57 coaches online • Server time: 23:32
Forum Chat
Log in
Recent Forum Topics goto Post Blackbox Teamsgoto Post Secret Stunty Cup IVgoto Post FUMBBL HAIKU'S
SearchSearch 
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic
Voidhawk



Joined: Jul 28, 2004

Post   Posted: Aug 07, 2004 - 03:53 Reply with quote Back to top

Hi!

I guess a lot of people complain about the generator, Im not going to do that. Smile I was thinking of how to check if its really good. I mean you need to "throw" the dices a couple of thousands times around to see the distribution. But my point is that we have already done that. Doesnt every log file contain the dicethrows? Is it possible to just collect these numbers and see how the Java generator behaves? It should be possible me thinks, but it possible requires admin access? Does anyone have any thoughts on this?

Voidhawk Wink
CorporateSlave3



Joined: Feb 07, 2004

Post   Posted: Aug 07, 2004 - 03:57 Reply with quote Back to top

You really ought to search this topic...there is a thread where someone did put the Java random number generator thorugh its paces and after several hundred thousand die rolls it came out...random.

_________________
***Did you know? 42.7% of all statistics are made up on the spot?
BadMrMojo



Joined: Aug 02, 2003

Post   Posted: Aug 07, 2004 - 22:50 Reply with quote Back to top

I made a voluntary home brew version of what you're asking for specifically because of all the people asking things like this.

There's no error checking or anything but it works well enough to get the idea across. Source is available if you want to check that, too.

_________________
Ta-Ouch! of BloodBowl
Condensed Guide for Newbies
kurre



Joined: Oct 14, 2003

Post   Posted: Aug 07, 2004 - 23:19 Reply with quote Back to top

Cool script BadMrMojo.
SunTzu



Joined: Feb 26, 2004

Post   Posted: Aug 11, 2004 - 13:20 Reply with quote Back to top

Something we at www.megamek.net has started to suspect is that the java RNG is flawed in the way that it tends to do long runs of numbers, like 5 5's in a row, then 5 7's, etc... Very strange.
Mnemok



Joined: Feb 11, 2004

Post   Posted: Aug 11, 2004 - 13:27 Reply with quote Back to top

To make the long rows of numbers problem go away (or any random problem), why not just (for each die roll) roll the die ten times and then roll another die (ten sided) to decide which one to use. The extra computing time would be close to zero. If random isnĀ“t enough make it more random.
Is this a good solution or am I out flying when I have these thoughts.

(BTW I also have a feeling the rand.gen. is flawed) Laughing

_________________
Dyslexics have more fnu.
Eddy



Joined: Aug 04, 2004

Post   Posted: Aug 11, 2004 - 13:48 Reply with quote Back to top

randomness also creates long streaks of the same results.

_________________
'The generation of random numbers is too important to be left to chance.'
Robert R. Coveyou
Laviak



Joined: Jul 19, 2004

Post   Posted: Aug 11, 2004 - 14:20 Reply with quote Back to top

I can't remember the exact details .... but you can also check the frequency of runs, or any other sequence, as well as the frequency of the individual values.
Someone with i bit more knowledge of statistics would probably be able to provide a bit more info Smile

Basically, I think the idea is that there should be 6x as many runs of '1', as there are runs of '2', and 6x as many runs of '2' as there are runs of '3'...etc.
Renno



Joined: Jun 17, 2004

Post   Posted: Aug 11, 2004 - 16:05 Reply with quote Back to top

You know when I have 3-5 games of bad luck in a row such bad luck (while my opponent is having really good luck) that I have no possible way of winning it gets old.

Then I go and have 1-3 games of crazy good luck and I know my opponent can't be having any fun while he has his bad luck that I get the same sick feeling.

I want a fun, competitive game where both of us have a chance of winning not these bad and good luck streaks I'm experiencing.

I don't mind losing but my god when you have no possible way of winning because of a bad luck streak you just can't have any fun.
Grumbledook



Joined: Aug 02, 2003

Post   Posted: Aug 11, 2004 - 16:32 Reply with quote Back to top

we checked after 100k games and the distribution was fine
Renno



Joined: Jun 17, 2004

Post   Posted: Aug 11, 2004 - 16:40 Reply with quote Back to top

You know grumbledook you say that but does that log sparser read all rolls including armor rolls, kick of table, et cetera?

The odds of getting 5, 1s or 5, 6s in a row is in the millions yet in less then 200 games I've seen it at least 7 times each.

That alone tells me something isn't adding up.

I've had streaks of 3 and 5 games in a row the luck was so horrible it just got stupid. I mean hell I'm here to play a fun game not be frustrated and afraid to throw 1 block or pick up the ball let alone throw or catch it.

I know I'm not the only one experiencing this set of extremes.

If you have 100 rolls in a row of the same number 6 times each a different number your parser is only gonna show 1-6 being rolled 100 times not extreme stretches of those numbers.
Renno



Joined: Jun 17, 2004

Post   Posted: Aug 11, 2004 - 16:43 Reply with quote Back to top

P.S. If a casino was using this generator I'd make a fortune off the double 1s. Wink
LordRummickIV



Joined: Apr 20, 2004

Post   Posted: Aug 11, 2004 - 17:18 Reply with quote Back to top

You can't complain about the random numbers. You have to think the 16% probabilities for each number, only will be 16% after a lot of rolls.

(And a lot is more than 1000. In a short serie of 20, you can get 20 sixes, and you won't complain)
BadMrMojo



Joined: Aug 02, 2003

Post   Posted: Aug 11, 2004 - 17:19 Reply with quote Back to top

In "less than 200 games" you have seen over 100,000 rolls*. The odds of getting 5 consecutive numbers are 1/6^5 or 1/7,776. On average, 100,000 rolls you will get strings of 5 consecutive numbers 12.86 times.

Roll a die 100 times. Honestly. Sit there and do it. Write it all down as you go.
Then do it again. And again. And again. And again.
You will very possibly get streaks of 5. Or maybe you won't. That's the definition of random. You can't predict it.

Looks like you just didn't notice those strings of 4s. Of course you didn't. It's psychology, not math.


* this is not an imaginary number. Check one of your .zip results from a game in which you didn't reload (so you have the full game in the .zip file). Look at the .log file in Notepad or TextEdit and scroll down to about 5 lines up from the end. It will tell you all the numbers that you rolled. Try a couple and you'll find that there are roughly 1,000 rolls per game, depending upon how risky you play. 100,000 rolls is a conservative number, to be honest. You're probably at least 50% higher than that.

_________________
Ta-Ouch! of BloodBowl
Condensed Guide for Newbies


Last edited by BadMrMojo on %b %11, %2004 - %17:%Aug; edited 1 time in total
BadMrMojo



Joined: Aug 02, 2003

Post   Posted: Aug 11, 2004 - 17:21 Reply with quote Back to top

DaBoss wrote:
P.S. If a casino was using this generator I'd make a fortune off the double 1s. Wink

Casinos are using this generator and it's because of people who think like you that they stay in business. Wink

_________________
Ta-Ouch! of BloodBowl
Condensed Guide for Newbies
Display posts from previous:     
 Jump to:   
All times are GMT + 1 Hour
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic