My save doesn't like men

[FyunchClick]

This was looked at very closely during testing. For each Kerbal generated, there is an equal chance of being male/female. Through random luck, it sometimes happens that the pool is lopsided in one direction. Statistically, the more Kerbals you rescue, the closer to a 50/50 distribution you should get. However, it is also possible to get a thousand female Kerbals, although that is unlikely.

I don't think so, and it's just the missions, the staff you can hire is pretty much 50-50. I've done about 20 rescue missions and never a male. So the chance of a streak like that is 50% to-the-power-of 20 which is 0,000000953, or almost exactly one in a million. And there's about half a dozen more of us.

It's a bug alright, though pretty benign, and I've decided that it's because with every new save I start, I get a different Kerbin out of the infinite Kerbins in the kerniverse, and on this one they have a very different biology.

[Moach]

it is indeed a good exercise in the observation of random samples generation and the often surprisingly "unrandom" results it can appear to produce...

I've had plenty of experience with random lately, being the lead dev. on a mobile bubble shooter game... -- you know that feeling that you never seem to get the bubble of the colour you want? - you're not going insane - that's MY doing

mwahahaha...

in the game I'm making (should hit the appstores within the next few weeks) - we've devised a somewhat intricate "biased random" color selection system -- this was done because "actual random" simply proved to feel "not random enough", and gave us no options to fine tune each level's difficulty (it biases both ways, mind you)

problem with random things, I mean, is that humans are hard-wired to look for patterns in things - it's an instinct, long story... anyways, in order to achieve a sense of "fairness" with a localized random process, one must often take contrived measures to allow for a (seemingly) more homogenous grouping

this is just to satisfy our human expectations that a randomly selected set should somehow be evenly grouped -- even though there is really nothing other than sheer dumb luck that would provide for that...

so truth of the matter is: true random isn't

and with that - I do seem to have more girl Kerbals than boys - I know it's random... but problem is -- it really IS random, where perhaps it shouldn't be... it works for human populations, sure... with 7 billion samples, your statistics are bound to converge...

but much like there are families of 5 children all of the same gender, a localized truly random sample group has a large tendency to appear biased - especially given our human disposition to see patterns, even when they're not there

my suggestion to Squad would be: -- make it less random -- use some statistics "magic" to make it appear more random by really making it less so...

a simple method: when groups are uneven beyond a set tolerance, draw odds twice - then from the two, pick the result that favours better balancing (if available)

it still is fundamentally random, but it will try to correct itself so that apparent patterns are disfavoured to some extent

Cheers

[tater]

With the Asteroid Days mod, mine is not a normal possible outcome of all females. It is, with any new career, 100% female, always, no exceptions. I disabled the mod, in the careers with no males, then got... males, immediately. I tested this in multiple careers. Others have observed the same thing. So the question is, does everyone observing this have Asteroid Days installed?

[Nathair]

So the question is, does everyone observing this have Asteroid Days installed?

I have it installed. I get rescue missions for both genders.

Edit:Running Windows PC

Edited September 10, 2015 by Nathair

[tater]

I might have dumped it before 1.04 at some point, now that I think about it. It would be interesting to see KSP version vs that mod, and rescues/VIP/tourist contracts (when I last had AD installed, all those were 100% female). But it might have been 1.03.

It might also vary by platform build, perhaps? I'm playing on my Mac, not my PC.

[Renegrade]

in the game I'm making (should hit the appstores within the next few weeks) - we've devised a somewhat intricate "biased random" color selection system -- this was done because "actual random" simply proved to feel "not random enough", and gave us no options to fine tune each level's difficulty (it biases both ways, mind you)

I generally use a Knuthian shuffle for any of my random game-development-ness, you can effortlessly stack the deck with those, and guarantee short term distributions.

but much like there are families of 5 children all of the same gender, a localized truly random sample group has a large tendency to appear biased - especially given our human disposition to see patterns, even when they're not there

Well, in my last few stock-ish games, I've rescued about 70 Kerbals. 70 of the Kerbals were female, and 0 were male. The odds of that happening are 8.47e-22. That's like three games, no boys aside from Bob, Bill, and Jeb.

