I just want a reasonably intelligent AI |
What the game really needs is an AI that is able to deal with human players, one that notices it is coming up against Missiles so builds ships with PD. Or that recognises that a warp point a) needs defending and b) that if it is defended by hostiles it should send a fleet to clear it, rather than sending colony ships and attack ships through it piece meal.
We also need an AI that is able to reasonably work out what are the best facilities to put on planets, I have seen the AI build 20 mineral miners on a planet with 4% minerals, so its actually running at negative production due to maintainence costs.... Surely its possible to put in a line that says if resouce
Some AI Stuffs
The sad thing is that most of your complaints are easy to fix. Point-Defense should be a standard addition on the smaller ships as well as the larger ships.
The stock AI does monitor their warp point locations, but sometimes they don't have enough fleets to cover them all and they patrol between a few of them. But they are marked as defensive locations in the scripts. It also keeps track of where enemy forces are - but I think the calculations used to determine the strength of the enemy force might be an overestimate and the condition to attack won't always be met Do you notice if this also happens when you're at war?
-----
A few of us are working on
A few of us are working on it 
. SEIV AI's were evolved over a period of a few years, each person adding more and more to make a worthy opponent. With the options we now have, hold on to your hats 
The ability to know when a human is using a certain type of weapon is possible to code but a little tricky. If human is using ships with greater than x number of missiles then build this into the design else build this..etc...
The AI already has the ability to assess a target, if it thinks its strength is greater than the target system it will send ships to attack. What it doesn't do is predict what its going to find on the way to its 'build up' point, that means you can ambush ships on their way there.
I have just altered the AI to wait for it to assess itself to have 50% more strength than its opponent, rather than 1%. Because if your opponent is a player, their mostly likely going to be able to use a near equal strength to beat an AI, no matter how good it is. This is untested as to effectiveness but I hope it works, I do not yet know what parameters the AI uses to 'assess' its opponents
Personally I think all 'build up' points should be set very far away from their targets. The reason being is that ships will be less likely to be engaged on their way there. However this could have the effect of the AI moving ships actually backwards rather than waiting them in place, if they are built on the front line, so to speak. This might be coded around but I am not yet sure.
Something like:
If the enemy is within X systems of our systems, the build up point is close by but again there are so many variables you are talking about, this simple statement will cause other possible AI confusions, which in turn need to be coded out... I added this comment as it shows some of the complexities of fiddling with the AI.
As far as facilities I haven't looked at that yet but I am about to alter the rate certain ones are produced (research being key), thats my next thing to look at for my AI in fact.
---
Designing an advanced, diplomatic AI + shipset
Consult the great SEV Wiki
I know what some of the
I know what some of the problems are and that what we really need is a sentient intelligence to make a decent AI, and that the dream of a lot more than just gamers.
For some reason half my post seems to have gotten lost *GRRR*
I was wondering why there is not a line that says if resource < 50% (for example) then DO NOT build that resource extraction facility.
There is also the problem of races that build hundreds of carriers but don't bother with any fighters to use with them, and even worse the AIs that build fighters (and drones) without any weapons.
Also it should be fairly simple to put in a line that says every X turns auto upgrade facilites, as players we can do this manualy so it should be easy enough for the AI to do this. You could make it a bit more subjective by adding lines about are you at war if so don't upgrade planets with ship yards.
Then there is the shocking lack of defences that the AI uses, I have yet to come across an AI planet with anymore than 2 or 3 weapons plaforms, and none that have troops stationed in defence.
Levels not Time
cordas,
It would be better to have the AI look at tech levels for facility upgrades. Basing it upon time will cause it to make too many upgrades or try to upgrade when there's nothing to upgrade. I base my upgrades upon levels of advancement through a facility tech. Early in the game when I've got low research I upgrade after maybe 3 or 4 level advances of some tech before doing the upgrade routine. Later I raise it to 5 levels and then to 10 before I'll pop the upgrade. Plus when I decide to upgrade some type of facility I concentrate my research on that type until I've advanced the number of levels I want to before doing the upgrade routine as upgrading is expensive and time-consuming when you've got large planets chocked full of goodies to upgrade.
Not upgrading planets with shipyards isn't very smart either since most planets will or should have them as well as resupply depots regardless if you're at war or not. One big problem is that the AI's don't upgrade shipyards and they should.
I still don't understand your fetish for ground units thinking they will improve the AI's. Ground units are worthless when you bring a fleet of tough attack ships along and just exterminte the vermin infesting the planet.
Omnius
cordas For some reason the
cordas
For some reason the greater than signs (Great than or equals too espeically) tend to cause the forum posts to be cut off or do weird things. I know having posted code and having it disappear, probably some of which looks really odd even now where I have missed it...
