Subject:
|
Re: Connect Four game issues
|
Newsgroups:
|
lugnet.org.ca.rtltoronto
|
Date:
|
Tue, 16 Sep 2003 19:03:21 GMT
|
Viewed:
|
688 times
|
| |
|
|
Quoting David Eaton <deaton@intdata.com>:
> In lugnet.org.ca.rtltoronto, Iain Hendry wrote:
> > David Eaton's post exposes my fears. The logic behind this sounds so • hard.
> > I don't understand how to program that kind of stuff... I supose the best • I
> > could ever hope to do is look at what's been placed, and calculate sort • of
> > a... hmm. I guess "What's the best move THIS time?" kidna thing. Look
> > for "if I drop one here, how many line up?" kinda thing for each of the 7
> > scenarios and then drop it. Crud.
>
> To give some reference for plausibility, a look-ahead will reveal (I
> think):
>
> Look-ahead | Max Possibilities
> ------------+-------------------
> 1 | 7
> 2 | 49
> 3 | 238
> 4 | 1,120
> 5 | 4,263
> 6 | 16,422
> 7 | 54,859
> 8 | 184,868
> 9 | 821,283
>
> So, you're entirely right. The only way to do this in the RCX world is to
> create
> an algorithm that analyzes the board and figures out the best move, not
> necessarily the move that will guarantee victory N moves down the road.
> Unless
> there's anyone who's an advanced game theorist who knows that it's possible
> to
> squeeze in the "perfect" algorithm, I don't think the fact that it's a
> "solved"
> game will factor into it. (Though I'd be happy to be proven wrong! I've
> been
> curious about the "perfect" algorithm for a while now)
>
> You might be able to pull off a small look-ahead algorithm in addition,
> that
> would effectively look, say, 4 moves in advance. That might fit into an RCX
> somehow. But I don't think you can get it to the point where you'd look up
> 42
> moves in advance. That's just too big for an RCX to handle. Heck, a
> look-ahead
> of 9 was soaking up 200+ megs of RAM when I was calculating the above
> numbers!
> (course you could reduce that by a factor of 4 pretty easily, since I was
> doing
> 1 byte per square, and theoretically you could do 2 bits per square)
Wouldn't you only have to look ahead one step at a time? The opponent only has
7 choices on any given turn. It's pretty much impossible to say like playing a
computer in chess, figure out how the game will turn out by the first opening
move.
I think it's hard enough trying to remember all of the current values for each
slot, and the standard rating system you had for the 7x7 isn't quite right,
since if you had 2 in a row, and the next would get you 3 in a row, in a lower
valued slot than a central slot - you would logically place at the 3 in a row
instead - but (if i'm understanding correctly) your design would put it at the
higher valued slot even if it didn't make for 3 in a row?
Should we take this to .geek?
|
|
Message is in Reply To:
 | | Re: Connect Four game issues
|
| (...) To give some reference for plausibility, a look-ahead will reveal (I think): Look-ahead | Max Possibilities ---...---+---...--- 1 | 7 2 | 49 3 | 238 4 | 1,120 5 | 4,263 6 | 16,422 7 | 54,859 8 | 184,868 9 | 821,283 So, you're entirely right. (...) (21 years ago, 16-Sep-03, to lugnet.org.ca.rtltoronto)
|
72 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
