all know all? To make the game fun/challenging it's all in the settings... Adjust the game for more challenge to keep it fun.

If you are beating 5 elites on one team on a small/medium map with minimal supplies available then you have a point. :confused:

[rich12545]

Why? Because it's ridiculous. I'm not trying to be confrontational either but playing a wargame against a know all AI isn't right and takes away a lot of the fun, for me at least. So if a lot of people want to keep it the way it is then maybe make it an option. We should be able to play the game against a decent AI without it having any, or maybe a few optional minor, advantages.

If I want to try and sneak a sub past a couple of ships to get the transport, it's not even conceivable to try. And there are so many other examples.

I've really been playing the game off and on since I got it, waiting for the AI to improve. Now that it's finally getting better I've been looking forward to spending a lot of time with EoS. But not with the AI this way.

[StellarRat]

I back Rich on this. What is the point of having subs or other units that are supposed to be invisible like SEALS if the AI knows they are there? Right now, against the AI you might as well use surface ships and regular infantry. Cranking up the defense for hidden units to simulate partial invisibility leads to all kinds unrealistic results like a sub destroying an entire enemy fleet or some Green Berets killing an entire division of tanks. It makes spy or scout type units impossible to simulate. I think this could fairly easily programmed by only letting the AI know the last actual sighting or attack of subs (or SF) then having it move there with some random offset to allow it to be unpredictable. If it actually gets within real sight range it could move to attack and call in any other units that have movement remaining to attack. The hidden type units could be "flagged" as hidden and only those units would have this special procedure applied to them.

[rich12545]

The AI should, essentially, play by the same rules as the human player. There could optionally be slight, maybe by percentage, production and/or combat handicaps to the AI to make up for the human advantage in thinking ability.

[Brit]

Wondering if you'll be removing the ai cheat that allows them to see all units in the game.

Yes, I want to. It'll take some more code to allow the AI to deal with the uncertainty, plus the AI might play a little worse, but I'd like to remove it. Having the AI play without knowing the placement and strength of the enemies adds more strategic possibilities (like more interesting submarine warfare, increases the need to remain unseen before an invasion, etc).

[Eidolad]

How about tweaks to that 100% AI radar sight...limit the AI sight range to that AI empire's current tech jet fighter patrol range (and place a colored "border" show the human player how far the AI sees)? Or limit the AI to "radar flashes" every few turns...and let the player sometimes know when these flashes occur? Except on elite level where the grognards want to play where it is rough, I suppose. And slowly wean Mr. AI away from 100% radar:

The new AI will have to take to the skies and have effective air patrolling to "see". I've never seen a computer wargame where the AI has effective patrolling. There would seem to be a great deal of coding.

a) if AI really doesn't know where to put his patrols...he would have to maintain a very expensive air force just for patrolling.

Single planes that are easy meat for even basic ground forces..so the air patrols will have to avoid flying over enemy and neutral cities

or: use spy planes primarily...hard to kill and low priority for enemies otherwise?

c) When a patrol plane gets damaged...where to go to repair? Currently the AI doesn't seem to build many airforce bases for repairing.

d) Then there's the coding for the "response to enemy contact". The AI, upon spotting an enemy force of transports...needs to vector in, and launch, a successful air attack to destroy *just* the priority threat Or send a naval/ground group into the area to engage them...but if the player has a screening force..that AI "response" will just go off and either waste themselves attacking...or go do something else. I.e. once a group of player transports is located, it is "tagged" as a primary bad guy by the AI. If the AI "response" fails to find that "tagged" group, it should stay there and begin patrolling?...make the enemy move somewhere else.

e) could I argue that a patrolling fighter plane is more valuable than some other ground unit to defend? Yes, most especially for an AI island city. Even a single fighter has a chance to sink a transport...

edit: a poor example of AI "response" is in Civ4...where a group of air units will be moved by the AI into a threatened city...oops they seem to not notice my stack of ground units who take that city in the very next turn...

[Lt Belenko]

Why? Because it's ridiculous. I'm not trying to be confrontational either but playing a wargame against a know all AI isn't right and takes away a lot of the fun, for me at least. So if a lot of people want to keep it the way it is then maybe make it an option. We should be able to play the game against a decent AI without it having any, or maybe a few optional minor, advantages.

If I want to try and sneak a sub past a couple of ships to get the transport, it's not even conceivable to try. And there are so many other examples.

I've really been playing the game off and on since I got it, waiting for the AI to improve. Now that it's finally getting better I've been looking forward to spending a lot of time with EoS. But not with the AI this way.

