Resource centre for ZX Spectrum games
      using Manic Miner and Jet Set Willy game engines

 

 

Previous message		Next message

 

 

Message: 2828

Author: filsoft_2000

Date: 01/05/2002

Subject: Re: Abstract JSW rooms / Psuedo-random JSW designer

 

> A (non-trivial) pseudo-random room-generator would have to be
> semi-automatic, i.e. it would require human post-editing.

I agree with you. There are many aspects that a human could do better
than a computer (aesthetics for one).

> Unfortunately the problem of deciding whether a given MM/JSW room is
> possible to pass is NP-complete, i.e. it would require exponential
> time to compute. On any given time-frame, there are six possible
> moves: do nothing, walk left, walk right, jump, jump left, jump

right.

> So in two time-frames there are 6^2 = 36 possible moves, in three
> time-frames there are 6^3 = 216, and so on.

What would be very interesting to see is a program that could study
existing MM/JSW games and create new ones based on the rules of
gameplay and design it has learned.

Such a program could theoretically create the most difficult games
ever, and leave Broadsoft behind in the most difficult Miner Willy
games stakes! :-)

I'm interested in self-modifying code. Imagine a MM/JSW game engine
that changes the difficulty (speed and number of guardians, item
location, etc.) according to how the player is progressing.

> It would take very strong heuristics to overcome such combinatorial
> explosion without loss of soundness (if the algorithm says it's
> possible to pass this room then it really is possible) and
> completeness (if the room is possible to pass then the algorithm
> says so). And yet it's amazing how a human expert can just look at a
> room and decide in seconds whether it is possible to pass!

Assuming there are no invisible blocks or teleporters, yes.

> But it is an interesting idea to poke random bytes into a room and
> massage the resulting chaos into order. I tried it once and came up
> with a tunnel-like room, but I didn't save it.

I've explored this idea with both MM and JSW but I haven't saved any
of the results.

> I'm sure the technique promises much more interesting rooms when you
> hit upon the right functions...

I haven't really explored pseudo-random guardian placement. All of my
experiments have been with platform and item placement. One idea I
may try is to have templates and/or patterns which the program can
use but this may result in formulaic room design, with none of the
random 'lego' design I have seen so far.

I quite like the pseudo-random JSW 48K title screen designer I wrote,
though. It often spews out pleasing garbage.

Phil

 

 

	Back to the Yahoo! Archive
Previous message	home page	Next message