Yes and no, simple is a relative word . At the moment the AI upgrades ship engines by a code like this
"Small Engine":
set comp_id := Sys_Get_Component_With_Name(sys_long_Player_ID, "Small Quantum Engine")
if (Sys_Is_Component_Available(sys_long_Player_ID, comp_id)) then
set final_comp_id := comp_id
endif
if (final_comp_id = 0) then
set comp_id := Sys_Get_Component_With_Name(sys_long_Player_ID, "Small Jacketed - Photon Engine")
if (Sys_Is_Component_Available(sys_long_Player_ID, comp_id)) then
set final_comp_id := comp_id
endif
endif
if (final_comp_id = 0) then
set comp_id := Sys_Get_Component_With_Name(sys_long_Player_ID, "Small Contra - Terrene Engine")
if (Sys_Is_Component_Available(sys_long_Player_ID, comp_id)) then
set final_comp_id := comp_id
endif
endif
if (final_comp_id = 0) then
set comp_id := Sys_Get_Component_With_Name(sys_long_Player_ID, "Small Ion Engine")
if (Sys_Is_Component_Available(sys_long_Player_ID, comp_id)) then
set final_comp_id := comp_id
endif
endif
etc,etc.
It acts like a chain. It has an upgrade like this for EVERY ship part, much like their would be for every facility. Its perfectly possible to do this bit however.
Then you'd need to alter this bit, I think, its all guess work at the moment as i've only glanced at it:
// Add facilities based on colony type.
case (gbl_colony_type)
AI_STR_COLONY_TYPE_HOMEWORLD:
call Add_Facility_Construction_Percentage("Resource Storage - Minerals", 5, 100)
call Add_Facility_Construction_Percentage("Resource Gen Modifier - Minerals", 5, 1)
call Add_Facility_Construction_Percentage("Storage Space Amount", 5, 3)
etc, etc.
Then you'd need to alter the item lists, then you'd need a code to tell the AI whether to upgrade old ones or build a new facility. Who know's maybe thats already in there, I haven't looked at it yet but it needs modifying if it is.
You get the idea, possible but there is a bit of work to it.
---
Designing an advanced, diplomatic AI + shipset
Consult the great SEV Wiki
cordas wrote:What the game
We also need an AI that is able to reasonably work out what are the best facilities to put on planets, I have seen the AI build 20 mineral miners on a planet with 4% minerals, so its actually running at negative production due to maintainence costs.... Surely its possible to put in a line that says if resouce < 50% then don't build resource extractor.
Another thing the AI needs to do is be able to understand it needs to upgrade its facilites, if only so I don't have to when I conquor the planet .
Errr this is starting to turn into a bit of a rant I know, but whilst I am on the subject.
The AI needs to be instructed that if it is building carriers it might be an idea if it makes fighters to use with them.
Oh and stop the AI from building drones and fighers with no weapons.
Planetry defences are also needed, a couple of weapons platforms can make a huge difference, and ground troops are essentail for defence (actually I think the amount of milita units should be upped my a factor of probably 10).
Another thing I think the AI needs to learn is the value of sensors, whacking a dozen or so levels of sensors onto your ships makes a huge difference.
Well I think thats enough ranting for now.
FYI to the OP: The < and > symbols are used by the board as part of its formatting, so you need to replace then with special symbols &l t; and &g t; (just remove the space) to get them to display correctly without messing up your post.
Its simple... some people tend to conquer, rather than destroy, enemy planets. Evidently your not one of 'em.
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
Nearly Forgot
Upgrading facilities
For ground constructions, the AI is setup to build Facilities first, then units, then facility upgrades, and last unit upgrades (if they were possible). So upgrades do happen when there is no space to build new facilities or units are not in demand - which is a rare situation in-game for the AI.
It's easy to look at time with ship upgrades because each design type has a date designed variable and you can make a statement if X time has passed - but that doesn't exist for facilities and I'm not sure if the AI can actually look at the level of facility either.
-----
Having now played both the stock and the balence mod
I find I like the way the stock game works better than the balance Mod except when it comes to the AI at the same settings the AI in the balance mod is a harder fight and a serious threat at the beginning of the game. Get stuck next to a major AI empire and you will have a lot more of a problem fighting it however the way round this is to deal with it diplomatically as drawing a strong AI into an alliance early can give you the boost needed to really take off.
cordas wrote:What the game
We also need an AI that is able to reasonably work out what are the best facilities to put on planets, I have seen the AI build 20 mineral miners on a planet with 4% minerals, so its actually running at negative production due to maintainence costs.... Surely its possible to put in a line that says if resouce less than 50% then don't build resource extractor.
Another thing the AI needs to do is be able to understand it needs to upgrade its facilites, if only so I don't have to when I conquor the planet .
Errr this is starting to turn into a bit of a rant I know, but whilst I am on the subject.
The AI needs to be instructed that if it is building carriers it might be an idea if it makes fighters to use with them.
Oh and stop the AI from building drones and fighers with no weapons.
Planetry defences are also needed, a couple of weapons platforms can make a huge difference, and ground troops are essentail for defence (actually I think the amount of milita units should be upped my a factor of probably 10).
Another thing I think the AI needs to learn is the value of sensors, whacking a dozen or so levels of sensors onto your ships makes a huge difference.
Well I think thats enough ranting for now.
Good points... Though I'd tend to think that the difference between the A/I and a good human player is not that the human player avoids dumb mistakes, it's that she ensures the dumb mistakes are on the periphery of her strategy so they don't get noticed.
ground combat
As I said, follow my instructions when it comes to "capturing" planets.... This is a stratergy I have used in SE4, when I would simply bombard the planet and carry on demanding they surrender until they did.
