Actually, there is a game where the AI can match and exceed the best that humanity has to offer: Chess.



The thing about Chess is that the rules of the game are relatively simple to comprehend, which means that when an AI has to interpret those rules it would be able to map out possible outcomes very easily. In this case, the game mechanics are ideal for allowing an AI to be 'smart', because there are comparatively few variables. This is why when a videogame is designed, the mechanics should be easy to understand: It doesn't just allow the game to be approachable for human players, but it also makes it easier to optimize the AI. Unfortunately, game mechanics tend to be interlocking in nature, to the point where if you change one thing it is likely to affect a completely different kind of clockwork elsewhere. Tweak military AI? Then economics and diplomacy are affected. Change those to accommodate the tweaks? That reverberates back into the military AI. Still, the simpler the mechanics are, the easier it would be to make something effective.



Anyways, I would recommend that the developers abandon the current difficulty system in Endless Space and polish up the AI. Right now, the various difficulty levels are an ineffective crutch, because they artificially cripple or inflate the AI's ability to do anything. Instead, difficulty level should affect the amount of assets that participants in the game start with, and from there the AI plays normally. This would force the developers to make the AI competitive, which would make the game more appealing to a human player. Humans don't like it when their opponent is obviously cheating, because it not only feels unfair, but it also implies that opponent is incompetent. Trying to best someone who doesn't deserve their advantages feels unsatisfying, because they are not capable of challenging the player in the first place.





Newbie - Players have one planet, four colony ships, five destroyers.

Easy - Players have one planet, two colony ships, three destroyers.

Normal - Everyone has one planet.

Hard - The AI has one planet, two colony ships, and three destroyers.

Endless - The AI has one planet, four colony ships, and five destroyers.

After playing a couple of games with The Harmony i`v found three things i believe is key to success in early/mid game. The tech you need you`ll find in the Galactic Warfare, Diplomacy & Trading and Exploration and Expansion trees. Keep research in the Applied Sciences tree to a minimum.



1. Harmonize Planets: With The Harmony, start out with focus on food production, fill up planets pop cap and go for the Harmonize Planets in the Diplomacy and Trading tech tree. It is extremely important to manually adjust your planet population in all systems so you always keep so many planets as possible pop capped and get the +50% FIDS bonus.



2. Applied Casimir/Atmospherics: Get only a few systems with as many planets as possible in the start (max 3 systems). This should not be exceeded before researching Applied Casimir Effect (Colonization program). You can then further expand with a couple of systems after getting the Applied Atmospherics (Orbital Counseling). This two will combined give you a 2 x -22% expansion disapproval.



3. Contamination Barriers: Get this tech in the Galactic Warfare tree as soon as possible. This will remove all Dust penalties in all system where its built and suddenly you can compete with the other factions

Nikolay_PI wrote:
How it is ? I am do not understand. Or "3-4" Or "pump".



Game flow for this approach:



turn 1 - Home World

turn 3 - 2 system

turn ~10 - ~10 pop and 3-4 system

turn 20 - ~20 pop and 3-4 system, ~2 destroyer first contact with AI

turn 30 - ~20 pop and 3-4 system (ship building block pop), ~10 destroyer (5 ship on 2 wormholes), first AI attack (with out war)

turn 40 - ~30 pop and 3-4 system, AI declare war for you (in AI eyes you pretty weak)

After this AI spam you with fleets and you can not replace you loses with same speed. Just by small pop. You do not have any working race bonus at this time and destroyer\cruiser has useless traits.



At least, at Endless level so.



P.S. With United Empire i am have 10 system and ~60 pop at same time and can struggle.

Harmony is nearly unplayable above Normal difficulty anyway. So arguing "That wouldn't work on endless" doesn't really do much.

Autocthon wrote:
If you're getting your butt handed to you on newbie the simple fact is that you're not all that great,

If a person is not that great, shouldn't newbie be a level where they can still do well, compete, and have fun without getting crushed? The game is very poorly balanced. Why do some people find this so hard to accept?



Take a look at Civ V. This game is exquisitely well balanced. On level 1 (chieftain I believe) it's almost impossible to lose and a great way for a newbie to learn the game. The difficulty moves up in reasonable steps with none being too much of a jump from the other.



I played endless space on normal and then moved down to newbie and found there to be no discernible difference in difficulty. Both were hard.



This issue is something the Devs should really look at addressing right away. Otherwise, they are scaring away potential players and buyers of expansions. I for one have stopped playing for now in hopes that the balance issue will be fixed. I certainly won't be buying any expansions for a game I don't enjoy due to the difficulty.

Looking at the AI on Endless, it is apparent to me that it is moronic at best. The only reason the AI would ever do well is because of the resources that allows it to witlessly pump out ships. Having reduced all of the AI's worlds to a single one for each empire and placing a full fleet of dreadnoughts over every world, I got to watch their development for a bit.



