Slitherine

An artificial intelligence?

I've tried countless games and keep seeing the same thing an ai that's dumber than dirt and uses no tactfulness at all. Even on pure strategy maps it's awful.

Why are there not programmers like the past that could program an ai to actually play better?

I see they make difficulty levels but all they do there are increase or decrease the numbers thereby just creating handicaps. Taking away the fog of war for the ai thus creating cheats.

Many people say they like the Combat mission series but have you really ever tested the ai out in that game. It's so stupid my dog can beat it with his hind leg. lol
Just run up the sides in any map and feint a frontal assault and you can flank it over n over to a simple victory.

Is there one decent squad or platoon or even company level game out there with a decent ai that knows how to play or only knows handicaps by the numbers and cheats like no fog of war and dice roll advantages like Heroes of Normandie uses imho right now.

There are various reasons
* AI programmers are not necessarily any good at the game. Its not possible to make a good AI unless you know what it should be doing.
* Ai is extremely difficult to do. Every time you get one thing working well it tends to compromise another area of the AI. Decisions are so interelated and so complex its very hard to define them in a completely logical way.
* Every action the AI takes has an opportunity cost. Evaluating what the best move is is so complex its hard to explain. The human brain discards billions of possibilities out of hand. AI cant - it has to evaluate all of them.
* How do you evaluate a situation? Just coming up with a system to evaluate a situation extremely complex - almost impossible. If the AI had a perfect way to assess the situation it woudl then be faced with another huge problem - what to do about it!
* Every time you change a lone of code it affects the AI. Even changing some numbers on a data file can require the AI to be rewritten. If you had a game that was complete without a single change and then you made the AI you'd have more chance of getting a good one. This is never possible for obvious reasons.

Chess is a very simple game and after decades of attempting to beat a human brain it can't. You make the game even slightly more complex and you magnify the problem. Each tiny piece of complexity adds vast issues for the programmer and performance. The games we make are vastly more complex than chess. Making a real AI for them is currently not possible. Maybe it will be one day. Now we have to set up interesting situations for the player rather than pitch them in to an equal fight.

AI is good at multi tasking so the only time it really gives a player an equal challenge is in games where the player has too many things to think about at once and they get overwhelmed by the volume of decisions an AI can make, not the quality of those decisions. E.g. an RTS could be like that.

I've been PC wargaming for 36 years and despite playing many wargame titles over the years I've never found one that's made me think "Wow, this AI is bloody good!"
Some games have got better AI than others, but even so, any reasonably skilful human should be able to beat any AI opponent in any game without much trouble.
I'm a PC dummy and know zilch about programming, but I've often thought that programmers should tell the AI what NOT to do.
For example in many games the AI will make stupid suicidal low-odds attacks, so surely programmers could programme a "don't do that" line into their programming?