In SEV it seems the AI will never surrender under any circumstances (in my experience), but it is possible to land relitivily few troops on a homeworld and capture it within a couple of turns of starting a new game, this instantly doubles your research and resource gathering and gives you a 2nd space yard (assuming you start with 1 planet) with the added bonus of often getting another atmosphere breathing race in your empire.
This is an incredibly powerfull way to start the game, and to me is too irrestible to avoid. Yes you can bomb the planet back to the bacteria and colonize it yourself and gain a reasonable bonus of a high resource large planet that might just be breathable by you, but its actually harder to do and less rewarding than invading it.
Yes this does come down in part to my gaming style, I play like the borg (tm ignored cos I ain't a trekkie ) and will assimilate every race I come across, but its the style I like and don't want to change because its too easy. I would rather that it was made harder to do. Too my mind anything that makes my game plan harder to implement is going to help the AI.
How about this then!
Ah I see, its simpler than I thought then, much simpler if all that code is already in place. If your saying facilities already upgrade.
To post this as simply as I can, Couldn't two or more of these lists be coupled with a few of these statements:
if (sys_long_Game_Date <= (Sys_Get_Game_Start_Date() + 50)) then
Run standard list.
For each colony:
// a. Add Ground Constructions
// 1. Facilities
// 2. Units - How many to build
// 3. Facility Upgrades
// 4. Unit Retrofits
endif
if (sys_long_Game_Date <= (Sys_Get_Game_Start_Date() + 70)) and (sys_long_Game_Date >(Sys_Get_Game_Start_Date() + 50)) then
Run Upgrade list.
For each colony:
// a. Add Ground Constructions
// 1. Facility Upgrades
// 2. Unit Retrofits
// 3. Facilities
// 4. Units - How many to build
endif
if (sys_long_Game_Date <= (Sys_Get_Game_Start_Date() + 120)) and (sys_long_Game_Date > (Sys_Get_Game_Start_Date() + 70)) then
Run standard
For each colony:
// a. Add Ground Constructions
// 1. Facilities
// 2. Units - How many to build
// 3. Facility Upgrades
// 4. Unit Retrofits
endif
etc.
This gives 20 turns for upgrading, 50 for building and can be repeated. Far from perfect but its only the beginning of what we can do surely.
We could throw in a 'is not at war condition'. And we could make this a simple loop with bit of clever math.
If Upgraderun = 0 run this list
Upgraderun = 1
Standardrun = 0
end if
If StandardRun = 0 run this list
etc...
-EDITED Because I put one of the lists the wrong way around...
---
Designing an advanced, diplomatic AI + shipset
Consult the great SEV Wiki
AI checks for upgrades each turn
Maran, it already checks each turn if there is a facility upgrade it can do (see Fx(Get_Ground_Constructions_Percentages) in Script_AI_Construction - the problem is its priority in comparison to other items. Since its usually a one-off upgrade, perhaps it should be done before anything else?
-----
I am about 75% certain its
I am about 75% certain its doing it because of the order the construction list is given. The percentages/(base priority points for space stuff) ,I am guessing, doesn't cause the AI to change the order of these four build catergories, only help it decide what to do when it reaches each one.
People saying units hardly ever build reinforce this theory.
So, if I am right, it will always do them in this order at the moment:
ITEM_TO_CONSTRUCT_FACILITY:
ITEM_TO_CONSTRUCT_SHIP_DESIGN:
ITEM_TO_CONSTRUCT_UNIT_DESIGN:
ITEM_TO_CONSTRUCT_UPGRADE_FACILITY:
What I am proposing is to reorder the queue right from the off or heck even randomise it if I could work out the (rand) command the engine uses
Here is the basis of code I have so far if your interested, its set to build for 25 turns then upgrade for 10; those values are easy to tweak for your AI's tastes. If there isn't any upgrading left to do, then it won't hurt the normal building order, it will however build units in those ten turns but that again, is easy to change depending on what you want:
(Old variables above this)
mupgrade_counter: long
mupgrade_use: long
mupgrade_second_counter: long
begin
// Maran - This has been added and expanded to force the AI to upgrade more often than virtually never.