I'm not sure if Asteroid Day has something to do with it or not (I usually do have that installed, I like the big fixed panels); but it strikes me as rather unlikely that I'd be getting a 70:0 ratio (that number is really durned close to being Avogadro's number inverted, geez).

As a counter example, this simple throwaway code:

#!/usr/bin/perl

$m=$f=0;
for($i=0;$i<25;$i++) {
	if( int(rand(2)) ) 
		{ print "Male\n"; $m++;} 
	else 
		{ print "Female\n"; $f++;} 
}
print "$m/$f males/females\n";

when run 20 times, resulted in roughly equal distribution. Only once did it favor one gender, and that was a 5/20 split (with the longest run being only 7 females), not a 0/70 split. And that isn't even cautious use, a really poor quality system rand() could really mess that code up.

I suspect that Squad's code looks something like this:

Name_and_Therefore_Career = MakeAStringFromANumber(rand());

Gender = rand()&1;

You can end up with some nasty bias that way. Throw into the mix that contracts are being randomly generated, probably from the same pseudo-random-number-generator state, and you could very well end up with a situation where a given type of contract could only have a given type of gender because of some bias in the PRNG.

Of course, C# might have a rand() call that ties in with any hardware RNGs, which might allow it to ditch any evil bias it normally has in software mode (also it may vary per underlying OS or OS version etc, with some versions having less than stellar properties).

Maybe I'll throw together a pure stock game to see if it still happens..

(Interesting tidbit: I used to hire kerbals with low stupidity to run my labs as scientists (a bit of RP there, there's no real benefit it seems), but after kerbals had careers, I'd noticed that scientists offered for sale were generally always stupid :/ )

[NecroBones]

so truth of the matter is: true random isn't

Yep, anyone who does programming long enough will eventually run into this.

(Note: I'm not saying there's no bug. It sounds like there might be one.)

Apple had a similar problem with their shuffle feature on iPods and iTunes. If you have 5 songs from the same artist in a playlist of 1000, it seems odd to have 3 in a row come up in shuffle. People complained about the seemingly non-random nature of it, and how songs would cluster up, and some days it would seem to have a "favorite" artist.

But that's the nature of randomness. It creates false patterns, which our brains are very good at detecting. The human brain is designed to easily find patterns, not randomness.

So when it comes to something like this, if you don't want "clustering" effects, then sanity-checks need to be added. One possible example for male/female kerbals would be to alter the recruit/rescue generator like this: Default to 50% chance for each of male and females, however if either gender's population falls below 40%, then you only add to the lower population.

The unfortunate side effect of this would be some edge case disasters in your game. For instance, if you loaded up a ship with 20 females, and crashed it, you'd get nothing but females to replace them for a while. It would be as if killing your female astronauts encouraged women to join your space program.. Heh.

But this can also be done with fuzzy logic as well. Instead of padding the lower population 100% of the time, it could be more like 90%, so you would occasionally still get the higher population's gender as well.

Edited September 11, 2015 by NecroBones

[MaxwellsDemon]

I have noticed this as well-- all of the rescuees are female. It's not as important to me as what their profession is, as I like to have a mix of pilots, scientists, and engineers, but I did notice it as an oddity and wondered if it was the "rescue the damsel" thing. I do have Asteroid Day installed.

[Gargamel]

Sure, for tiny sample sizes but that's a far cry from asserting that a (any) random function must be heavily biased 90% of the time. That's definitionally backwards.

If I flip a coin thirty times and I expect somewhere around fifteen heads and fifteen tails, I do not expect twenty eight heads and two tails and therefore suggest we should just get rid of the coin toss as a random function.

Why should you expect any one particular result? Assuming each coin toss is non-dependent upon any of the others, 28-2 and 15-15 is equally likely, along with all the other possible combinations of 30 flips. It's only when you start looking at dependent results (ie, runs of heads in a row) do the odds start to vary. If we toss the coin thirty times, and repeat this set of flips 50 times over, and every set of 30 is around 15-15, then I would contend that is anything BUT random.

