AI not challening? |
The computer plays like a chump on high difficultiy no bonus 
If I give it a bonus will it expand faster and actually increase it's research points and what not? Has anyone had any success getting a challenging game from the AI?
Do I have to play Human vs. all AI?
human vs all is impossible to win
played twice and cant win - they kill me via intel ;/
Play against yourself or your friends
I'm afraid I've given up on the AI for the moment. I'm playing a game with four empires which are all under my control.
Doing it properly takes ages and you can't have a favourite empire or it throws the whole idea out of the window. You just have to play each one differently and with no mercy.
Only the Awkward question; only the Foolish ask twice
Try to play “God”
If I give it a bonus will it expand faster and actually increase it's research points and what not? Has anyone had any success getting a challenging game from the AI?
Do I have to play Human vs. all AI?
Well, until the AI becomes challenging, especially regarding ship design(I hope it won’t last too long) there are other ways to have fun with SE-V : Try to play “God”. Try to reach a technological level which will allow you to completely redo the map : blow planets and stars, create new ones, build ring worlds and sphere world everywhere.
You can even try to do some modding.
But I agree with you it’s a pity the AI is so dumb
-------------
Feel lucky earthling ?
Shamless plug...
The bonus is the true difficulty setting and will allow them to build/produce things somewhat faster. The stock AI isn't really well developed in terms of ship designs and things of that sort - so they the numbers.
The Balance Mod does make some strides in this area if you're interested in it. See the link in my signature below. 
-----
Balance mod design scripts
Yes, good job, I think your ship design scripts should definitely be included in the stock version.
-------------
Feel lucky earthling ?
AI v SI
No offense meant, Captain Kwok as your AI is much, much better than the stock AI. But I'm still going to keep playing against me, myself, I and the other me...and sometimes other real people...just once I get the hang of making friends!!! 
Why? Because the AI is still really using numbers to beat you...No cunning at all...where as we can use all the cunning we want!
Lets look at a simple cunning plan...
In tactical combat the AI Ships tend to fire on the first target that comes in range...this could be used against them. Make a Ship that is just all armour and shields...one that can take alot of punishment give it 1 or 2 weapons so it is still a threat...send them in first to draw fire...then once the enemy is busy send in ships with lots of weapons to kill the enemy.
Why is the AI dumb...becoz it won't switch targets to attack the lighter armoured but heavily armed ships even tho they are the real threat...it doesn't adapt to the situation...this is true of any AI in any game not just SEV.
One day in the future, maybe we'll see real AI and not SI(scripted intelligence)
Well now, I'm going to get back to playing SEV...whose turn was it? Oh yes, its my turn!
Only the Awkward question; only the Foolish ask twice
you could script which
you could script which target will be used - based on various selections - its just stategies used - you could made own one's
but it could take too much "if .... else .... then ... " 
SE3 Strategies
I think it was possible to specify the target priority back in SE3, but I has been years since I played SE3... You could choose to go for the biggest, the smallest, the slowest... something like that...
I do not know why this feature has been discarded...
wasn't
I do not know why this feature has been discarded...
It wasn't. Check empire options, the Strategies list.
I-Robot refers to a concept
I-Robot refers to a concept called a "Difference Engine". What a difference engine is is a universal calculation designed to assign a percentage to any situation (who has the greater chance of survival, who is the greater threat, shall i take the left door or the right).
It accomplishes this by taking every little factor into consideration and then assigns positives and negtives to every factor. It then calulates that to give a basic percentage. It then uses that percentage to choose.
I-Robot covered that with only 1 flaw. It will always make the 'logical' choice.
In most situations the logical choice is the correct one to make and works very well in strategic situations and would make for a very good AI.
The only bad thing about it is "irrelivent to my point", because the bad part behind it relates directly to emotional situation which are inherently illogical.
So how would you make a difference engine that could work with SEV combat system?
Every aspect of a ship will have a threat value associated.