if sys_long_game_Date = (Sys_Get_Game_Start_Date()) then
set mupgrade_counter := 25
set mupgrade_use := 1
endif
if mupgrade_use = 1 then
set mupgrade_counter := mupgrade_counter - 1
endif
if mupgrade_counter = 0 then then
set mupgrade_use := 0
set mupgrade_second_counter := 10
set mupgrade_second_counter := mupgrade_second_counter - 1
endif
if mupgrade_second_counter = 0 then
set mupgrade_use := 1
set mupgrade_counter := 25
endif
if mupgrade_use = 1 then
set retval := FALSE
set num_items := 0
set list_count := gbl_lst_item_type.count()
if (list_count > 0) then
set index := 0
loop
set index := index + 1
set item_type := gbl_lst_item_type.get(index)
set item_id := gbl_lst_item_id.get(index)
if ((item_id > 0) and (item_type > 0)) then
case (item_type)
ITEM_TO_CONSTRUCT_FACILITY:
call amt_to_spend.assign(gbl_resources_facilities)
set item_cost := Sys_Get_Facility_Cost(sys_long_Player_ID, item_id)
ITEM_TO_CONSTRUCT_SHIP_DESIGN:
call amt_to_spend.assign(gbl_resources_vehicles)
set item_cost := Sys_Get_Vehicle_Design_Cost(item_id)
ITEM_TO_CONSTRUCT_UNIT_DESIGN:
call amt_to_spend.assign(gbl_resources_units)
set item_cost := Sys_Get_Vehicle_Design_Cost(item_id)
ITEM_TO_CONSTRUCT_UPGRADE_FACILITY:
call amt_to_spend.assign(gbl_resources_facilities)
set item_cost := Sys_Get_Upgrade_Facility_Cost(sys_long_Player_ID, item_id)
endcase
if (amt_to_spend.IsGreaterThan(item_cost)) then
set retval := Add_Specific_Construction_To_Queue(queue_id, index)
if (retval) then
set num_items := num_items + 1
call amt_to_spend.Subtract(item_cost)
case (item_type)
ITEM_TO_CONSTRUCT_FACILITY:
call gbl_resources_facilities.assign(amt_to_spend)
ITEM_TO_CONSTRUCT_SHIP_DESIGN:
call gbl_resources_vehicles.assign(amt_to_spend)
ITEM_TO_CONSTRUCT_UNIT_DESIGN:
call gbl_resources_units.assign(amt_to_spend)
ITEM_TO_CONSTRUCT_UPGRADE_FACILITY:
call gbl_resources_facilities.assign(amt_to_spend)
endcase
endif
endif
endif
exitwhen (index >= list_count) or (num_items >= max_items)
endloop
endif
return retval
endif
if mupgrade_use = 0 then
(Same code again except rearranging the two instances of this)
case (item_type)
ITEM_TO_CONSTRUCT_UPGRADE_FACILITY:
call gbl_resources_facilities.assign(amt_to_spend)
ITEM_TO_CONSTRUCT_UNIT_DESIGN:
call gbl_resources_units.assign(amt_to_spend)
ITEM_TO_CONSTRUCT_FACILITY:
call gbl_resources_facilities.assign(amt_to_spend)
ITEM_TO_CONSTRUCT_SHIP_DESIGN:
call gbl_resources_vehicles.assign(amt_to_spend)
endcase
Though there may be a better but more complex way using some of these variables, you might be able to use the top 10% of resources to upgrade and the rest for building or some such.
call total_resources.clear()
call total_resources.Add(Sys_AI_Empire_Stats_Get_Current_Resources(sys_long_Player_ID))
call total_resources.Subtract(Sys_AI_Empire_Stats_Get_Total_Queue_Usage_Per_Turn(sys_long_Player_ID))
What do you think?
I can post you the full code if it will help (and compile )
-EDIT
I had to edit that code, it was so error prone, it now compiles in the full version and i'm going to test if it works.
---
Designing an advanced, diplomatic AI + shipset
Consult the great SEV Wiki
I should say if all you
I should say if all you wanted to do was have it upgrade before building then changing this to be the first function called of the four should do:
ITEM_TO_CONSTRUCT_UPGRADE_FACILITY:
call amt_to_spend.assign(gbl_resources_facilities)
set item_cost := Sys_Get_Upgrade_Facility_Cost(sys_long_Player_ID, item_id)
However to me thats not as powerful as doing it with a few turns to spare.
Take an example where the AI builds its first building on a planet, then immediately, rather than filling up the remaining building spots, its upgrading, which is inefficent for a new planet
However your way would mean no time was wasted using upgrades, hmm tough one to call. I'd still like to cycle it so the AI occassionally builds units for a few turns however.
---
Designing an advanced, diplomatic AI + shipset
Consult the great SEV Wiki
Very interesting
Might work better if we where a bit more prescriptive of what an AI should use rather than leaving it to adapt to experience. All AI warships should have some PD on them by default in my opinion but perhaps it would be possible to have an AI favour a particular weapons set shields and type of armour. It may be that they do this already.
I also think that all AI's should research, research to a fixed amount of there research capacity and upgrade these installations independently of other upgrades. This is particularly important if neutral AI's are to be more of an obstacle and less of a speed bump given there limited resource base there are also certain universally sound tactics which should be part of any AI's repertoire such as mining and dropping satellites at warp points in there empire as well as stationing fleets at access warp points to hostile races which they do already.
Teaching an AI to build a balanced fleet with synergies would also help as they are given to building large fleets of very similar ships at the present. This is some what hidden by the fact that the AI does not maintain class names as players do, but have a look at the ships on the scanner in combat.
Once again I ask Captian Kwock to publish the AI changes from the balence mod as a start as the AI is more chalenging in the Balence Mod and that would make the stock game more chalenging without changing the game more generaly in the way the Balence Mod does
AI Scripts
I'll post the v1.00 scripts with the next release. The design creation scheme should be fairly adaptable to someone who is interesting in making it compatible with stock.
-----
I don't understand any of the coding stuff but...
