Vertigo
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  17:35 

well it really doesn't matter where you get the rolls from no dice roll steals another dice probabilty.
one of the needed constraint for a random number is that every subgroup also has even distribution.
(subgroup meaning 20 random rolls, or first 20 or such a like not a group with every roll that was 6) 


Spyrer
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  19:00 

Well, I once spoke against the dice code, then I shut up...
I must however say that my Java BB games are much more "extreme" than my regular ones.
But indeed people tend to notice bad rolls more than good rolls...
My record in bad rolls is 5 consecutive double 1s (not bad eh?)
And the most positive rolls I saw (unfortunatly against me), was 5 6s (AV, injury, dead)
Last game I block one of my own gobos with a natic, got 3 blocks, reroll, pow, block, pow, threw the gobbo down, KO and the natic collapsed... now that would have been really bad, if the other guy didn't start his turn with double skulls... (The game was very wierd) 


Darkwolf
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  19:32 

I think the only REAL difference in jave versus real life BloodBowl is the human factor. I remember a few times ignoring a 12 and NOT using my DP skill b/c I felt bad I just killed my buds best player. Or taking a RIP for a SI instead. You get no sympathy factor online, you take the dice as they land, no mercy, no forgetting to use DP, no magic sponge or anything. Also, in our face to face leagues we changed the injury table a bit.
14 BH
5 SI
6 RIP
While the java uses a truly random number generator, it just lacks the 'magic' that humans have when we roll dice. 
_________________ Check out the latest Darkwolf "***Did you know?" in his bio! Each month, a new Darkwolf factoid!
http://fumbbl.com/FUMBBL.php?page=coachinfo&coach=768 

SkiJunkie
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  20:28 

Quote: 
well it really doesn't matter where you get the rolls from no dice roll steals another dice probabilty.

Well I didn't mean steal exactly. I meant that if you are only looking at the d6 rolls to see if they are evenly distributed then it is possible they will not be because you are excluding the d8 rolls (and all other rolls). In reality the random number generator creates numbers between 0 and 1. So ALL rolls are between 0 and 1. So you must look at ALL the random generated numbers in a game to determine true distribution, not just the random numbers used to create d6 rolls.
I am just saying that even when you see the d6 distribution there will be games where it looks very very wrong because of this, and you will get more threads like this one complaining about it.
But regardless, I've just added the distibution of rolls for d6 and d8 rolls in a game. After the next patch you will be able to find these near the bottom of the log file.
As far as taking a look at the random generator code, that just isn't going to happen. I would never try to write my own, and I would never trust one written by someone without some serious credentials.
The only computer random generator I have ever seen or heard of that gave true to life random numbers was at MIT. Some computer scientists there hooked up a server to a camera pointed at a lava lamp. The server, when asked for a random number, would take a snapshot of the lava lamp and generate a random number based on what the lava lamp looked like. Thus true analog random numbers that don't completely depend on a restrictive digital computer. 
_________________ SkiJunkie
________________________________
Java BBowl  http://www.geocities.com/javabbowl
Mirror  http://javabbowl.noip.org
Mirror  http://home.austin.rr.com/javabbowl 

Grumbledook
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  20:42 

so we need to spend the donations on a lavalamp huh ;]
the bulbs in my room blew a couple of days ago anyway i need a lightsource and i already got a cam 


Christer
Joined: Aug 02, 2003

Other popular sources of random numbers are radioactive isotopes. Something to do with measuring the radiation they emit. 


Revener
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  21:53 

/spam on
I have an idea! What if Christer manually throw all the dice himself
Or just ignore me 


MattiasF
Joined: Aug 02, 2003

Posted:
Aug 08, 2003  22:37 


slackman
Joined: Aug 02, 2003

Posted:
Aug 09, 2003  23:46 

i think the problem lies wherein the same number generator is used by both you and your opponent. so if you roll double 1s, after barely succeding w/ skill rrs on a couple 2,3s or such, your opponent is much more likely to start rolling 5s and 6s.
the luck calculator itself weighs the % chance to succeed against whether or not it actually does succeed. i saw a guy roll a 1 on a dodge, use a rr, and roll a 6, and his luck went DOWN after both rolls were taken into account. is this right or wrong?
every 2+ dodge has a 1/6 chance to fail. this means that if you dodge 6 times on a 2+, one of them WILL FAIL. another problem arises when you confuse probability and odds. the odds of rolling double 1s is 1/36. the probability is the same. but after you roll the first one, the odds dont change, but the probability becomes 1/6, as you are only rolling a single 1/6 chance.
you often will use a team rr (or skill rr in certain situations) when you need a 2+ and roll a 1. but how often do you rr when the result is a 6? you dont, because you dont need to. as someone mentioned above, he tried doing that, and found that you do get just as many "double 6s" as you do double 1s.
in stunty league, i roll a grip of 1,2,2 or 2,2,1 on 3d blocks. but then i throw a grip of 3d blocks. i dont keep track of the times i roll 6,6,5 or anything similar, because they succeed.
the point here is, javabbowl dice hate me, and ive come to accept that. sure, it still pisses me off, but its not going to change. either i get everything my way, and win by a huge margin (i once won 40 w/ my dwarves) or its a close game, even if i do lose. good coaching comes from eliminating dice rolls, not counting on them succeeding on a 3+.
no matter what you think, im sure my first point is the most important. you and your opponent are using the same number generator, w/ the same original seed. if you fail everything, it is going to be quite likely your opponent will seem to never fail at all.
slackman42 