The hull size and it's componants
Componant threat value can be created through calculation, and that calculation can be automatically be answered by effect and amount.
Weapons would have a high threat because their 'effect' is to cause damage so it would have a high multiplier.
Cargo would have a low threat because their 'effect' would have a low multiplier.
The difference engine would have to disregard range and make a direct attack on the greatest threat. A simple rule can be in place to shoot at nearest if the threat is out of range, but would have to continually check the range so that it switches target when the threat is in range.
I've left the rest pretty open...discuss.
Your lord and master (below Foamy) LordHavoc
Cargo = low threat?
I'm not sure about making cargo = low threat. I often prioritize cargo ships fairly high, simply because they tend to be defensless and carrying lots of stuff I don't want arriving.
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
Cargo...near or far?
I think that the cargo threat would have to depend on what type of cargo being carried and the nearest friendly(to the AI) planet.
Cargo ships that are X hexes away from a planet could have a low threat value but as they get nearer to a planet their threat value could rise.
Only the Awkward question; only the Foolish ask twice
Yeah
Cargo ships that are X hexes away from a planet could have a low threat value but as they get nearer to a planet their threat value could rise.
Only the Awkward question; only the Foolish ask twice
That makes sense... I was just nitpicking 
I tend to prioritize cargo ships that carry troops, fighters, or any other units generally.
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
There's a solution for that...
Sincerely,
Kahn
Almost forgot one other factor...
In order to bring variation to each different AI player (after all, you don't want them all playing the same, that too is predictable and boring), at the beginning of each game the AI players would each adopt a personality based partially on their race attributes, and partially a random decision. Personalities would also have a random modifier to determine aggressiveness. A particular personality could be for instance, an aggressive colonizer, or may show a preference for carriers in combat, or be a defensive expert with lots of starbases, drones and minefields. The possibilities are endless.
That's why many games have adopted a scriptable engine for AI's using Python or Lua for the language. That way, just about any type of behavior you can possibly think of can be scripted.
Sincerely,
Kahn
Logical decision not always correct
Then there's the analogous situation where we face off with the same fleet, and if the opponent misunderstands my purpose, for instance, thinking it to be an invasion force and the cruiser and destroyer screen to be the diversion, rushes and is promptly caught in a trap because the transports were the decoys and my real purpose was to smash his fleet, and I was willing to sacrifice the obsolete transports and obsolete troops to draw him in, and then I simply bomb his planet into rubble now that there are no defenders.
The problem with such a purely logical approach is that it fails to function when the information about the situation is incomplete, and if the AI always picks the logical decision based on what it knows, it will be mercilessly abused.
Quote:The problem with the
Quite true. However in a bloody conflict a 'plan' is never the same from its initial formation to its end result. We change what we are doing to fit the situation, often making many small logical choices.
But I see what you are saying, AI in its current form is unable to set a long term goal that is capable for turning the tide of the battle. For example, making a drastic 'illogical' choice that could potentially cripple a races economy like tackling a heavily armed planet.
For the actual battles, you could pre-program a bunch of tactics and have the AI randomly decide which one to use on creation of a fleet. Although not ideal would add some spice to combats as you wouldn't know whether the AI is going to do a feint, pincor or a hit-n-run. (sorry, spelling)
That point could be linked to the AI limitation of only assessing real threats instead of potential threats (covered above).
In a situation where you are attacking a fleet with both offensive ships and transports, the most logical choice is to attack the offensive ships because they can damage your ships. To attack the cargo ships is 'less' logical (not illogical!) because to attack the cargo ships while offensive ships are still present would cause more damage to your fleet because you aren't returning fire (giving the offenders 'more time' to attack your ships).