So to answer your original post. Yes, "random" in computers means something different that in statistics. Most computers cannot create a random number. It just can't. And if something they did with the AD mod affected the seed generator (or more likely, fudged up a line of code somewhere), then the RNG may have a bias installed.

[Kerbart]

Apple had a similar problem with their shuffle feature on iPods and iTunes. If you have 5 songs from the same artist in a playlist of 1000, it seems odd to have 3 in a row come up in shuffle. People complained about the seemingly non-random nature of it, and how songs would cluster up, and some days it would seem to have a "favorite" artist.

But that's the nature of randomness. It creates false patterns, which our brains are very good at detecting. The human brain is designed to easily find patterns, not randomness.

No really, that is a bug in iOS. I get the same artist three, four times in a row all the time, at least twice each commute. "The nature of randomness?" It never happens when I play music through Foobar on my PC. And they use exactly the same playing list.

On the selection of stranded kerbals: my male crew is not complaining.

[NecroBones]

Ooh boy, and that can be a whole separate discussion in itself-- What does "random" mean from a computer algorithm point of view, and how do you approximate "randomness" as much as possible? If this were a discussion in a cryptography community, there could be long drawn out discussions and arguments about this, I'm sure.

The good news is that (at least in Linux anyway), you can get some pretty good randomness these days, with accumulated "entropy" from a variety of sources. But traditionally, all you get is pseudo-randomness, in a completely predictable and repeatable sequence based on the seed value.

- - - Updated - - -

No really, that is a bug in iOS. I get the same artist three, four times in a row all the time, at least twice each commute. "The nature of randomness?" It never happens when I play music through Foobar on my PC. And they use exactly the same playing list.

On the selection of stranded kerbals: my male crew is not complaining.

Well part of the reason I mentioned Apple specifically is that they defended it as being random. But part of what you're seeing is that it only shuffles (with a chosen seed) "once", and doesn't re-shuffle when you stop, start, or skip around. Other players behave differently, or use sanity checking of some kind. It doesn't mean Apple's approach is broken, it's just not what you expect. It has to be made less random to sound more random, which is counter intuitive.

[Tex_NL]

...

and with that - I do seem to have more girl Kerbals than boys - I know it's random... but problem is -- it really IS random, where perhaps it shouldn't be... it works for human populations, sure... with 7 billion samples, your statistics are bound to converge...

...

Problem is, it ISN'T random. The game is biased to giving almost only females. And you are biased to think SQUAD didn't frag-up.

With the Asteroid Days mod, mine is not a normal possible outcome of all females. It is, with any new career, 100% female, always, no exceptions. I disabled the mod, in the careers with no males, then got... males, immediately. I tested this in multiple careers. Others have observed the same thing. So the question is, does everyone observing this have Asteroid Days installed?

I have noticed this as well-- all of the rescuees are female. It's not as important to me as what their profession is, as I like to have a mix of pilots, scientists, and engineers, but I did notice it as an oddity and wondered if it was the "rescue the damsel" thing. I do have Asteroid Day installed.

And there are many more reports just like those.

Getting 20 females in a row completely random and unbiased is indeed not impossible, just highly improbable. About one in a million. Please explain why literally dozens of people have reported this behaviour. If chances are indeed about one in a million only one or two KSP players world wide should have experienced this.

The game IS biased. This behaviour IS NOT random. This IS a bug.

[Pax Kerbana]

It doesn't matter what people think the code should do or will do, only what it does do. And I can tell you that every single rescue contract I've ever gotten or taken has been for female Kerbals. I checked last night.

The astronaut center applicants are equal distribution.

Something is wrong despite people wanting it to be right.

I have never run a mod except for urbizio welding and that is not a factor because that was removed for 1.0 and is not installed.

[Tex_NL]

...

I have never run a mod except for urbizio welding and that is not a factor because that was removed for 1.0 and is not installed.

But you do have the official Asteroid Day 'mod' installed. Right?