Candlejack
Joined: Aug 02, 2003

Posted:
Aug 10, 2003  00:15 

err..., the random generator has no "pool" of numbers from where it picks the dice results. the results of any dice rolls are completely indpendent from it's predescessing results.
Correct me if I'm wrong
Candlejack 
_________________ 
The Sanity Resort 

Revener
Joined: Aug 02, 2003

Posted:
Aug 10, 2003  00:42 

Still what slackmans says happens quite a lot if you fail everything your opponent usually manage all his roll. Not counting those fun games where both sides fail miserably in everything like not being able to pick up the bal for 8 turns or something 


MattiasF
Joined: Aug 02, 2003

Posted:
Aug 10, 2003  01:20 

slackman wrote:  i saw a guy roll a 1 on a dodge, use a rr, and roll a 6, and his luck went DOWN after both rolls were taken into account. is this right or wrong? 
Chance to succeed on a 2+ roll is 5/6, chance to fail once and succed with the reroll is (1/6)*(5/6) == 5/36 which is less than 5/6, plus he burt a reroll. I say its correct..
slackman wrote:  every 2+ dodge has a 1/6 chance to fail. this means that if you dodge 6 times on a 2+, one of them WILL FAIL. 
Sorry can resist....
There is just a probability to fail 6 2+ rolls, not a atomatic failure.
The exact chance to succed in 6 2+ rolls is:
(5/6)^6 ~= 33,5%
The chance to roll 2+ on one die is 5/6, to do that (independenty) 6 times, you simply multiply the chances.... 6 2+ rolls will succed one time out of thee. Go figure... 


MrKlipp
Joined: Aug 02, 2003

Posted:
Aug 10, 2003  05:32 

This is no more correct than saying that if you roll lots of 6's on real dice probability means that your opponent will roll more ones. Each die roll or random number generated has the same chance of being any number as any other, probabilistic effects only become relevent when looking at many hundreds or thousands of rolls. 
_________________ Looking to get your minis painted? Look no further.
The Finishing Touch 

WolfyDan
Joined: Aug 02, 2003

Posted:
Aug 14, 2003  12:05 

In fact there is a huge mixing of a priori and a posteriori knowledge with all these homebrewed stats being used.
It is importasnt to note that stats do not have to 'work out' in the long run. That is to say that if you roll 20d6 and they all come up a 1, then you have not 'burnt up' or your bad luck. Previous rolls have no effect upon later rolls. Probability distrubitions are based upon a wonderful concept called infinite trials, so if anyone here plays a game with an infinite number of rolls and then still fails all those 2+ dodges, then I'll have some sympathy
This leads us to a term called confidence, which is, simply put, the probability of probability. Take the 4GFI example christer shot me down with ( ), he rightly stated it had about a 48% chance of working. For simplicity's sake, lets assume this is 50%. So now if I attempt it 6 times, and it works in only 1 of them the whole thing can be resolved to the same as flipping a coin 6 times and it only ending on a heads once. This amalgamation of stats into stats can actually prove very tricky.
But I digress, what I meant to say was that it is dangerous to say that the chance of a reroll coming up the same is 1/6. Although 'puristically true' it does not take into account that sometimes rerolls will not be taken. Noone can remember the 5 reroll 5, because hardly anyone will ever need to reroll a 5. Thus the stickiness of rerolls will become a problem, because they are not taken in random situations (as the other dice rolls are).
I do understand the psychological aspect involved (gawd knows I do), but the pattern that I had picked up on was worth questioning, I thought. Thanks for the replies Christer & Ski Junkie 


psikobunny
Joined: Aug 02, 2003

Posted:
Aug 14, 2003  12:44 

My point is this: All of these examples and patterns that people put forward as a flawed dice roller, are anecdotal which equates to useless. Something you (general not specific) notice as a pattern probably isn't and none of us can be considered a valid observer of a full game. 



 