I think we'll see large advances in AI in the next few decades now that processes and memory are getting faster and larger. I probably wont be long before a dedicated AI board becomes a standard in pcs, sitting nicely next to the graphics card (I don't think physics cards are going to take off very well personally)
Your lord and master (below Foamy) LordHavoc
Problem is it developes too
Problem is it developes too slow - I played a game against medium amount of computer players all cooperating, hardest difficulty setting, but with not all warp points connected. Now, I've finally begun opening up warp points, and I've got temporal weapons tech level 60-something, same with armour, shields, ship designes maxed out, as with most other combat techs - and the computers still are on general tech level 20ish. Which means I can send in one ship and kill them all...
I guess the only solution is to start a game with as few players as possible, have them battle it out, and hope one comes out as a really powerfull one. I remember in SEIV, the sudden shock (and joy) of having warp points suddenly being opened up in the middle of my core systems....
To quote DimmurWyrd
Amerelium, are you playing the balance mod...if not try it. You might find it a bit better.
And now for something that someone else wrote in another post... Which I think game developers should really start thinking about! And yes, I know the quote thing is all messed up but I've got to get back to work!!!!
Thoughts from an AI 'Expert'
I actually have a ph.d. in AI (well, its in CS, but the 4 yrs of dissertation work was in AI). When I first got out with the degree I applied to a number of game companies (being an avid gamer with time to kill back then in school days) but did not get that far with any (furthest was with whatever group did that 'Immortal Kohan' game) - because I can't program graphics worth a D*mn, or whatever. So now I go around 'rationalizing enterprise supply chains' & whatnot as a consultant, which has nothing to do with AI but pays the bills (v. well) - but leaves me no time even for gameplaying
Anyway, one possible 'ultimate' AI approach involves an area called 'goal recognition' (my diss work) - the AI has to be able to recognize the goal(s) of the human/other AI player and then employ a strategy (action sequence) that addresses throwing a monkey wrench into the opponents goals. This involves the AI having an internal goal/plan model for the domain (e.g. SE-V) and searching that model for relevant goals that match what the opponent is doing (such as a fleet with transports making a beeline for planet X would indicate the goal 'take over planet X'), then picking its own 'plan' (action sequence) that addresses the opponent's goal. (Of course, the bright AI would have its own goals that it was pursuing independently...) (oh, and all this would be done in a fuzzy or probablistic way so that it wasn't totally predictable by a human or other AI...) (Oh, and goals can be at different levels - strategic to tactical - that would enable the AI to reason about feints within a larger overarching strategy)
Anyway, this is (relatively) easy to do in 'simple' board games like checkers/chess where the search space of goals is not huge compared to high-end computers (one approach is called 'alpha-beta pruning' for example). Supercomputers can exhaustively search the possible-move-space to find the optimal move in each case - which is why they are getting difficult for even world champion humans to beat. Even with current PCs, you could probably put together a decent approach to an (extremely) complex game like SE-V. I don't know that it could be done with scripting - it would probably need to be built in as an AI engine subcomponent. Anyone know anyone who is willing to pay mid-6 figures to a frustrated would-be AI game architect ?
BS
We could all chip in
I would, but I'm a poor science teacher. Maybe all us SEV players could chip in
The AI you describe really sounds like the AI I would hope to find in a game...but I think we'll have to wait awhile till we start to find AI like that outside chess games.
Thanks for your ideas BS
Only the Awkward question; only the Foolish ask twice
few things
TakAhLah, if you build your fleets around the AI's specific disadvantages you will never be happy with anything but another player. We can't expect miracles here.
Personally with the AI I've found them very aggressive and difficult at the start if you start nearby another empire, but then towards the end it's like they lose interest and give up. I havn't gotten far in a game in 1.17 yet, but given the topic I'm assuming it's not too different.
I think mostly they just need to expand more, and build more- faster into the late game so we don't start dwarfing them technologically and militarily. (less sagotage though, holy christ! ) Also their ship designs and research aims (I'm assuming, I don't know excactly WHAT they're researching) could be more streamlined.
*goes to read up on the balance mod*
AI needs more research!