[Hotaru]

So here's my question: has anybody had a save with predominantly or exclusively male rescuees?

[Pax Kerbana]

But you do have the official Asteroid Day 'mod' installed. Right?

No. I only have KSP update patches. Is Asteroid Day something I would have to download separately? If so, no.

[tater]

No. I only have KSP update patches. Is Asteroid Day something I would have to download separately? If so, no.

What OS are you on?

Actually, anyone who has had this problem, what OS?

[Pax Kerbana]

What OS are you on?

Windows 7...

[snip]

I have had only 3 male Kerbals in my career roster: Jeb, Bill, and Bob. I think I'll spend a ridiculous 300k+ roots on a male hire just in the spirit of affirmative action.

Ooh I know what I'll do... I'll just send the rescues I don't need to Eve.. Hehe

Edited September 12, 2015 by Vanamonde

[nsgallup]

In one of my saves Val turned into a male

[Goddess Bhavani]

In one of my saves Val turned into a male

Can't unsee this. Rename to Valentino Kerman

So here's my question: has anybody had a save with predominantly or exclusively male rescuees?

Me. Possibly due to installing a contract pack mod. I'm about to change some Kerbals manually via persistence or would anyone here like to help me create a handmade kerbelle space program?

[Mr. Speed]

Wow, so i'm not the only one. At first it was just like flipping a coin and it lands on heads a few times in a row, but it seems more and more like there is some sort of bias in the contracts. I must have had atleast 20 rescue contracts pop up, all damsels in distress.

Yes, I have the asteroid day mod installed.

[nsgallup]

Can't unsee this. Rename to Valentino Kerman

I just pretended she didn't age well

[FyunchClick]

I get the feeling parkaboys graphic novel mission report has run into the same bug.

[Jovus]

Wow, so i'm not the only one. At first it was just like flipping a coin and it lands on heads a few times in a row, but it seems more and more like there is some sort of bias in the contracts. I must have had atleast 20 rescue contracts pop up, all damsels in distress.

Yes, I have the asteroid day mod installed.

Chiming in here to say that it's the exact same thing with me - not a single male Kerbal outside the Big 3 in my game.

I feel like there's a joke somewhere in the fact that there are all these women who got stranded in space...

[NoPanShabuShabu]

Why should you expect any one particular result? Assuming each coin toss is non-dependent upon any of the others, 28-2 and 15-15 is equally likely, along with all the other possible combinations of 30 flips. It's only when you start looking at dependent results (ie, runs of heads in a row) do the odds start to vary. If we toss the coin thirty times, and repeat this set of flips 50 times over, and every set of 30 is around 15-15, then I would contend that is anything BUT random.

So to answer your original post. Yes, "random" in computers means something different that in statistics. Most computers cannot create a random number. It just can't. And if something they did with the AD mod affected the seed generator (or more likely, fudged up a line of code somewhere), then the RNG may have a bias installed.

All the stuff about RNGs aside, you would definitely expect to get something close to 15-15. The probability of getting between a 10-20 to 20-10 is about 95.7%, which are pretty good odds. There are a "whole lot" of ways that 30 tosses of the coin will get those more even distributions, and that number goes down the farther you get from 15-15. There is only 1 way to get 30-0, and only 1 way to get 0-30. If you did a series of tosses and got 15-15 all the time, sure that's definitely a problem, but on the other hand, if you did 50 sets of 30 and did not get any 15-15 splits, then that would very likely be a problem too. The odds of that happening are something like 0.04%, which is possible, but not likely.

Edit: BTW, I've got about 50 females in a row right now.

- - - Updated - - -

I recently rescued a female Kerbal named "Danny". I realize now that I was prejudiced in thinking that the Youtuber was a male. I can't lie, I was apprehensive when the name came up to be rescued, thinking that maybe something had gone wrong in reverse, but luckily the streak remained unbroken.

As I mentioned in the edit above, I've got about 50 females going in the two careers I'm running (hard and moderate RSS.)

Edited September 13, 2015 by NoPanShabuShabu