1) Some of the AI had systems that weren't completely inhabited, but didn't colonize any planets after 50+ turns. Some of them were also broke.

2) Removing my blockade of their system and taking away my defenders from a neighboring one, the AI didn't try to take over that system. Instead, they threw 3000 MP fleets against my 25,000+ MP fleets, which were beyond the undefended system.

3) Went ahead and razed that undefended system. The AI didn't colonize it, and still attacked my defended worlds relentlessly.

4) Loaded up a savegame, and then gave tremendously good peace deals to the AIs. Then made Open Borders and Cooperation agreements that were in their favor. About 20 turns afterwards, they declared war on me, promptly losing the 24+ resources they were receiving from me. They were able to reach other empires that were roughly equal to them, yet they still insisted on fighting me.

5) Another savefile, and gave every AI 4 worlds in addition to the resource agreements. They still declare war on me, and I noticed that they didn't bother to improve their undeveloped systems, try to conquer each other, or do any real preparation for a war with me.



The AI is a complete failure. Dumb and bloodthirsty.

Stargem wrote:




1) Some of the AI had systems that weren't completely inhabited, but didn't colonize any planets after 50+ turns. Some of them were also broke.

2) Removing my blockade of their system and taking away my defenders from a neighboring one, the AI didn't try to take over that system. Instead, they threw 3000 MP fleets against my 25,000+ MP fleets, which were beyond the undefended system.

3) Went ahead and razed that undefended system. The AI didn't colonize it, and still attacked my defended worlds relentlessly.

4) Loaded up a savegame, and then gave tremendously good peace deals to the AIs. Then made Open Borders and Cooperation agreements that were in their favor. About 20 turns afterwards, they declared war on me, promptly losing the 24+ resources they were receiving from me. They were able to reach other empires that were roughly equal to them, yet they still insisted on fighting me.

5) Another savefile, and gave every AI 4 worlds in addition to the resource agreements. They still declare war on me, and I noticed that they didn't bother to improve their undeveloped systems, try to conquer each other, or do any real preparation for a war with me.



The AI is a complete failure. Dumb and bloodthirsty.

I find it great the Devs work with the fans with regards to what feature to implement; however, they have done a poor job in continuing to tweak and improve the existing features.



I bring up Civ V a lot, not that I think the Devs there are great, they are not, they couldn't care less about the community, but they have done a great job in improving the existing game. They've balanced features that needed it, removed ones that didn't, and added new ones that were missing. They did all this while working on expansions.



The Devs for this game have to work that type of balance. For now it appears they have focused on the game breaking bugs and releasing expansions but have forgotten to tweak it and polish the game.

ThatMG wrote:
I have gone back to classic as of now, sad though because I liked the Harmony race idea.

I have the same problem in the beginning but I think it's because of the Harmony.



The Harmony doesn't have most of the advantages we are familiar with, and it has some strange rules quite different from other factions.



It forces us to adopt the whole new method to win.



And that could be a challenge even for a seasoned player.



After trying the Harmony, I played other factions in Disharmony and Beta.



And there is no difference, you can surely win on Endless difficulty if you have plenty of experience in Classic.



So the problem is simple. The Harmony is difficult, not Disharmony.

Stargem wrote:
Looking at the AI on Endless, it is apparent to me that it is moronic at best. The only reason the AI would ever do well is because of the resources that allows it to witlessly pump out ships.



The AI is a complete failure. Dumb and bloodthirsty.

I have to agree with part of what you said.



But I think that's because the devs don't have sufficient time and resources to enhance that.



And why? I am trying not to be too mean on this. I agree with what Autocthon said.

Autocthon wrote:
If you're getting your butt handed to you on newbie the simple fact is that you're not all that great, and therefore are very unlikely to have done any for of extensive testing, unless you're playing Harmony, then it's because Harmony is nearly brokenly weak right now and requires a massive amount of experience to keep up with the AI.

Pointing the devs in wrong direction is not helpful, and that will make the devs have no time for something really important like more diversity or challenges.



So the devs don't have time for what you and I really care about, and that's unfair to blame them.



And about the AI thing, I have a different perspective.



The AIs do have some issues even on Endless difficulty.



They manage heroes not as well as the players do.



And they don't even actively train heroes with every tiny opportunity like how we do.



Their heroes don't have effective battle actions and the AIs don't know how to use the right ones in the battle.



That makes their fleets inefficient. They have to maintain huge fleets and it drains their economy.



The poor economy affects their speed to expand and research.



And even worse, they are like newbies who are not familiar with new mechanisms.



They seldom capture the star systems instantly, bomb or do any new tricks we are good at.