I havn't gotten far in a game in 1.17 yet, but given the topic I'm assuming it's not too different.
) Also their ship designs and research aims (I'm assuming, I don't know excactly WHAT they're researching) could be more streamlined.Smarter AI
Actually, I dont build my fleets to exploit the AI disadvantage and I'm not talking about that as the whole problem...The simple cunning plan was to highlight a very basic problem of the AI.
I'm talking about the fact that AI in games really isn't very good as a whole...not just in SEV and it could be and should be...no I'm not expecting a mirical what I was expecting was for AI to get better over the years just as graphics have.
This improvement of AI has not happened...play the Art of war(written years ago...it had CGA graphics...it was a cool game for its time) The AI does the same thing as SEV...attacks when it can't win. Uses the wrong formations. Does not select targets correctly. Uses ONLY numbers to win a game. It never sees the big picture. Which I think it could if game developers took some time to do...but that is asking for a mirical!!!!
Coming back to what you said about your games, Shion Kreth upping the difficulty is not the real solution...copy and stick from another post!
If the AI is too weak to compete from the start without any advantages then it's not going to EVER be strong enough because a X% advantage in resources or construction capacity etc is a fleeting one when it's not backed up by the ability to KEEP EXPANDING/EXPLOITING that advantage... This just means that at first you will have a hard time but the exact same problem will crop up every time and you will find that if you weather the initial fury you will exceed the AI's ability with ease and then crush them without any real effort and that means that JUST giving them "more toys" does nothing to really increase the AI's competetiveness.
Thank you Dimmurwryd
So in some peoples opinions it is too much to ask that the AI be improved in games...I for one will keep asking for SMARTER AI!!!
Only the Awkward question; only the Foolish ask twice
Hmmm While I find most of
Hmmm
While I find most of the suggestions interresting, the sheer truth of it is, that the AI isn't very smart, and won't be for a long time.
It needs like a little child, borders and simple rules.
Where it so far excels is in chess and very strict games.
For a game like this to work, the AI would have to work with as little parameters as possible.
However, SEV is a beast with tons of stuff the AI never understands, making an AI that actually works for this game, is out of the question, however making one that simplifies the rules, and ignores fluff like diplomacy - weird ships and gizmos and just goes back to the basics, of building killer ships and sending them out to conquer the galaxy,would be nice.
So far, even on 1.17 the AI does weird shit and I can wipe him off the table at High - low - medium bonus, altough medium and high bonus has a tendency to lag my game seriously as the AI has buildt 2342342342 sats and bases.
Which btw, doesn't save him.
A fleet however WOULD.
Janster
Hopefully we will get some
Hopefully we will get some improvements to the AI once the bugs are fixed.
Somebody left a quote tag open!
And that the AI gets fixed.
edit:
Alas, I can't fix it. MODS, HELP!
_______________________
There are 10 types of people in the world:
Those that understand binary, and those that don't.
Ensure the AI has the advantage.
For me a game has to be challenging enough or I’ll loose interest and shelf it. I noticed right away that SEV had a weak AI so I carefully watch the tech levels of all players and make sure my tech levels do not exceed that of the average AI. I have also changed the following values in the settings.txt file and in the game I am currently playing the AI tech levels are 3 & 4 times higher than mine.
Computer Player Bonus Points Multiplier Low := 6.0
Computer Player Bonus Points Multiplier Medium := 9.0
Computer Player Bonus Points Multiplier High := 15.0
//---------------------------------------------------------------------------------------
Computer Player Bonus Queue Rate Multiplier Low := 3.0
Computer Player Bonus Queue Rate Multiplier Medium := 5.0
Computer Player Bonus Queue Rate Multiplier High := 8.0
I haven’t tried to do battle with them yet but they have carriers and cruisers with big guns and I’m still running around in frigates. When starting a game I also set the maximum number of ships and units as low as they can be. I’m hoping this will help keep my game from getting too sluggish.
