Game Crash

deccan · #1 September 15th, 2004, 10:19 AM

Just my neophyte suggestion: wouldn't it be better to have it append a timestamp + process id instead of a random number?

reverend · #2 September 15th, 2004, 10:21 AM

dom16882 here as well. Clearly not as random as it should be.

Cainehill · #3 September 15th, 2004, 12:01 PM

With random numbers, of course it is possible for the same number to come up in a row, even on different computers.

After all - flip a coin 8 times, there's a one in 64 chance it comes up heads every time. Roll a die 3 times, there's a one in 256 chance that it'll come up a 6 each time.

In an infinite universe, it's not just possible, but a certainty that random numbers will come up the same across _all_ instances on a given day.

On the other paw - mine is dom16822. It's a sign from Dog!

Really does seem like it's using the same seed for the RNG, or not setting a seed at all. You should probably use the clock for generating the seed, or use deccan's suggestion of a timestamp and whatnot (on multiple user systems, I've used a combination of a date/timestamp plus a random number, testing for the existence of that already and incrementing the number if need be, just for paranoia's sake.)

PDF · #4 September 15th, 2004, 12:21 PM

Hooo, mine is also dom16822, every time I play, and on 2 different comps with different OSes (one W98SE, one NT).
Looks like the RNG used is in fact a "CNG" - constant number generator !

Johan K · #5 September 15th, 2004, 12:39 PM

Maybe the clock stands still in Windows? I get different random numbers in Linux.

Seems like I should take a look at it in Windows too.

Thufir · #6 September 15th, 2004, 01:19 PM

Quote:

Johan K said:
Maybe the clock stands still in Windows? I get different random numbers in Linux.

Seems like I should take a look at it in Windows too.

That sounds like a good idea. I was doing some tests on magic duel (on a Windows box) and saw some surprising results, so I was beginning to wonder whether I was getting a truly random sample. With an s6 vs s7, I saw the s6 win only 2/20, which conflicts with what we've been told of the formula for magic duel. Here's where I posted the results:
post

I'd really like to know whether loading from saved files resets the seed in a way that'll generate true random samples, so I can know if my testing is valid, or not.

Thufir · #7 September 15th, 2004, 01:22 PM

Quote:

Johan K said:
Maybe the clock stands still in Windows? I get different random numbers in Linux.

Seems like I should take a look at it in Windows too.

Oh, and btw, this may sound weird, but in the limited tests that I did, it wasn't as if the clock stood still, but rather that it moved in 10 minute chunks. I don't know if that's possible, but maybe if somehow the Last byte of the time value were right shifted, or something like that...

I know, sounds unlikely, but that would be consistent with the way the magic duel was behaving in my tests.