somebody <x### [at] ycom> wrote:
> I might be wrong but all chess games I used to play used analogous fake
> difficulty / cheap tricks too, in that instead of using a different,
> "smarter" engine, they use the same engine but allowed it longer time to
> calculate ahead (alternatively, they artifically limit the calculation
> time/level for lower difficulty levels).

  That's precisely fine-tuning the AI parameters to match the difficulty
level.

  There's not really such a thing as a "smart" engine. There's only number
crunching: Crunch more numbers and the program will be a stronger opponent.
Crunch less numbers and it will start making more mistakes and poorer moves
for the simple reason that it's not seeing far enough. Not completely unlike
human players: A beginner will see only one or maybe two moves forward, and
this for only a few pieces. A pro will see a dozen or more moves forward
(more for the relevant pieces, less for the irrelevant ones).

  There are a few chess engines out there which at easiest difficulty
levels will make deliberate mistakes. In other words, rather than blindly
choosing the best move it has seen so far, it deliberately at times chooses
a slightly worse one, just to lower the difficulty level. This is also
adjusting the AI (in a way that it sometimes ignores better moves as if it
had never even seen them).

> Coding multiple chess engines for
> different difficulty levels doesn't make any more economic sense than coding
> multiple AI's in action games.

  I never talked about multiple AIs. I talked about tuning the AI strength
according to the difficulty level.

-- 
                                                          - Warp

Post a reply to this message