It's only "unethical" if you tell them. The implied insult in "I did not try my best" is "you have not, really, beaten me". True or not, when humans say that, it is often considered poor sportsmanship. And, as Alexandre Vaillancourt points out, we routinely use this trick (let someone win and not tell them) in our daily lives.

The starcraft campaign AI is completely non-competetive and nobody noticed until they disassembled it.

The AI can only do what it is programmed to do; by definition, every AI plays to the best of its ability. Whether you enjoy playing against any given AI provided by a game is another question.

If we really didn't create non-competitive AIs not a single chess AI would be enjoyable for anybody but grandmasters, because you'd have not the slightest chance of winning or understanding what was going on. The same is true for a larger and larger number of games - particularly with board games we're pretty much at the end of the rope now that computers can even beat world class Go players. I can't think of a single boardgame a computer wouldn't be better than humans (if sufficient resources to programming an AI were invested obviously).

Ethics discussions is more for things like GTA or Kitten Cannon, since those might be mistaken for encouraging bad real-life behavior. At best I would phrase this question as "is it dishonest" and answer with yes, but in the player's interest.

@person27 I dunno. Ethics isn't a narrow field. I think both would have to do with ethics. I mean, we had a Socratic seminar on walking on the grass of someone else's lawn in my ethics class. Actually got pretty deep haha

@Voo I don't even thing grandmasters are able to get fair games in anymore: we have generated 140TB lookup tables that offer the best next move for every chess position with fewer than 8 pieces left on the board, and (strong) desktop machine chess engines are now rating in the 3400 range, about +500 points over the best human players.

GMs would likely have a hard time competing with Stockfish, even on a smartphone.

I recommend this Kotaku article about what happens when you make AI that actually is half decent and uses intelligent strategies. Turns out that most players aren't as smart as they think they are and are easily beaten by any half-intelligent tactics.

I just wan't to cite my former Lead Level Designer when I was scripting enemy behaviour (in a shooter game): "The AI is not meant to be perfect. The task of the AI is to give the player a great experience, a hard time and then die as spectacular as possible."

@Temporal I can still see them enjoying the games even if they have no chance of winning, I mean I do enjoy playing against stronger players - but yeah 500 points difference probably makes even that one moot.

@mr23ceec - I wholeheartedly disagree. The problem with opponent not doing their best is that he denies the challenge, regardless of the outcome. And this is seriously frustrating with lots of modern games - they are WAY to easy and thus get boring very quickly. Also, it's unethical to lie, even lie by omission. The opponent, human or AI, is always EXPECTED to do their best, so if he chooses to do otherwise and does not explicitly say that, he lies.

To paraphrase the prime minister of the UK; "Easy means Easy". If a player selects the lowest difficulty setting, they don't expect to have too much difficulty. They'll instead be enjoying other aspects of the game; story, atmospherics, graphics. I would shy away from a level called 'impossible to lose', unless your setting is named 'impossible to lose'.

Also note, sometimes it's not even possible to do a competitive AI, because that one would consume more computation power than the game's budget allows for. Also, even if you have a perfect AI, it's hard to scale it down if you need lower difficulty levels (after all, it moves perfectly!) and not all players may be on a level to be able to compete with it (after all, we are just imperfect humans).