Good Luck & have fun…
yeah :)
but the turn processing will kill you around 300 turn woy should have loooooooooooong "processing all players turn"
well, the AI makes fleets at
well, the AI makes fleets at least - he just sent a fleet of 36 ships agains my lone guard...
Re: AI not challening?
First Off hello.
(I see this post is... well old but since im new to the game i read... well most of it)
But I wanted to add my thoughts on the Idea of AI being or becomming good/smart.
I think "BloodStone" nailed it very well, and that type of AI (IMO) is not possible with simple scripting.
Also any AI created can not and will not be able to play better then its creator(only differenc is its creator is human and is capible of forgeting or overlooking information and a computer can not. So those who are building the AI need to forsee EVERY possible setup that AI will encounter. Like the game of chess since chess is simple movements once you alow for all possible moves(actions) and counter moves(reaction) you would never move a pice that can lead to a checkmate, If i remember correctlly (let me know if im wrong) when these super chess AIs play humans and dont end in the human in checkmate they tended to end with nether side winning since the AIs first priority is to not lose. Since in SEV there is only win/lose outcomes with many uncontrolible components (not just the reactions of a single input) it is easyer to beat. In the same since Tic-Tac-Toe has so few actions/reactions it is possible to make an so called AI that will NEVER lose.
It boils down to the same requirements for a good Human player.
Knowing the Game and know all possible actions/reactions one can take. Humans might know something but forget it thats why AI for games always have a random input to them.
This next part is just my thoughts of what one would need to know or beable to find out for an AI to act "smart" in SEV
The Goal. -- Destroying all others -- geting to X tech level, ETC
The Techs and there Actions, -- for SEV Techs allow for new actions to take place. Knowing the techs allows for knowing the actions/reactions one can take. the more of these the AI or you know of the more actions and reactions you or the AI can defend for or use. Knowing how to use the techs is where the creator would need to tell the AI.
After those are given the rest is simple logic. And I dont feel logic is without emotion, because its only logical to include all effecting information when comming to an outcome. Logic is simply the path one takes to get to the location they are at, adding or ignoring information will change the place you end up and sadly both locations have a "logical path" for geting there.
(I dont remember now but there was a logical outcome to I-Robot, me and some friends had a discution on it, I simple forget what it was atm.
And good night all. If this post made no sence to you im sorry since this post is a resault of staying up for so long and being board.
Re: AI not challening?
WAAAAAY OT i know
The logical outcome in I-Robot was based on asimovs three laws of robotic :
1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey orders given to it by human beings, except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
the logical step the AI, which led the robotic research in I-Robot, took was
Taking Rule one ... or through inaction allow a human being to come to harm ..
recognizing that the being which hurts human most is humanity itself ..
deciding to protect humanity from itself by bending the second law a litte .. excet where such orders would conflict with the First Law .. and accepting that there will be casualties ..
Re: AI not challening?
Asimov's Robot books were a study in how robots got around the three laws. Ironically though, most robots were trying to be more human.
Re: AI not challening?
using higher bonus settings does increase difficulty, however most of the effect is early game, typicaly with ratings around 20,10 in bonus the computer is almost impossible to beat early game but if you have the time to move around for a year or two the advantage is leveling off to become far to low lategame
many people seem to have a strange idea of makes a decent ai, the ai needs not be very clever only patient, for example the ai could run a few thousand complete ship battle simulations every time it encountered a new design,to ensure that its ships use optimal strategies against its foes, this would mean that the ships the ai produces would always bee at the very least decent.
in a similar fashion the ai could, on a very fast system, randomize every aspect of its turns predict a number of turns ahead, and evaluate, this is exactly what a chess ai does and it works, every other aspect of what a chess ai typicaly does is simply optimizations to make it drop bad chains of "thought" more quickly.
there are many ways to make this faster perhaps even fast enough to be used before this game is played on winxp emulators ...