That makes huge fleets the only solution.



That's why they build the huge siege fleets what many players complaint about, and make those silly decisions.



I think that's only because they just don't get enough love from the devs who are busy dealing with the players.



If the devs do teach them how to be meaner and tougher, they can just capture our star systems and raze them, being anything but the dumb opponents we are facing now.

cgboy2003 wrote:
The AIs do have some issues even on Endless difficulty.



They manage heroes not as well as the players do.



And they don't even actively train heroes with every tiny opportunity like us.



Their heroes don't have effective battle actions and the AIs don't know how to use the right ones in the battle.



That makes their fleets inefficient. They have to maintain huge fleets and it drains their economy.



The poor economy affects their speed to expand and research.



And even worse, they are like newbies who are not familiar with new mechanisms.



They seldom capture the star systems instantly, bomb or do any new tricks we are good at.



That makes huge fleets the only solution.



That's why they build the huge siege fleets like what many players complaint about, and make those silly decisions.



I think that's only because they just don't get enough love from the devs who are busy dealing with the players.



If the devs do teach them how to be meaner and tougher, they can just capture our star systems and raze them, being anything but the dumb opponents we are facing now.

You're a perfect candidate for MP games. These problems with the A.I. don't exist when you play against human players.

Nasarog wrote:
You're a perfect candidate for MP games. These problems with the A.I. don't exist when you play against human players.

Hmm, you can see that I am just explaining what the problem is.



Neither Disharmony, nor the AIs. Not even the mechanisms. The mechanisms are FINE.



I am not the ones who are complaining, and I do know what MP is like.

cgboy2003 wrote:
Hmm, you can see that I am just explaining what the problem is.



Neither Disharmony, nor the AIs. Not even the mechanisms. The mechanisms are FINE.



I am not the ones who are complaining, and I do know what MP is like.

Oh indeed, I do. I doubt an A.I. exists that can give such a challenge when compared to a human player. If the game is properly balanced, I think a human player would be your perfect opponent. Otherwise I agree with you. I was just trying to offer a helping hand to tide you over until the bugs are worked out.

Nasarog wrote:
Oh indeed, I do. I doubt an A.I. exists that can give such a challenge when compared to a human player. If the game is properly balanced, I think a human player would be your perfect opponent. Otherwise I agree with you. I was just trying to offer a helping hand to tide you over until the bugs are worked out.

I added some quotes to my earlier post and tried to make my opinions more clear.



Please read it again, and I think there is some misunderstanding.



I was trying to explain why the AIs can't be what we expect because of something else, and it's not the devs' fault.



So please stop saying the MP things to me, that will make others confused and think I was complaining about the devs.



I was NOT.



I don't think there is any problem and I just enjoy the game very much. You probably mistake me with someone else in this thread.

I routinely set my slider to +100% science once I pop-cap my first 3-4 planets

I have 3-4 planets at roughly double pop early in the game.

as long as I pump out a blockade on wormholes

How it is ? I am do not understand. Or "3-4" Or "pump".



Game flow for this approach:



turn 1 - Home World

turn 3 - 2 system

turn ~10 - ~10 pop and 3-4 system

turn 20 - ~20 pop and 3-4 system, ~2 destroyer first contact with AI

turn 30 - ~20 pop and 3-4 system (ship building block pop), ~10 destroyer (5 ship on 2 wormholes), first AI attack (with out war)

turn 40 - ~30 pop and 3-4 system, AI declare war for you (in AI eyes you pretty weak)

After this AI spam you with fleets and you can not replace you loses with same speed. Just by small pop. You do not have any working race bonus at this time and destroyer\cruiser has useless traits.



At least, at Endless level so.



P.S. With United Empire i am have 10 system and ~60 pop at same time and can struggle.

Stargem wrote:
Actually, there is a game where the AI can match and exceed the best that humanity has to offer: Chess.

Yes, but the choices are limited when compared to a 4x game.

Stargem wrote:
Actually, there is a game where the AI can match and exceed the best that humanity has to offer: Chess.

Thoughtful elaborations.



Endless Space sometimes makes me feel like playing chess.



But it has more bones and flesh, and it's more entertaining.



Like what I mentioned above, the devs care too much about the tiny little opinions from the players, which is basically good, and makes this forum and all the g2g stuff.



But if you are just too worried about "all kinds" of opinions, nonsense or not, that can just limit your creativity and shackle many things like AI.



You can see that I am trying to encourage these guys to do what they should do, and hope they remember who appreciates all the effort and fine work they make will eventually appreciate it.

Biggest problem I have had in my 4 games so far playing disharmony is the ridiculous amount of unarmed invasion fleets the AI uses. Fleets composed entirely of AI ships with only invasion weapons dont even try retreating. They just are canon fodder for my admirals and repair buffs for my fleet using the battle damage cards.