OK. I see your point. I was thinking over all game play not an individual unit or two. I've dropped more than a few paras at a distance from anything hoping to make several drops before moving on the enemy but they come under immediate attack.

[rich12545]

Yeah, I wonder how long it'll be before Brit will be able to do it. Looks like there might be quite a bit to do first.

[jeep]

I agree this would be nice to have, but I have to say that I've never seen a wargame AI that really does this. Humans deal with uncertainty with creative, changing solutions. The AI would need to do so with a fixed formula. Once you understand the formula, you can trick the AI nearly every time.

To be blunt, humans ALWAYS have an advantage over the AI in a game like this. It isn't really fair to expect EOS to have the most advanced AI in the world (though it would be cool).

Given that the AI always can see everything (units & map) for planning purposes (visual range still applies to attacks of course), subs and paras still are useful for the human player. Para are fast to deploy, and subs are great for sinking the BBs. You just have to get used to the idea you can't sneak around the AI.

AI improvements, at least right now, are better spent on better land and sea tactics. I definitely see improvements in the last update. Once the elite AIs get to the level that they can stand a good chance of beating human players in a only slightly stacked fight, then we can worry about its omniscience.

Making an AI that can be tricked is usually something you DON'T want to do. Once it is easy to exploit, you won't be having much fun....

[rich12545]

I don't agree. Basic in a wargame is to have the ai start equal with the human, by that I mean using the same rules. This is wargame 101. Then you can have options to make the ai easier or harder with little things such as an advantage to production and/or combat. For example you can make production + or - 5%, 10% etc. You want to adjust it so the human wins most of the time but has to work to achieve it. That was the working theory behind the original Empire Deluxe and it still is true. That makes for a fun game and, after all, that's the object of us playing games.

One thing you don't want to do is allow the AI to see all the units. It's just way too frustrating. Maybe that works for you, but for me it's a game breaker. In fact I'm going to play other games until this is corrected and that's the reason I asked about how long it will take.

[StellarRat]

Actually the AI could easily respond to invisible units without cheating. Probably better than a human handles the situation. I've written software that does this. Computers have advantages over humans in many areas. You write the AI to use those and they make up for some of what the computer is poor at.

[jeep]

Just out of curiosity, what advantages does an AI have over humans in a game like EOS?

I understand in games like chess etc it can run many more scenarios to predict outcomes than a human could. But that kind of thing doesn't really apply here.

I do agree a good compromise solution would be something like excluding "invisible" units like subs from the AI's planning. AI would only respond to them while they are sighted. This avoids the larger problem while still giving you the function you wanted.

You will still need to guard those transports, but you should be doing that anyway

[rich12545]

The thing with a game like EoS with random maps is that it's virtually impossible to script unit's behavior. So the best thing is to give units general orders that can be used on every map. Expand, attack, retreat when overpowered, move to repair, escort, etc. The ai won't do all this as well as a human of course and that's why you should be able to give it some advantages as outlined above.

[StellarRat]

Just out of curiosity, what advantages does an AI have over humans in a game like EOS?

There are many, but I'll name a few.

1. It has a perfect memory. This is useful in locating and remembering maps and unit locations. This is also one of main reasons I'm saying that it actually has an advantage in hunting hidden units. And it's not going to forget to check something or fail to issue important orders like a human could do.

2. It will never do something irrational out of spite or because it's had a couple beers.

3. It won't screw up and move or attack the wrong unit or wrong place because of a slip on the mouse or a wrong key press.

4. It can calculate the odds instantly before it attacks based on whatever it knows and decide if it's a good situation. Provided it has a good formula for this it won't take unnecessary risks.

5. It doesn't get distracted from it's objectives. This is one the principles of strategy. Of course, this only helps if it has good code to determine what it's objectives should be.

If programmed by an expert an AI can really do a pretty good job against a human. Actually computer game "AIs" are actually more accurately called expert systems because they don't learn. A true AI is a learning system. The ones in games generally follow a formula that a good human player would use to win, but they don't "learn" new tactics to defeat an enemy.

[hoolaman]

It might be possible to have the AI cheat strategically, but not cheat tactically.

For example, it may launch air or sea patrols when it "knows" there are enemy units nearby, but have it send them in random directions or towards a spotted enemy city instead of the laser precision it has now.

StellarRat's idea sounds on the right track, but I imagine since some units can see subs a more universal approach would be required.

