So poor a strategy fight ?

The AI is not too smart.  ZombieRus and some others are actually working right now to make it play smarter. 

You will want to try Mods--Distant Stars and Maelstrom have tougher units in them as do other mods. 

Playing online is best with other players unless your game is laggy. You might ask here for European players to identify themselves.

Play on the hardest difficulty settings with pirates on--might make for more of a challenge--place the AI's on the same team against you as well.

Yup, the tutorials suck and the learning curve is steep.  This is the only place where the game really dropped the ball.  As much as we may rant about the game's shortcomings, it's still top-notch in most respects and even its weakest points still compare favourably against most RTS games out there.  The tutorial is the one place where it really failed.

Nope, the AI is a fairly straightforward opponent that is easy to manipulate once you understand its general logic.  As a general rule of thumb, the AI is incapable of countering large swarms (~100 squadrons) of bombers, and has immense difficulty breaking through upgraded starbases.  When comp-stomping, I find the most enjoyable way to play is to speed-run, forcing you to pull every ace you can to win quickly. 

No AI will ever have the level of intricate strategic planning a human is capable of, so at a certain point if you want to play against a truly adaptive and intelligence opponent you will need to go online.  Sins is a multiplayer gem for all its shortcomings.

Guess you're just getting out of bed and I'm just getting into bed.

Regards from the other side of the world

@ Darvin3 : You're right ! Have a good night !

I know that the best, as it is in most strategy games - Civilization, Game of Thrones, Shogun II for exemple, is to play against people. We are far from having smart AIs that could be of some challenge to humans, unless they are granted a certain amount of cheat - which is not very funny. That is in fact weird enought, since computers are beating humans at chess, and have a pretty good level now at very difficult games like go...

The problem is that what I enjoy in that kind of game is to think, to plan, to find out the best and the shortest way to develop my civilization, to discover the correct balances between economy, resaerch, fight.. By the way, I am poorly gifted with reflex and reactivity. That is why I spend much more time with the game paused than with the game running... I'd be a very bad and boring opponent for most players !

Thank you very much for your answers.

Part of the problem is that the AI doesn't spend its resources very well. Thus at higher AI levels you may see it use a wider variety of ships and tactics, simply because it can now afford to build more different types of ships. But yes, Sins is at its core an RTS, infinitely more complex than chess. It will take quite a while for game AI to catch up to its possible depth.

Why not have an Ai that spams LRM with carrier caps and avoids suicidal tendencies on star-bases? Easy to program if you ask me..

you go do that and get back to us, riddleking.

Actually, it wouldn't be too difficult to change the aggressor to spam LRF instead of LF like it currently does, which would significantly increase the difficulty.  "Not suiciding on starbases", though, would be a lot more challenging.

Chess is a discrete game by its nature, meaning it's easier to break down your options algorithmically.  In a RTS game, there are so many possible factors to track that you need a fairly sophisticated algorithm just to identify which factors are relevant, much less to actually determine your best course of action and how it fits into a long-term strategy.

More important is multi-tasking.  I have an abysmal APM (at least compared to most active competitive RTS players) and still maintained diamond league in Starcraft II while I was active.  Unless you're aiming for E-Sport level competition or something you don't need crazy reflexes to play RTS respectably.

Well, I remain convinced that game developers – most of them, if not all of them – do not spend much money and time to improve their AI. They bet the game will meet its market on the multiplayer field.

I agree with GoaFan : a strategic game, RTS or not, can be much more complex than chess. In fact, chess just needs computing power in order to explore a tree of possible moves deeper than a human can explore it. That’s almost all.

But that is not at all the case for go, which is a very good example of great improvements in AI.

Go has 10^600 times more possible games than chess. That means that you need a computer 10^600 more powerful to beat humans, which is, for the moment and for a very long time, impossible.

That is why go robot developers cannot solve easily the problem with computing power. They have to face a very complex game, much more complex than chess, and much more complex than any RTS game. Good go players have to show imagination, courage, trickiness, ability to take risks, experience, ability to decide, capacity to analyze a situation from different points of view and, above all, an immense knowledge of go concepts, situations, vocabulary and tools. And there are almost always several possible good moves, each of them depending on your style of playing, your sensitivity, your character as well as your opponent’s.

After five years playing go, reading dozens of books, learning, being taught by pretty good champions, I have great difficulties to beat the best robots – not to say they beat me . And research on the subject is at its beginning.

Why is it so difficult to beat go bots? Because unlike RTS games, the game interface is very simple and all the effort has been applied on the AI development.

That is why I remained convinced that, unfortunately, strategy games could be very interesting against computers, but they are not…

PS : English is not my mother tongue, forgive me if I am not clear enough !

A potential problem is that you're assuming that brute-force (analyzing all possible movement options, in Go or Sins) would be needed to create a very strong AI. Although I'm not familiar with Go, at least with Sins it's possible to create well-defined strategies that an AI should be able to carry out as well (reasonably easily, in theory at least). If, of course, the AI is moddable enough, which at the moment it is not. And there aren't actually that many strategic options, even at the upper levels of play, I think - sure, there are dozens and dozens overall, but for any given situation there are probably only a couple or 3 or 4 worth considering, and I would think that an AI would be able to weigh them and come to an optimal decision.

A computer would also be able to queue up the proper units to counter enemy fleets more efficient than any human player, probably... and optimal in-combat micro isn't very complicated either, probably something a computer can do pretty well also. Broad scanning of the battlefield and making conclusions about what to move where could probably be done somewhat better by an experienced player than a computer, but making the computer relatively skilled at it shouldn't be too hard.

Making an AI very skilled at macro though would be difficult though, I think. Not impossible, but it would take a decent amount of work. I think. Another thing that plays into this is expectations about player skill (and subsequent build decisions), which a human can do pretty well but a computer can't.

So although there are a virtually infinite number of possibilities for a Sins game, I still think making a very very strong AI would be viable, if the AI can be given the appropriate amount of CPU time and if the necessary variables and commands are at our disposal.

But unfortunately, the devs have not opened the AI up to any significant modding... and the AIs are still stuck on the level of ramming their hapless fleets into starbases to die. Modding a normal Sins AI to the level of one of the top-tier players (or better) would be my dream, but we don't have the tools at our disposal to do that...