Mezir
Joined: Aug 02, 2003
|
  Posted:
Jan 04, 2006 - 13:27 |
|
macike wrote: | However 'bad' or 'good' streams are more likely to occur because they are not random. |
Since you can easily test for this (taking the average of the difference between each two consecutive rolls and comparing it to the standard deviation, I think) I am willing to doubt the validity of that statement. "Random enough" also takes into account that it will create bad and good streams just as much (no more, no less) as "true random" would. |
_________________ Build a man a fire and he's warm for a day; set a man on fire and he's warm for the rest of his life. |
|
koadah
Joined: Mar 30, 2005
|
  Posted:
Jan 04, 2006 - 13:55 |
|
koadah wrote: |
You still need to be sure that the same JVM is running the game, the pure java browser and the applet. Otherwise you may as well pick any C library and test that. |
OK, I've checked and different JVMs should use the same algorithms to create their pseudo random numbers so we should be fine.
I don't know whether this applies to javascript. Let's say that it probably does so that Vicimus can have a good nights sleep. |
_________________
O[L]C 2016 (big teams, progression) Swiss 9th Oct! --- All Star Bowl - recruiting NOW!! |
|
macike
Joined: Jun 25, 2005
|
  Posted:
Jan 04, 2006 - 14:10 |
|
Mezir wrote: | macike wrote: | However 'bad' or 'good' streams are more likely to occur because they are not random. |
Since you can easily test for this (taking the average of the difference between each two consecutive rolls and comparing it to the standard deviation, I think) I am willing to doubt the validity of that statement. |
I am willing to doubt that two consecutive rolls form a 'stream'. |
_________________ Hold him, thrill him, kick him, kill him!!! |
|
Mezir
Joined: Aug 02, 2003
|
  Posted:
Jan 04, 2006 - 14:12 |
|
Which is why you take the average. Long streams of consecutive numbers being the same, or close to the same, will cause the average to be lower than the standard deviation for a d6 (1.7), where as too few 'streams' will cause the average to be too high.
Edited to put the correct figure for stdev in. |
_________________ Build a man a fire and he's warm for a day; set a man on fire and he's warm for the rest of his life.
Last edited by Mezir on %b %05, %2006 - %01:%Jan; edited 1 time in total |
|
macike
Joined: Jun 25, 2005
|
  Posted:
Jan 04, 2006 - 14:27 |
|
Mezir wrote: | Which is why you take the average. Long streams of consecutive numbers being the same, or close to the same, will cause the average to be lower than the standard deviation for a d6 (3.5), where as too few 'streams' will cause the average to be too high. |
No way! Long streams mean snake eyes re-rolled with snake eyes or DD DD (Defender Down) regenerated kills. |
_________________ Hold him, thrill him, kick him, kill him!!! |
|
Rynkky
Joined: Aug 03, 2004
|
  Posted:
Jan 04, 2006 - 14:36 |
|
Good proof about good streams and bad streams was "Bad karma" option that I implemented in roll predictor (Prophet was the true name of the project). When "Bad karma" option was selected, your opponent was having extremely bad luck at the beginning of his/her turn and in the contrary you were having extremely good luck at the beginning of your turn.
How this was implemented?
Easily enough just by searching forward the random number stream until a bad stream (when opponent's turn) or good stream (when your turn) was found. Usually next sequence (good or bad) was found in 40 to 60 rolls. This itself proofs that the streams exist.
IIRC bad stream was if the sum of next 5 rolls was 7 or lower and good stream was if next 5 rolls were 27 or higher. |
|
|
Mezir
Joined: Aug 02, 2003
|
  Posted:
Jan 04, 2006 - 16:47 |
|
I'm not trying to disprove the existence of streams. I'm just saying they occur no more frequently in the RNG than in actual rolled numbers. How often have I rolled quad skulls in JavaBB? Not that stunningly often. How often have I rolled them irl? I know of at least two times, and my number of rl BB games ain't stunning.
What about the time when I killed a Kroxigor on a -2d block? That's a serious "good karma" stream. And it was real life.
Streams exist. In any sequence of "random" numbers, no matter how they were generated. Rynkky's "karma" function just rolls the dice until you get to one. If you could predict dice that you're going to roll by hand in the same manner, you could do the same.
I guess we're talking at tangents again, huh? |
_________________ Build a man a fire and he's warm for a day; set a man on fire and he's warm for the rest of his life. |
|
macike
Joined: Jun 25, 2005
|
  Posted:
Jan 04, 2006 - 16:54 |
|
Mezir wrote: | I'm not trying to disprove the existence of streams. I'm just saying they occur no more frequently in the RNG than in actual rolled numbers. |
In accordance with Rynkky's expirience I would dare to say that they do. |
_________________ Hold him, thrill him, kick him, kill him!!! |
|
Rynkky
Joined: Aug 03, 2004
|
  Posted:
Jan 04, 2006 - 20:34 |
|
The whole talk about streams was about using same RNG for both players. When other player has good luck with stream of good numbers, this usually (yes I know I use indeterminate term) leads to stream of bad numbers because things even out.