I've read how classic allowed any fleet to capture, even if they didnt have siege weapons. Thats the way it needs to be. Invasion weapons should reduce the time needed to capture, not be the only way to do it. Also, because of how EXTREMELY POOR the AI handles their invasion fleets, i.e., flying an entire fleet without usinmg any combat capable ships for cover is RIDICULOUS in the extreme and needs to be changed. Either the AI needs to be reprogrammed NOT to fly entire fleets of ONLY invasion or Invasion weapons need to be dual purpose, as in they need to also be able to be used in combat, but be only half as effective tonnage wise as combat weapons. Maybe even 1/3 as effective... but more than half and less than a third would be too much and too little.

Poor A.I should not get a free pass.The A.I fleet building and decision making can always be improved with continues work.

Stargem wrote:


The thing about Chess is that the rules of the game are relatively simple to comprehend, which means that when an AI has to interpret those rules it would be able to map out possible outcomes very easily. In this case, the game mechanics are ideal for allowing an AI to be 'smart', because there are comparatively few variables.

I think it's less a matter of how "simple to comprehend" the rules are but more about how much effort has been put into it.

I think dozens of programmers had put years into the development of a chess-programs before those started to be competative with really good players.



Chess computers, aren't even close to mapping every possible outcome for every possibility. The small fraction of possibilities they are able to look ahead is just a lot bigger than the very small fraction that humans do.

There's still massive potential for improvement that the authors of Chess-Engines like Houdini, Komodo or Stockfish are trying to explore while fighting for the crown of the strongest chess engine. (years of doing nothing but trying to find further improvements in an AI, that no human can beat in order to make it beat the AI of another programmer/team!)



I still think that a lot could be learnt from how chess-programs work and adapt it to other turn-based games like ES.

The key is not just do what a very simple algorithm tells you, but instead look at the result of it several turns later.



Like: I wouldn't assign "AIValue" to buildings. I would make the AI simulate the effect of that building. "How would building it affect my FIDS in 1, 3, 5, 10 turns?"



On the tactical level, not knowing where your opponents units are, simplyfies it dramatically when compared to chess. You don't have to take all that into consideration because you cannot know it anyways.



But it would be a good start if the AI would stop doing stuff that is obviously a blunder. Sending in undefended Invasion-Fleets for example. That's like a chess program putting its undefended pieces on squares you control. But they are very, very far beyond that kind of stuff. Their "mistakes" are more of the type "I made a move that accomplishes too little and in 5 turns could result in me not being able to cover my piece without a pawn-move that weakens my pawn structure which will be a disadvantage in 30 turns because then the opponents king can reach the center slightly faster and then block of my king from protecting my pawns efficiently which results in him queening one of his pawns 1 turn earlier forcing me to sacrifice my remaining knight for it."

Ail wrote:
I think it's less a matter of how "simple to comprehend" the rules are but more about how much effort has been put into it.

I think dozens of programmers had put years into the development of a chess-programs before those started to be competative with really good players.

Programmers also mainly learned exploit better the centuries of strategical studies of Chess. And they learned exploit cumulative stats of chess databases. That makes them stronger in any area from opening, tactical, strategical and even game ending through the use of ending huge databases.

Ail wrote:


Chess computers, aren't even close to mapping every possible outcome for every possibility. The small fraction of possibilities they are able to look ahead is just a lot bigger than the very small fraction that humans do.

Yep, the improvements they made is evaluating positions and this is by learning better to include in programs the centuries of chess studies and the databases for anything, from opening to ending.

Ail wrote:


There's still massive potential for improvement that the authors of Chess-Engines like Houdini, Komodo or Stockfish are trying to explore while fighting for the crown of the strongest chess engine. (years of doing nothing but trying to find further improvements in an AI, that no human can beat in order to make it beat the AI of another programmer/team!)

Yeah but at this point you wonder who is interested anymore. I don't think that own more win titles for a chess program will give it much more sells. Well apart for team plays human+program.

Ail wrote:


I still think that a lot could be learnt from how chess-programs work and adapt it to other turn-based games like ES.

The key is not just do what a very simple algorithm tells you, but instead look at the result of it several turns later.

I don't think so, the key is evaluating positions. And that's why games will never reach such level of efficiency they'll never had behind centuries of studies. Nor dabatases of millions of pro games and databases of other millions of other stats.



EDIT: Mmm Ail if you was meaning evaluating position some moves ahead, then sorry it wasn't clear for me.

dizzy_xc wrote:
Biggest problem I have had in my 4 games so far playing disharmony is the ridiculous amount of unarmed invasion fleets the AI uses....

That and how stupidly the AI doesn't use fleet regroup and so often do vain pointless no hope attacks.