Too chuck in my two pennies.
Is there anyway you could get the AI to choose what it upgrades 1st, the obvious being space yards. I would even have it do this XX% of the time even if its at war.
I would then have it choose to upgrade things like the political centers or shrines (if its religious) and resuply depots. Then work down the list of anything that has system wide effects before moving onto "lesser facilities".
If its possible it would also be great to have the AI check to see if its running short of any resource and if so to look to see if it can upgrade any of those facilites *then bunk this up the upgrade list* and maybe to spend research points on improving that tech for X levels then upgrading those facs again.
The AI really suffers at the moment from not upgrading its facs, even more so if it is spending research points on upgrading these techs. Could you also add in wieghting into the equations so that the higher the level difference is between what it has built and what is available to it the more likely it is to upgrade.
Personally I would think that doing this on an empire basis rather than a planetry one should be much quicker, but it might cause some confusion when they have built different levels.
Hmmm on thinking about this, how does the AI decide what it is going to research? Is it possible to tie this into the upgrade loop. I know as a player I tend to go through cycles of different types of research, concentrating for a dozen or so turns of facility techs, then a dozen or so weapon techs, then half a dozen ship improvements, then back to the begining and every so often i will bung in some civics.
If the AI works in a similar way (or could be made to) then this would make it even easier to manage upgrades, when it finishes a facility tech push it auto upgrades whilst its doing it ship and other techs. Then when it finishes its next tech facility tech cylce it could be pushed to go again.
Also any planets with full ques and no space yards should be doing facility upgrades at least every XX turns, irrespective of what else is going on.
Thank you for that
-----
I might even have a go at it myself not that I am all that good at coding dyslexia and not haveing done it for a few years realy dont help
AI Strategy paths: Intensive vs. Extensive
Seems to me many of these A/I suggestions make assumptions about optimal strategies that may be independent of each other.
It may well be that upgrading facilities first makes sense. But that may be premised on a "teching" strategy rather than a "rushing" or "booming" strategy.
Come to think of it... it may be that the optimal strategy for the game (itself) would be to take a pass on the project of building an A/I for a generic game, and instead open up the possibility of designing and scripting scenarios -- vis. Starcraft/Warcraft's methodology.
NOOO!!!!!!!
NOO!!!!!!! (no offence) Please don't do that, what I love about TBS against RTS is that the games tend to be far more open and free forming. Creating scripted scenarios ends up making a very different game to what I want SEV to be.
If you or anyone else wants to make scripted mods then great go for it, and I will give them a play I am sure, but they tend to force you into one type of play and one type only.
One the things I really loved about SE4 was that there where so many valid stratergies available to the AI and the players (in later versions once the bugs where ironed out) that each game was different, and playing online against friends was challenging because we each had different styles and stratergies to use against each other.
You know, I tend to agree
You know, I tend to agree that scripting is a good thing...script the AI ships, script its formations, strategies and fleet stuff.
its okay...
Giving the AI freedom and tons of options equals a shoddy AI...
Chess AI can be really good, even in a simple program...but you have to restrict its 'options'
Janster
Fair objections
Fair enough, Cordas... I can see why you feel the way you do. But you know of course that the logical outcome of your argument is "come out and play."
Games against players will be challenging and free-form, as you say.... Games agains the A/I will always be competitive puzzles, even if the A/I has gotten to the point of "attacking the opponent's strategy" -- even this ability to counter the player will be limited by the A/I's library of strategies.
One specific goal for Multiplayer, by the way, is to ensure that the A/I plays a good "turtle" game -- one doesn't want Multiplayer decided in the first 10-15 turns by the player who is first able to discover and conquer a neutral empire.
Of course use scripting, but
Of course use scripting, but my impression from what crward was suggesting was to use the type of scripting used in RTS games, which tends to be very formulaic and lacking in any form of imagination.
I am all for scripting the AI so that it behaves reasonably in given situations (such as building a fleet to break a warp gate blockade, rather than sending in ships one at a time to be killed without them firing more than one shot), but I still want an AI that is going to play every game differently rather than follow the same script for expansion.
After all this game is far more than chess where you have a finite number of peices and finite number of legal moves possible with the opponent having the same limitiations. Or a RTS game where the idea is to balance harrying with building overwhemling numbers to crush the enemy.
A/I and player puzzles
I am all for scripting the AI so that it behaves reasonably in given situations (such as building a fleet to break a warp gate blockade, rather than sending in ships one at a time to be killed without them firing more than one shot), but I still want an AI that is going to play every game differently rather than follow the same script for expansion.
After all this game is far more than chess where you have a finite number of peices and finite number of legal moves possible with the opponent having the same limitiations. Or a RTS game where the idea is to balance harrying with building overwhemling numbers to crush the enemy.
That's the problem with the A/I though: the game is "far more than chess." The result is that a "good" A/I (to deal with, for example, the ground combat game), is easily beaten by changing games. The interim trick, I suspect, is to create scenarios where the player is constrained to play against the A/I's good game. Perhaps over the course of creating these scenarios, certain A/I scripts could become more resuable for the open game.
Possibly, or the game could
Possibly, or the game could just get down to forcing you to play one style of play against and AI and not be able to do anything else...
Its a matter of balance
Thoughts.
Great thank you, I will be doing the same with mine when I get a bit further along with them, there are still several aspects I dislike about my AI's but they are looking more like 2 complete races now.
I haven't touched any coding myself for several years, I can tell my conditional statements are not very well written but they get the job done for now, i'll probably look back in a month or so and smirk at what I had written
It may well be that upgrading facilities first makes sense. But that may be premised on a "teching" strategy rather than a "rushing" or "booming" strategy.
Come to think of it... it may be that the optimal strategy for the game (itself) would be to take a pass on the project of building an A/I for a generic game, and instead open up the possibility of designing and scripting scenarios -- vis. Starcraft/Warcraft's methodology.
There's room for both as far as I am concerned. Just like there is for gameplay mods and such. An AI assuming a certain strategy is fine and playing to all that strategies strengths is fine, if that's what the AI is designed to do, more flexible ones are just as possible to code but harder to perfect in such a way.
Personally I want to release a good few stock AI's first, that way the community will have something to work with. That and I loved SEIV's opened ended playstyle also, watching the AI development over the years was interesting, with the scripting we can do now it should be even more so.
Eventually there should be quite a diverse mix of them, all usable in one game, which will be challenging in itself to play against. Especially if they are being updated and tweaked so you never quite get 'comfortable' with them
---
Designing an advanced, diplomatic AI + shipset
Consult the great SEV Wiki
Upgrading Facilities
Upgrading Facilities seems oddly bugged. I ran some tests on this, both using the default AI & the Balance Mod 1.00 AI, and couldn't get either to upgrade my facilities, no matter how I tinkered with the ministers. It would just sit there doing nothing despite a mega cheap storage upgrade sitting there to build, and a huge excess of income.
I think the bug is either in the SetGroundConstructionPercentages function, whatever the exact name of it is, or in the actual code of the game itself. I did some playing trying to make it work and made no progress, though I'm still only just feeling my way around it all.
Either way..... it would be ideal to seperate it into a seperate AI minister. So you can turn the upgrading control on without having it also try to build structures. Though then you would need to get the minister to check if the current construction item was a very short project and the upgrade important or not.... otherwise it would just never get to put in an upgrade if it only did it if the production queue was empty.
If anyone can work out how to make it go.... give me a yell please.
Also be nice if the AI scrapped an unimportant building for a usefull one discovered through later levels of tech But thats even harder to code.
Two items
Item 1: If a player could a) delete obselete designs AND b) prohibit AI's from building specific facilities, ships, and units, I think it would enhance the game. It would prevent strife between the player and the AI's. It would also allow a player to force the Ministers to support the Empire's "political" policies. For example, you could decide that your Empire will not use certan weapons, units, ships or build certain facilities or that you will only use them on a limited basis. You could build a certain number of items, them stop all future production. This would help a player to tailor the game more precicely to his liking.
Item 2: I am interested in seeing Babylon 5 style mods. Including the Vorlons and Shadows and their penchant for working behind the scenes through other races. I would love to work on this myself but, I have no skill or knowledge in this area. But, I am certain that many of you have such talent.
Carriers
I must agree with Cordas, if you build a carrier, the AI should also build the fighters to go with it. If you have more than one type of fighter, you will need to select which ones and how many of each.
I also think that a Carrier should allow either Fighters or Drones.
And it is very annoying to build a ship which turns out to be a disappointment and have your AI fall in love with it and build it in preference to anything else. I would love to delete these designs completely because making them obsolete does not discourage the AI from building them anyway. I waste a lot of time checking constuction ques and removing unwanted items.
Fighters & the AI
The odd thing is that the AI will use fighters if it has them it just won’t build them off its own bat or wont in any useful quantity. I dropped in on one Stock AI and had it build 10 of its standard fighter at every planet. Not one which I had any contact with and I didn’t load the fighters onto any of the half dozen of the empty carriers available and left the AI to itself after a couple of turns. almost a game year later I invaded one of its systems after a border incident and as I often did blockaded the system with small fleets of missile frigates at each of the incoming warp points whilst I went round and captured all the planets the turn after I took the last one there was a combat announced. This turned out to be the AI I had helped and later attacked forcing the warp point with a large fleet am mix of carriers and cruisers about seven in all. I ordered the my squadron to run to force a pursuit fight where my out gunned and out numbered frigates would have the advantage. Then to my surprise the AI scrambled about thirty fighters and even though I had ample PD on each frigate they could not cope with a fighter swarm for reasons of space all my Pd was blaster based and so short range.
There followed a nasty little fight where I killed about half of his fighters and slightly damaged one of his cruisers which was at the lead of chasing group and I lost three frigates quite quickly. It did not constitute a major defeat as I still had the escort fleet to the invasion in system but it proves something I think and was a nasty surprise.
Food for thought
The biggest problem for the
The biggest problem for the A.I is lack of memory when scouting alien planets.Until MM programmes the game to remember planets last known ownership like GC2 and MWT2 do with fog of war,the A.I won't attack because it is unaware of the aliens presence in other systems.I hope this is a priority for the next patch.
