what do you think about adding random errors in game AI to make it easier to beat th cpu?

[Nop90]

In 4ttude there is a strong AI algorithm, there are 15 rules, every rule has a difficulty level (from 0 to 10) and is applied only if the game level is the same or higher

At level 0 the cpu does only random moves, at level 1 blocks a row if three opponent marks,  or complete a row of 3 own marks if available. You can beat it setting a simple Atari (two rows with 3 marks).

From level 2 the cpu starts stopping two opponent marks  in a row and tries to complete a row with 2 of it's marks.

At level 2 the game starts to be very difficult because, despite the simple cpu strategy, an humann player can miss some almost winning rows, while the cup never misses one.

The good thing of this situation is that the player has to learn to check all the board and this makes him stronger, on the other hand the casual player can get bored by the starting difficulty.

To compensate this I'm considering to randomly introduce a 20% of mistakes in evaluating the rules of the same level of the game.

This meas  that at level 1 the game can miss to block 3 marks in a row one time out  of 5.

At level 2 it can miss to block 2 marks in a row 1 time of 5, but never misses to block a 3 in  row because it is a level 1 rule.

What about it. It's better to leave the cpu as strong as possible, or is it a good idea to introduce random mistakes,

Do you know if such an idea is used in other games?

[Fadest]

I am not really aware of AI algorithm, but I like the idea, it is simple and elegant to give some weakness to the AI.

But when it doesn't play the best move it should do, the AI should still make a "good" move, just like someone who did not see the trap and play his strategy. And not a random useless move or it would break the feeling of playing against a good opponent.

 

Maybe you could put this as an option (strong AI vs emotionnal AI :D)

Edited September 2, 2019 by Fadest

[Nop90]

Every evaluated move has a ranking, I will not over rank a bad move, but randomly under rank the best move so that a less good move (if exists) could be selected.

 

Thanks for your opinion, tonight I'll try to implement it.