I think this could fairly easily programmed by only letting the AI know the last actual sighting or attack of subs (or SF) then having it move there with some random offset to allow it to be unpredictable. If it actually gets within real sight range it could move to attack and call in any other units that have movement remaining to attack. The hidden type units could be "flagged" as hidden and only those units would have this special procedure applied to them.

[StellarRat]

StellarRat's idea sounds on the right track, but I imagine since some units can see subs a more universal approach would be required.

My system does handle units that can see subs. In fact thats all I wanted it to handle. Maybe you didn't read it carefully enough. Even units that can see subs have very limited range that they can see them in. I really don't care too much if the computer uses its current system to spot units other than subs and other units designated as "invisible". For units other than invisible types it doesn't affect game play nearly as much, but for subs and special forces it pretty much invalidates their reason to exist.

[rich12545]

I still think the computer should follow whatever's in the ruleset. Period. However the human plays, the computer should play.

[StellarRat]

I still think the computer should follow whatever's in the ruleset. Period. However the human plays, the computer should play.

It is possible, but its a huge hunk of work to make one that plays decently. I did a space war game with no terrain (obviously) or strategic components (simulated individual battles only) it took two of us working about 12 hours a week on it for a year to do just the AI. It didn't cheat and could beat humans including us about half the time.

I'd hate to think how long it would take to do the same for EOS when there are so many other elements besides tactical combat to consider.

[rich12545]

You could be right. I don't expect miracles. Just a decent ai playing by the same rules. If that can't be done, I'll play something else. It would be a shame because I like this game and it has a lot of potential, but it's not the only game out there. I've been waiting a year for it to show its potential but now I'm not so sure it will.

[StellarRat]

I doubt you are going find one that doesn't cheat in some way. In all likelyhood you just won't know it. Brit is doing an awesome job supporting this game. I'm sure he will take care of any blatant AI cheats. Most other developers probably wouldn't even discuss how the AI "works" with their customers because of the huge amount of cheating most of them do. They really don't care about post release improvements other than fixing bugs that cause crashes. The AIs are lousy because main focus in most games these days is fancy graphics and sound to impress the teenyboppers. Even unit and play balance and actual game system take second place in most games. If you look at ratio of programmers to artists and sound engineers on most modern games you'll see its over 7 to 1 in favor of the non-programming jobs. That should tell you something.

[rich12545]

I don't buy "most games." I typically buy games from companies like Battlefront and Matrix plus a few other like companies. These are war/strategy games. So we can't compare to those. I'm comparing to, for example, Empire Deluxe Enhanced. EoS is by far the better game if you look at features. However, EDEE doesn't cheat at all. The ai follows the same rules the player does. The game was unplayable out of the box because the ai was so dumb. However there was a script that could be modified. And the unitdb could be modified. I did and made a mod that made the game very playable. The ai does a good job without cheating. If it's too easy for you then modify production rates and combat advantage. That's easy. However, EoS is a much better game due to its features and that's why I'm here instead of there. What needs to be done is make the ai better. I've been harping on this for months. Frankly I think a lot of time was wasted working on fluff. But now it looks like Brit is working on the ai in earnest and a lot is getting done so I'm very hopeful. I'm probably going to continue playing simple games until the ai is working properly and put up with it being "all-seeing." I've asked how long this is going to take but Brit hasn't answered that. If it's going to be months I'll do something else and come back. If a couple of weeks I'll just continue playing. It would be nice to know though.

[StellarRat]

Isn't modifying the production rates to make the AI more challenging the same as giving it a cheat? It's not using the same rules as the human at that point.

[rich12545]

Yes it is a cheat but a very minor one. It's more like a balancing cheat. Since everybody's ability is a bit different it's impossible to make an ai to suit all gamers, most games have this kind of thing including EoS already with novice, veteran and elite. I don't really know the difference between those three. I think it was posted once maybe last year but it would be nice to see it again.

The two guys who made the original Empire Duluxe said their goal was to make the game challenging but expected the player would win most of the time. I think that's a good axiom for any game.

[StellarRat]

I guess my take is that if AI isn't ripping me off and realistic tactics work against it I'm OK with it for now. I could see most land based units, planes and surface ships being easy to find by spies and civilians reporting. It's pretty hard to hide an armored division. I can justify it in my head. However, the fact that subs are hunted down like dogs by the AI because it knows where they are at all times galls me to no end. Subs have always been hit and run weapons and that just doesn't work right now nor do my SF units that are always seen by the AI and blown up by artillery when they are hiding the middle of a mountain range with no other enemies around.