for instance one could give the ai a good basic gameplan and then allow it to run permutations of it in essence mutating the basic plan by trying small variations of it ...
thats what i basicaly do in pvp matches and ive never managed to deviate far from the basic plan without losing ...
also one could make several different types of ai to cut down on the various permutations a eco-ai a ship-ai a dip-ai etc ...
still with the 1.66 patch the ai is decent on maximum difficulty settings starting from the beginning assuming one does not exploit its weeknesses to much like diplomacy etc
it never gives me a challange but it would had i just begun playing
still one unfourtunate side effect of the overall game design is the crushing effect of figthers and drones, unless limited by player imposed rules fighters will be the wepon of choise a good ai would exploit this but we are far from there eh ...
Re: AI not challening?
I designed a mod based on Halo (just techs and comps and facilities, i cant design much 3d stuff without using GMAX) The AI is INCREDIBLY DIFFICULT!!!! I made one enemy empire with super advanced weaps another with armor another with sensors etc. etc. I made it so all AI worked together and BAM all enemy empires have insane techs and within 4 years each of the 9 have well over 300 ships each (not counting non combat hehe!!!) The AI is very slow in the beginning but very powerful once it establishes itself. I have had battles with my entire fleet of around 40 frigates, 14 destroyers 5 cruisers 4 light carriers (each with a standard garrison of 100 soldiers and 20 light fighters(anti-fighter craft) 3 heavy cariers(each with 120 sooldiers and 40 medium fighters) and 2 baseships. The construct speeds are doubled forplayer and quadrupled for AI.) The enemy engaged me with a fleet of 30 OLD ships equipped with weak shields and weapons. the enemy ACTUALLY USED STRATEGY!!!!!!!!!!!!!!!!!!!!! they drew my main fleet out and obliterated my flagship and base ship, crippling my fleets resupply capabilities. Their fighters, (MUCH better than mine) obliterated 2/3 of my fighters and when i recoverd enough to attack them they retreated bringing in another 20 ships from BEHIND after they retreated from the battle, these ships destroyed the dozen space stations and 3 starbases the fleet had been defending forcing my LARGER SUPERIOR fleet into retreat. And whilke i had less casulties in the end THEY won! an amonkrie fleet attacked that fleet the next turn, destroying 1/3 of the fleet and then retreating and closing the jump point behind them! But that! isnt the worst this mod has done to me!!! At my MOST POWERFUK military base, a captured neutral homeworld surrounded by over 20 space stations ad a dozen frigates (patrol boats built around 2 fast missles(archer missles) and a MAC gun(slow to reload extremely damaging projectile.) 4 waves of 40 ships eah ambushed me there after the bulk of my fleet left to defend the only jump hole to an enemy ocntrolled system (this is on a spiral map witah only two holes per sys, my personl favorite.) The AI then jumped in sys wit suprisingly advanced ships, just a turn earlier i thought they only had level 23 guns but they attacked with level 60 subspace rupture beams. Of the 24 ships and and 20 space stations only 3 ships escaped with the planet captured and the space sttions destroyed! the only escaping military ship was a captured heavy carrier i was planning to analyze. most of my fleet was cut off from the home stars and after they secured the system from me the covies brought inm over 200 ships in one jump!!!
Now thats a challenging AI!!! after i have edited a few things ill post the mod online, the main ting stopping me is slow development ni the beginning of the game. Note: The mod DOES work with the stock AI and comp generaed AI
if anyone wants me to upload it let me know otherwise id rather not till its done
What does this have to do with AI? it proves an AI can bne improved and enhanced through good modding.
Re: AI not challening?
Re: AI not challening?
And a repetition of Kwoks first suggestion ^^
But still it sounds like a fun game
Remember : Pillage! then burn.
- Cpt. Kaff Tagon
Try this
Create all the main empires and set the computer controled ones to have beter starting positions than you do. And limit your self by not using these advantages