Edit.It would also stop the A.I from always sending colony ships to systems it can't colonize due to treaty's.
It's possible to do that... Ugly, but possible
It's possible to program the AI in such a way that it remembers the colonies it sees. It's a huge amount of work to get it working right, but it's possible.
Over the next few weeks, as I work on my AI for the mod I'm making, I'm probably going to include this for the AI.
-----
Give me atmoshperic manipulation cannons, or give me death!
The AI apparently does not
The AI apparently does not put into consideration any treaty elements or other factors when it chooses a target to attack. I took a lone ally planet (the treaty did not include non agression settings) just for the different atmosphere breathing population. The AI got mad at me for this and declared war on me and that one planet seemed to be its priority. It was in the middle of my territory and was defended by satellites and a base at the warp point entrance. It sent its biggest fleet and broke through my defenses, (my base was extremely obsolete but they still took heavy losses) and attacked my planet. Turns out they had a treaty preventing them from bombarding planets, and since the AI never captures planets, the fleet just ran away and was soon annihilated when I sent a fleet to dispatch them. They wasted most of their ships in a fruitless effort, which left them open to attack, and I soon destroyed them. The AI should check if it is allowed to bombard planets before it orders ships to attack a colony.
Use the balance mod and the
Use the balance mod and the A.I won't give out stupid treaty's.
Just tried it yesterday
I just downloaded the mod yesterday, and I must say, the AI in the Balance Mod is a MASSIVE improvement over standard SE5 AI. I am only about 20-25 turns in, so I don't know if they stay this way, but so far I am very impressed. I am actually having a hard time trying to capture one of his planets because he has stationed 20 fighters on it and 2 weapon platforms with 3 point defense cannons each (I didn't know until now that point defense on WPs could fire at troops).
Captain Kwok has done a great job in improving the gameplay experience overall. For all of you who complain about how retarted the AI is, but have not played the balance mod, please go download it for yourself. I agree, they are still nowhere near as challeging as a human, but they're much better than the stock AI.
Bettter
Actually, they couldn't before... thats just something thats been getting modded in since the idea popped up 
And the AIs are gettign MUCH BETTER. Go try Maran's AI if you want to... from what I can tell its pretty good.
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
Quote:And the AIs are
When did he release his AI? Or part of his AI? Did I miss something in the past week?
Been Released
When did he release his AI? Or part of his AI? Did I miss something in the past week?
Its been released for a while... he's has .6 IIRC, with .7 "on the way".
He's only gotten his peaceful AI all the way out, the others are still in the works but I think they come bundled (incomplete version) with his main peaceful one.
Go look in the modders forum. I'll go try and dig up a link...
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
What i'm annoyed by is that
What i'm annoyed by is that i'm playing a game right now where i have 144k research points per turn... next closest ai? 9000 or so. The gap shouldn't be this insane..... some AIs don't even HAVE research points!
The "holy grail" in strategy games isn't a competent ai.... the AI in Galactic Civ for example can keep up with players in basic game mechanics same with Civilization 4, only exploits is what that has problems with. The AI should be able to keep a pace with an average player. This is my first game with AI on hard and i'm learning as i play and i'm still demolishing them utterly to the point where i could just send 1 ship to a planet with 20 satellites, a starbase, 7 ships, and weapon platforms and defeat it in seconds.....
Its going to be funny when I have ringworlds and the AI is still stuck on medium ships....
I remember the last game, SE4... the AI had no idea what so ever how to use stellar manipulation. If you played a universe where not every area was connected an AI could easily be isolated for an entire game simply because they weren't programmed to use it. However the AI could keep up to me on that game far more than the AI in this.
I know this is the epitome of a small company but you would think that after releasing 5 games that things would expand beyond this.
Tools
Try Maran's AI. Malfador has given us (most) of the tools to create a dozen great games through modding. Maran's uses stock game and expands peacefully, focusing on research and expansion. Nice AI, though I dislike several things about it.
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
Slow research kills them
Slow research is definitely something that needs to be looked at in stock. I've played both BM and against Maran's AIs, and both are much better, with Maran's AI being insane at the start. I just wish it wouldn't build colonies a few sectors from my homeworld on turn 4. Yay even distribution not working properly...
I've started work on my own AI. I'm trying to get it to a point where it will actually plan events several months ahead of time and get ready for them by using the variable storage. It's slow work since I haven't found a way to do 2D arrays, but I asked if MM could add support for them.
And stock SE:IV AI could use stellar manipulation. Around 2500 in one of my longest running games, the other remaining empire opened up a warp point right to my home sector. It made me mad 
-----
Give me atmoshperic manipulation cannons, or give me death!
Wish upon a star
It would be really cool if someone made some sort of AI programming GUI for us people who are good at strateegery but don't know nothin bout programmin. I know this is pie in the sky wishfulness , but I can still dream.
Kerry-
"...I say we must move forward, not backward, upward, but forward, and always twirling, twirling, twirling towards freedom!"
