Re: Turn Based Games

Raghar a écrit :

> Richard Delorme <[EMAIL PROTECTED]> wrote in
> news:[EMAIL PROTECTED]:
> 
>> Raghar a écrit :
>> 
>>> I very don't recomend to create checkers, or othelo becose it's
>>> somewhat easy to come with best tactic so it'd become boring quickly.
>>> Especially for AI creation. You wouldn't want to teach just tree
>>> searching.
>> 
>> I would like to know what is the best tactic at Othello. Strong
>> Othello bots actually play well mostly because of their perfect
>> endgame, as the last 27 moves are usually played perfectly by the
>> fastest programs. Most of the recent tricks to enhance the strength of
>> the Othello bots consist in very selective searches to obtain a
>> probable endgame score for the "best" move as early as possible
>> (usually when 40-35 moves remain to play), just to avoid the
>> hazardously weak midgame tactic. So I am quite interesting of that
>> best tactic, as I am only aware of weak ones, that just works because
>> of CPU-speed.
>> 
> 
> 
> I actually ment that 27 last moves that are played perfectly. It destroys
> all interesting from ending part of the game.
> What about tactic get corners and count properly? Seems to work well.

Without using mobility, this will lead to a very weak opponent, much weaker
than any human knowing how to play, even with a deep tree search. Othello
is much more subtle than it seems and is well described by the words "easy
to learn but difficult to master". Strong Othello bots use a pattern-based
evaluation function, with learning algorithms to best-fit the function from
millions of positions. Unfortunately, these strong bots are still weak
without tree-searching, obviously weaker than any skilled human player.
 
> I didn't spend too much time in doing research on Othello like games and
> what I did was rather my memory long ago. I actually like more games that
> could change rules even at the end of the game and possibly turn nearly
> win in very likely lose.

All games have their interest.

-- 
Richard