Vicimus
Joined: Nov 16, 2005
|
Posted:
Jan 03, 2006 - 04:09 |
|
I have heard lots of chatter about the Java Random Number Generator rolling double's too often or too many 1's as lot as a whole heap of other nonsense. Having a degree in Software Development, and knowing the basics of how Java generates Random numbers, this sounded extremely unlikely to me. But it did seem I failed a lot of re-rolled GFIs too. Was it just my imagination or was there really a problem? I wrote a quick program in Java using the Math object's Random() method (just generating a 1000 random numbers and then printing the amount of consecutive doubles etc). There was a slight sway but it was no larg enough to be statisticly significant (i.e. it was too small to conclude anything) AND on top of that it suggested the opposite of what people here seem to say (that there is a problem with rolling doubles too often).
Okay, so maybe the Fumbbl program uses some other strange method in order to get syncronised rolls. So I loaded up Fumbbl and used GFI with every player on every move for a couple of games against myself and counted the stats. The results are exactly as what would be expected. Out of 72 GFIs that failed on the first move, 12 also failed with the reroll. That's exactly 1 in 6 ... showing no problem whatsoever.
Conclusion is that *IF* there is so problem with the Java Random Number Generator in Fumbbl, the problem is negligable.
If you don't believe me I'm sure there are plenty of software students or just people with too much time on their hands around here that can recreate these results. |
|
|
SolomonKane
Joined: Jul 21, 2005
|
  Posted:
Jan 03, 2006 - 06:02 |
|
Is 1000 a large enough sample to be considered an accurate empirical result? I would think a sample more the size of a million would give you closer to theoretical percentages? |
|
|
Furious_George
Joined: Aug 13, 2005
|
  Posted:
Jan 03, 2006 - 06:16 |
|
well then, i suggest you get counting. And in response to the complaining issue. Yes. I whinge. Im british.
Heaven forbid that the dice were perfect, then id have to come to terms with my own incompetence to explain my losses. Dice are fine, functional or otherwise
I dont see a problem, and think folk ought to moan more about khemri teams and less about random number generation.
My Opinion (is fact!) |
_________________ Dead Men dont tell tales... But they sure play a mean game of Bloodbowl.
"Hugh Mann eh? Now theres a name I can trust!"
Me Loves Futurama |
|
MightyQuinn
Joined: Jan 13, 2004
|
  Posted:
Jan 03, 2006 - 06:17 |
|
I dont care what the outcome is, all I know is in the past 5 game, I have failed 75% of the time and I wont gfi anymore, but I hate to stall... |
|
|
annachie
Joined: Jul 27, 2004
|
  Posted:
Jan 03, 2006 - 06:36 |
|
Bah it's busted I tell you, busted.
Played a stunty game last night, and failed 11 bonehead rolls out of the 24 I made. I don't think I've ever had a decent result with that skill |
|
|
Rynkky
Joined: Aug 03, 2004
|
  Posted:
Jan 03, 2006 - 12:23 |
|
It is fine, it is just too predictable! |
|
|
odi
Joined: Aug 02, 2003
|
  Posted:
Jan 03, 2006 - 12:36 |
|
So far atleast it's been broken in 506 of my games |
|
|
Jinxed
Joined: Jul 04, 2004
|
  Posted:
Jan 03, 2006 - 12:50 |
|
It is not broken. My opponents roll pow and I rarely do.
Averages out. Works fine. |
_________________ Nuffle sucks |
|
koadah
Joined: Mar 30, 2005
|
  Posted:
Jan 03, 2006 - 13:01 |
|
Whinging, whining, moaning e.t.c. is part of the game. How dare you try to take that away from people with your reasoned, logical argument!
If not for nuffle, dodgy dice, cherry picking, one turning, dirty playing foulers what would we do in those moments of spare time that are not long enough to sqeeze in a game.
Vicimus is a heretic and should be burned at the stake, or given burnt steak, or burnt pie or something. |
_________________
O[L]C 2016 Swiss! - April ---- All Star Bowl - Teams of Stars - 2 more teams needed |
|
Optihut
Joined: Dec 16, 2004
|
  Posted:
Jan 03, 2006 - 13:28 |
|
Vicimus wrote: | Having a degree in Software Development |
Good for you! |
|
|
macike
Joined: Jun 25, 2005
|
  Posted:
Jan 03, 2006 - 13:29 |
|
Vicimus wrote: | I have heard lots of chatter about the Java Random Number Generator rolling double's too often or too many 1's as lot as a whole heap of other nonsense. Having a degree in Software Development, and knowing the basics of how Java generates Random numbers, this sounded extremely unlikely to me. |
I was taught that it generates pseudorandom numbers
Vicimus wrote: | I wrote a quick program in Java using the Math object's Random() method (just generating a 1000 random numbers and then printing the amount of consecutive doubles etc) |
How many seeds have you tested?
What were you going to prove? Can you prove randomness of pseudorandom numbers? WOW... you'll be famous!!!
My advice would be to read the Fumbbl Luck topic. Rynkky has wrote enough for anyone to understand (even without a degree in Software Development) that probability of rolling the snake eyes is:
- 1/36 in case of randomness,
- 0 or 1 in case of pseudorandomness (which was proven by Rynkky's piece of code).
If that's not enough for you, leave the basics and learn some advanced lessons. |
_________________ Hold him, thrill him, kick him, kill him!!! |
|
heinz
Joined: Mar 24, 2004
|
  Posted:
Jan 03, 2006 - 13:43 |
|
*cough* |
_________________ #fumbbl_academy - the old school alternative | #bloodlust - FUMBBL's first Vampire chat |
|
sk8bcn
Joined: Apr 13, 2004
|
  Posted:
Jan 03, 2006 - 13:48 |
|
I am the only one here to think but:
I AM A LUCKY COACH
ha ha ha ha ah ha. |
_________________ Join NL Raises from the Ashes |
|
koadah
Joined: Mar 30, 2005
|
  Posted:
Jan 03, 2006 - 13:49 |
|
|
macike
Joined: Jun 25, 2005
|
  Posted:
Jan 03, 2006 - 13:53 |
|
koadah wrote: | Hmm. Why is it that these threads start at the begining of the year? |
Because 00:00 on January the 1st resets the number generator. |
_________________ Hold him, thrill him, kick him, kill him!!! |
|
|