Why this is bad? Well in game bad stream most propably means turn over, meaning other player gets his turn and if the luck "changed" will have some lucky rolls. This is not always case but when I made the prophet I certainly noticed this kind of behaviour. |
|
|
Vicimus
Joined: Nov 16, 2005
|
  Posted:
Jan 04, 2006 - 23:06 |
|
Rynkky,
You have to be kidding me right?
I haven't looked at your prophet program, but without doing so I feel safe saying that I am sure the Random object would not work by purposefully trying to create comparitively even streams. What you are saying (as far as I understand it), is that if the first set of numbers that it produces are slightly lower than the average, then the next set of numbers will be slightly higher than average in order to "balance" things out. I am having a very hard time believing that's how it works.
Will a run "bad numbers" last for 3 rolls, 10 rolls, or 10 million rolls?
Tell me how many rolls a run of "bad" or "good" luck goes for, and I feel extremely confident I could write a Java Applet that shows this "balancing" theory wrong beyond any reasonable doubt.
Macike,
The fact that the seed is created ahead of time means nothing. It doesn't matter if it was created a million years ago or 1 nanosecond ago. The numbers it produces will still not be bias one way or the other. Unless what Rynkky says is correct, but I find it hard to believe. |
|
|
Rynkky
Joined: Aug 03, 2004
|
  Posted:
Jan 04, 2006 - 23:57 |
|
Vicimus wrote: | You have to be kidding me right?
I haven't looked at your prophet program, but without doing so I feel safe saying that I am sure the Random object would not work by purposefully trying to create comparitively even streams. What you are saying (as far as I understand it), is that if the first set of numbers that it produces are slightly lower than the average, then the next set of numbers will be slightly higher than average in order to "balance" things out. I am having a very hard time believing that's how it works. |
Algorithm isn't balancing anything out but it has tendency to balance out, even on relatively small sample. And you know it, if it's balanced (e.g. average is ~3,5) then any stream of bad rolls has to be compencated with stream of good rolls.
Make a prophet-like-program yourself or get older version of JBB (8.3 or older should be fine) and I will send you mine. Then play few games and make your own conclusions. I am just too tired of harp on same issue over and over again. |
|
|
koadah
Joined: Mar 30, 2005
|
  Posted:
Jan 05, 2006 - 00:09 |
|
|
Vicimus
Joined: Nov 16, 2005
|
  Posted:
Jan 05, 2006 - 00:13 |
|
Rynkky,
Of course it has a tendancy to balance out! It's the nature of generating a lot of random numbers. If they didn't balance out, they wouldn't be random (they would obviously be bias one way or the other). I know you realise this already. But the question is whether this balancing act is occuring by coincidence or if Java is purposefully balancing the numbers. For the "good stream/bad stream" theory to be correct, it would have to be the latter as if if was occuring by coincidence, the variation would be too scattered to fit into sets (or "streams"). I know you're cluey enough to realise all that already.
I don't know how your prophet program works, I haven't looked at it (but I would be interested). Are you sure it's not just a computerised way of crunching the numbers and playing the odds? I mean can it predict the roll every time? Is it accessing the seed somehow? Or is it taking an educated guess going on past rolls? Tell me the basics of what it is doing.
Anyway, I hate to be nit-picking bastard, but I wrote a quick script to test the "stream of good and bad numbers" theory, at-least as I understand it (which may be wrong).
Draw your own conclusions: http://home.kooee.com.au/issima/random2.htm |
|
|
sk8bcn
Joined: Apr 13, 2004
|
  Posted:
Jan 05, 2006 - 00:35 |
|
can a nice technical summary be done here? I mean a summary with point, argument exemple. I want to mix up in the debate |
_________________ Join NL Raises from the Ashes |
|
Vicimus
Joined: Nov 16, 2005
|
  Posted:
Jan 05, 2006 - 00:40 |
|
Okay,
Well, it started out by me saying that the Java Random Number Generator does not have a habit of rolling double 1's or double 6's or any other abnormal patterns of numbers. Although this was initially disputed, opposition to that statement seems to have vanished. Somehow the debate got twisted into whether the Java produces streams above average rolls followed by streams of below average rolls (a stream being say, ~50 consecutive rolls).
I think it doesn't. Rynkky and Macike think it does. A few others seem indecided. I also made these two pages to test both theories...
http://home.kooee.com.au/issima/random.htm
http://home.kooee.com.au/issima/random2.htm
Hope that helps. Join in the psuedo-intelligent debate. |
|
|
|
| |