Time to reboot

[anna Acanthus]

I'll just add DITTO and NC !

[Liriel Garnet]

(04-08-2012 09:50 PM)Callie Cline Wrote:  this is fascinating... i'm just chiming in cuz i have a great aunt who had 12 boys and guess what her 13th was? a girl finally. she was also one of 11 kids the only girl... (but not the youngest) it would be an interesting study in RL to see patterns like what you describe.

i know when my family bred dogs when i was a kid we'd have litters with sometimes equal male and female, and often buyers would "hold" a female or male, and we'd have NONE!!! our dogs had large litters being german shephards and dobermans and english springer spaniels, but anyway just wanted to chime in.

i'm sorry it's frustrating

Callie, this isn't about one cat producing a string of one sex, this is about every box that I open from several breeding pairs in a project line. This'd be more like whole neighborhoods producing only children of one sex (the lines) in the middle of towns that produce huge runs of children of one sex (individual breeders - because a *lot* of us have noticed month+ periods where we birth mainly one sex - with some producing boys and some girls).

True randomness will produce groups and runs, there's no question about that. But what I'm seeing isn't about one run, it's about everything being out of balance. There should be groups of 1, 2, 3, not *just* long runs. Long runs should be the exception, not the rule. Overall, I'm absolutely sure the numbers work out approximately even, but that isn't a reliable indicator of randomness either, it's simply an indication of overall balance in whatever method is being used to calculate the numbers.

Since any computer-generated 'random' number is pseudo-random, at best, calculations of 'random' results via a computer is highly susceptible to bias, both as Tad pointed out as well as via the potential addition of any other factors by the programmer, including things like ranging the output of the random number generator used.

[KittyCats Resident]

(04-09-2012 11:04 AM)Liriel Garnet Wrote:  Since any computer-generated 'random' number is pseudo-random, at best, calculations of 'random' results via a computer is highly susceptible to bias, both as Tad pointed out as well as via the potential addition of any other factors by the programmer, including things like ranging the output of the random number generator used.

There is no "kittycats" input added to create a cat gender, neither userdata or data of the cat itself, ruling out any relation between gender and traits, gender and users or gender and parents.



As "pseudo-random" doesn't say a lot to many, i'm explaining here what that could mean...

For Example:
Each time we have to "create" a gender, we would check what time it is. And we say, if the number of seconds ends at 1, 3, 5, 7, 9 its a girl, if it ends at 0, 2, 4, 6, 8 its a boy.

This would for the user give the impression its random, while its actually not at all...
Thats what we call "pseudo-random".

The user would also hardly be able to influence it, as the cats decide when they rez a box, and there may be delays when the box contacts the database...

This is only a simple example though, and the actual way servers create a feeling of randomness is far more complex, and less predictable.

The data i have provided earlier also focused on smaller time periodes, more specific for a day and for 15 minutes, as well as a specific fur, not just an overall global view. All of them which worked out.



Does randomness mean i will get nicely a boy and a girl alternating each time?
No, it would almost be like winning the lottery

Imagine this:

You have 10 buckets, each containing a ball with number 1, and a ball with number 2.

if you were allowed to pick a ball from each bucket (without looking), its extremely unlikely you would pick 10 times a ball with number 1.

It is however as unlikely that you would pick first a 1, then a 2, then 1 again, 2 again, and so on...

While with so many users, there may be a few of you who have bumped into either situation, the average user will get any of the scenario's in between.

This 10 buckets with 2 balls can give up to 1024 different outcomes.




If you want to read more about randomness, including how a more "true" randomness could be reached even by a computer, visit http://www.random.org/

Regards,
KittyCatS

[Tad Carlucci]

To me, the question is not randomness, per se.

I take two things of faith:

First, KityCatS Resident is using nothing but the frand() function (or it's equivalent) to determine gender.

Second, frand() is a computer function and therefore not truly random.

So I have every confidence that the proportions are about equal. In fact, if KittyCatS Resident were to cite numbers showing consistent, exact, 50/50 splits .. THEN I'd be upset because that is NOT random.

But I realize that frand() has it's problems. And sometimes, after a very VERY large number of times used, it can develop into patterns. Those patterns will still hover close to the 50/50 split. But if we look at the results we see repetition. It might be "5 males, then 5 females, then 4 males, then 6 females, then 7 males ... " What's wrong with that? Well, there should be some "1 male" or "1 female" mixed in! It's still about 50/50 and it still wiggles around .. but there are NO short runs. Similarly, it might be ONLY short runs, and no long ones.

Is this what's happening? We don't know. But, as a systems admin, I look to complaints. And if I were to see a growing number of people wondering why they're getting "only males" I'd look to be sure, first, that they're not ignoring a female or two. But I might just decide .. well it's been a few days and I use frand() so much that's probably a few million hits .. so what-the-heck .. let's reboot to ensure it's not frand() gone wonky. That's easier than carefully monitoring and it can't hurt.

I've not done the calculations for KittyCatS because I have not found a good way to estimate daily box production. But, when I did it for another breedable with a daily nest production of about 12,000 I came up with about 10 days before I figured frand() would exceed the minimum guarantee of 2^24 bits output and it entered the range where randomness could begin to break down and .. guess what .. at the time the server had been up for over two weeks. Maybe frand() was not the culprit, but a few days later the hosting service messed up an ATM change, rebooted the servers, and .. yep .. you guessed it .. complaints dried up for weeks afterwards. Was that what fixed it? Dunno. No way we'll ever know but it sure didn't hurt and the lack of complaints was refreshing.

[Sanura Snowpaw]

I mostly get runs of boys or girls in certain lines although overall I am fairly even on what I unbox. Sadly right now the ones I've been unboxing to keep turn out to be boys, it is frustrating but I still love them.

[Liriel Garnet]

(04-09-2012 04:42 PM)Tad Carlucci Wrote:  But I realize that frand() has it's problems. And sometimes, after a very VERY large number of times used, it can develop into patterns. Those patterns will still hover close to the 50/50 split. But if we look at the results we see repetition. It might be "5 males, then 5 females, then 4 males, then 6 females, then 7 males ... " What's wrong with that? Well, there should be some "1 male" or "1 female" mixed in! It's still about 50/50 and it still wiggles around .. but there are NO short runs. Similarly, it might be ONLY short runs, and no long ones.

This is *exactly* what I'm talking about. I'd expect things to be pretty even overall, mostly short runs with the occasional longer runs thrown in. That's not what we see. And *we* have effect on this as well, because we pick and choose which boxes to birth. But if everything is truly random, then the boxes we pick should still, over a period of weeks, start to even out.

to KC: I was talking about a limiting function. Some code packages have a random number generator that allows you to set top and bottom limits, for example, or specify integer returns rather than double float. Depending on the code package, using those limits can greatly reduce the repeat period of the generator.