-Kodos for President
Lame AI Practices
I've been playing a 1.20 game and got about 350 turns into the game. For several hundred turns, yes several hundred, the AI's were basically asleep. After trashing one AI the other two become pals through treaties. After that they just dozed off. I found one AI's carrier over 10 sectors away from his nearest colony and it just sat there for hundreds of turns without ever moving. One AI sent a "slow boat to China" colony ship to colonize a small planet in one of my systems but I parked a ship over the planet and it turned aorund again. It had movement of only 1 point yet wasn't damaged at all and had full supply.
The AI's need to pick better planets to colonize, the current selection stinks. It looks to colonize small or medium worlds yet bypasses larger better planets. It also needs to stop trying to colonize in human owned systems because thats just suicide.
The AI's need to build better facilities like monolith facilities or system/planet production boosters like robotoid factories or mineral scanners. The AI's need to build atmosphere converters and climate control as I've found them invaluable. The AI's need to scrap some facilities in order to build other good facilities.
The AI's need better target selection. I had one system in my 1.20 game with 3 major huge/large worlds with breathable atmospheres plus 2 domed colonies in one system. I had about 6 ships in system while the AI had 2 fleets totalling 34 ships. After the AI decided to break our treaty and attack it wasted one of my small worlds. The next turn it took out one of the big colonies and I figured it was time to write off that system. Then one turn later there's a major 19 ship fleet heading off to attack one of my planets in a system 3 system away. The other fleet just sat around in system but didn;t attack and then moved off. Thanks to the heads up from my scanners I was able to intercept that big fleet in it's home system that I had just destroyed and had 23 ships waiting to welcome the AI marauders. I slaughtered the AI fleet with very few losses and I was able to retain and rebuild my system's 2 destroyed colonies. The AI's need to be better at localizing their attacks and once in an enemy system they should stay and attack all colonies and clear the system.
This game has limited appeal thanks to the stinky AI's. I'm rather bored with the stock game and am contemplating shelving it until it's finished hoping the AI's will be better then. Nothing kills a game's appeal faster than a crappy AI.
Omnius
Has everyone ignored my rants?!?
Have you downloaded the balance mod!? Do my posts telling people to download it for MUCH better AI go unnoticed? How many times do I have to shout it out? Try this, download the balance mod, add as many AIs as your comp can handle, go to war with them all and you will guaranteed be challenged.
Stock AI and fixes
Destroyer224, it's also important that stock weaknesses are highlighted though so that they can be fixed. Not everyone wants/likes to use a mod to enjoy their game.
-----
Suggestion... prolly dumb but makes sence to me :)
Ive only had a brief glance down this thread but thought id repeat a simple suggestion a made a while back...
Pre-designed ships for AI races would be a simple yet effective way to make the Ai ship designs slightly less lame and would give each of the alien races a much more unique feel not to mention giving them some things to focus on in there research.
Have a selection of premade hulls for a race covering various levels of tech and make it produce them as needed and upgrade when they have to then have the AI focus its ship based reasearch in areas that unlock its better ships.
Marans AI Mod
That’s what Maran's AI mod does it will when its finished give three distinct competent AI's one peaceful and expansionistic one an aggressive user of Intel and the other least well developed is a hive race. I do agree that the AI in stock needs improving but I am happy for Aaron to continue to work on the rest of the game as AI moding seems to be well in hand
My mod does it already.
I have done it a while ago, but no one seem to notice...
Inter-Related Tactical Conversion v 0.54c
The Holy Grail
cordas,
Ah the Holy Grail of computer wargamers everywhere, an intelligent AI. Ofcourse you know that AI just stands for "Artificial Ignorance". I agree it would be a boon to the industry if some programmer could master this impossible task.
If only some programmer would understand that making AI's agressive just makes them suicidal and easier to beat. If only some programmer would make an AI that would emulate Mohammed Ali's famous "Rope a Dope" defense that he used to trash George Foreman. I use it constantly to trash AI's because I know the AI's will be stupidly aggressive and send in their forces to the same place piecemeal.
When my latest 1.17 game choked I ended up looking at the offending AI to see if I could determine what was causing the hangup. I saw some really stupid stuff going on. Ships with 3 move/attack orders (two of which were duplicate). AI ships should only have one movement/attack order at a time because things change and AI's that have so many orders in their que end up doing stupid things because their orders are out of date. The other bad thing I noticed when I switched back to my original race was the absolute mess the AI made of my position in only 3 turns. It moved ships all over the place willy nilly with no real purpose. It started building worthless satellites over all the planets. No wonder the AI stinks in SE5 because it just isn't programmed to build smart ships or facilities.
In SE5 one big improvement would be making the AI build ships smarter. One thing I've noticed is the AI's do not understand to build ships from the inside out and that makes their ships crappy as can be. Way too much supply and ordnance, poor choice of weaponry, and other poor choices of other spiffy items makes AI ships all too easy to wipe out.
I also agree the AI's ought to be using better facilities like the planet and system production/research/intel boosters. Another smart thing the AI's don't do is upgrade facilities.
I always think of George Carlin's classic comedy routine about tv's when I think of AI's. He said that "there's a brightness knob on the tv but no matter how much I turn it the tv never gets any smarter". There's a hardness knob on every AI and never have I seen an AI actually play smarter no matter how much I jack it up. All it ever does is give the AI more garbage to lose.
Omnius