Re: [TuxKart-devel] Ideas for AI

[Steve B. <sjb...@ai...>]

Oliver Jeeves wrote:
> On Tue, Jun 29, 2004 at 06:11:22PM +0000, Charles Goodwin wrote:
> 
>>I also don't believe we should make each driver in the race be of
>>increased difficulty so it is hard to get to #1.  Some people aren't
>>game wizards and they'll find it frustrating if they can't win on easy.
> 
> 
> I'm not exactly sure what you mean here,

He's saying that you have some number of AI players - graded from
easy to beat to very hard to beat.  Then, if you are a beginner, you'll
come in close to last - and as you get better, you'll come in higher
and higher up the ratings until *FINALLY* you can win a race.

If I've interpreted it right - I think this sucks as a game concept.

Probably I don't understand either!

 > but I think there should be an
> approach similar to Mario Kart - where the AI ahead of you becomes weaker, 
> and the AI behind you becomes stronger (or maybe just adjust the speed).

Yeah - that's what Tuxkart's AI does now - and people who figure it out
*hate* it.  No matter how hard or well you drive, you just can't get far
enough ahead to guarantee a victory.  The way to win is to drive around badly
(it doesn't matter - the AI will slow down to let you keep up) - then just
before the finish line, work hard - and before the AI players notice, you'll
win almost every time.

Clearly that sucks.

I don't see the problem with doing this to a *mild* degree - because it does
make the game more interesting...but that can't be the mechanism for keeping
the game interesting.

In the end, the tried-and-tested way is to have some number of levels of
difficulty.   Since we don't want to design too many characters (it's a LOT
of work), it makes sense to allow each character to play at different skill
levels and with differently powerful kart physics so they play at the skill
level the player asks for.

> But how do you define good choices? Is it better to take the corner in the 
> fastest possible way, or to take it in a slower way that leads to collecting 
> a powerup?

That's what puts the 'I' in AI!

Probably what you do is to build these preferences into the AI's "Character".

You know - the suave sophisticated Penguin drives as best he can, the maniac
Octopus goes for the powerups every time.  The other guys pick randomly with
variously skewed probabilities.

This possibly leads to 'richer' game play.  If you know the Octopus is behind
you, maybe you make more effort to nab the goodies before he does.  If you
know the Penguin is going to ignore them, maybe you take the best racing line
and drop a bananaskin there because you guess he'll go that way.

> Basically, what you're saying is that we need to define some heuristics to 
> determine what actions to take. It's not as simple as good/bad/normal 
> decisions. For example, taking a slower route to collect a powerup makes less 
> sense if you are in the lead.

Yep.

---------------------------- Steve Baker -------------------------
HomeEmail: <sjb...@ai...>    WorkEmail: <sj...@li...>
HomePage : http://www.sjbaker.org
Projects : http://plib.sf.net    http://tuxaqfh.sf.net
            http://tuxkart.sf.net http://prettypoly.sf.net
-----BEGIN GEEK CODE BLOCK-----
GCS d-- s:+ a+ C++++$ UL+++$ P--- L++++$ E--- W+++ N o+ K? w--- !O M-
V-- PS++ PE- Y-- PGP-- t+ 5 X R+++ tv b++ DI++ D G+ e++ h--(-) r+++ y++++
-----END GEEK CODE BLOCK-----