Thread: Share Your Fortune/Misfortune

from a few days draw

After 7 months with no SSR drawn, I've changed the way I draw Jewel Gacha. I'm taking them 1 at a time, never 10, and always at server quiet-times. Since then, I've drawn 2 SSR, which is roughly the right proportion (first 18 months playing, 22 SSR Eido and 20 SSR Kami; roughly 1 SSR every 2 weeks; switched to 1 at a time off-peak 4 weeks ago).

There's something called "running out of entropy" that can happen to busy servers that use random functions frequently. It can cause a significant change in drop rates - if a Linux server runs out of entropy, all random numbers generated are zero until new random seeds are generated. Recently learned about this in my new job, and am wondering if this is a thing in Kamihime Project. It's just anecdotal evidence & theory though. If I cared, I'd try to get broader data - rates of SSR drop on draws of 10 and draws of 1 with a server time associated, see if there's any correlation. But that sounds like effort, I have no idea if anybody else here is interested and don't produce enough data myself.

Originally Posted by Chrestomancy

After 7 months with no SSR drawn, I've changed the way I draw Jewel Gacha. I'm taking them 1 at a time, never 10, and always at server quiet-times. Since then, I've drawn 2 SSR, which is roughly the right proportion (first 18 months playing, 22 SSR Eido and 20 SSR Kami; roughly 1 SSR every 2 weeks; switched to 1 at a time off-peak 4 weeks ago).

As far as we can tell, single and 10 pulls have no difference in SSR drop rate.

What you are fucking yourself over with is SR drop rates, though. Since 10 pull guarantees you a SR (assuming that slot doesn't drop you an SSR). There's plenty of decent SRs in game, useful for Tower at the very least.

Originally Posted by Chrestomancy

--
There's something called "running out of entropy" that can happen to busy servers that use random functions frequently. It can cause a significant change in drop rates - if a Linux server runs out of entropy, all random numbers generated are zero until new random seeds are generated. --

Kamihime is coded by incompetent monkeys, so it's possible. However, there's no way that the random function is the only part of the seed (if there even is a seed). Chances are the seed is the date and time, including milliseconds when the request reaches the server. Maybe a random part is added in too, which could be hit by that thing.

However, who is to say that your chances are LOWER when your seed has a 0 in it? As such, I don't think such a thing is necessary to even look into. Since we have no idea how the seed works or even what the results will be, it will never be enough to reach any kind of conclusion.

Hey Slashey

All good points. I'm simply starting with my observed "no SSR drawn in 7 months", which represents over 300 jewel gacha draws, and something like 50 premium tickets. So yeah, it's utter conjecture - zero on the table could be "give out an SSR", but if it were, then running out of entropy would result in multiple SSR in one draw. That could be true, but my sample shows basic R weapons/eidolons only so I'm positing that the SSRs are at the top-end of the random sample. It's entirely possible I was just very very unlucky; statistical samples of 1 can't be generalised.

It's also possible that they are using a custom random sequence, using a seed from server time. Again, that would mean I've just been unlucky. I'm positing that they're using the random function in Linux and assuming it gives them a random number. In all fairness, I know slot machine games do exactly this, and I would think they would be more sensitive to a scenario that could return all reels showing the same icon than a game where the win has no monetary value.

As to SRs, I'm not really struggling from lack of them - I have 94 out of a possible 130. I've got most of the starting SR kamihime weapons in full break mode from all the duplicates, and most of those 10 draws were just giving me more of the ones I already had. Again, I've had more SRs (about 8 I think) in the last 4 weeks than the previous 7 months, because almost every 10 chain I've drawn in the last 8 months has returned a single SR duplicate. Having drawn something like 50 times trying for each of the new character gacha when ther was something I wanted in it, it is noticeable I got none of the new SRs either. And that's a part of my suspicions - usually when a new event launches with a new character possible, there are always raids to join and any raid I start fills up and completes in a couple of minutes - suggesting busy server. Drawing on the busy time of the busy day is where this problem would manifest - if it exists. So now I'm checking that it's a quiet time (GMT 10:00 to 17:00 on a week night) before drawing anything, and waiting for day 2 or 3 of a new character release.

I'm aware of the previous statistical sampling; one of the things I really appreciate about this community - thorough dedication to research! If this is a thing, from my perspective it is a change that happened towards the end of 2018. I had no issue in 2017 or most of 2018, was drawing 10 chains and getting expected distributions - roughly one SSR every 3 10-chain drawn. Sometimes I went 100 draws with nothing, sometimes 2 SSRs in a single draw. That changed to nothing every draw, but took me a while to notice. If it is a change, it would mean hosting servers were changed around then.

So yeah, probably nothing for anybody to worry about. Probably just me looking for patterns where it's just a statistical anomaly. But I thought I'd post here, just in case there's other people seeing the same statistical anomalies, or someone wanted to run some tests of their own.

Yo Kitty, next time I have 3k jewels, I'll tell you to make my pull.

Originally Posted by Laventale

Yo Kitty, next time I have 3k jewels, I'll tell you to make my pull.

sir yes sir

THE SSR KAMIHIME DROUGHT HAS ENDED

Originally Posted by Chrestomancy

After 7 months with no SSR drawn, I've changed the way I draw Jewel Gacha. I'm taking them 1 at a time, never 10, and always at server quiet-times. Since then, I've drawn 2 SSR, which is roughly the right proportion (first 18 months playing, 22 SSR Eido and 20 SSR Kami; roughly 1 SSR every 2 weeks; switched to 1 at a time off-peak 4 weeks ago).

There's something called "running out of entropy" that can happen to busy servers that use random functions frequently. It can cause a significant change in drop rates - if a Linux server runs out of entropy, all random numbers generated are zero until new random seeds are generated. Recently learned about this in my new job, and am wondering if this is a thing in Kamihime Project. It's just anecdotal evidence & theory though. If I cared, I'd try to get broader data - rates of SSR drop on draws of 10 and draws of 1 with a server time associated, see if there's any correlation. But that sounds like effort, I have no idea if anybody else here is interested and don't produce enough data myself.

I've observed something like this effect in other gacha games. They say to keep from pulling if you don't get something good in the first few tries and try again the next day. Perhaps the rate boost is both time-specific and gacha-specific and lowers drastically after you pull a particular SSR or after you've pulled a set number of times per day.

1 SR + 9 Rs on my account.
Ain't even mad. I'll edit this post since I'm gonna draw on my alts and see what they get... posting notable SSR only, just for funsies.

3rd account:


10th account:

->Open Spoiler<-

14th account:

->Open Spoiler<-

Isn't Gaia limited edition? She's in the f2p gacha pool now?

she's not on rate-up, but should still be in pool for some time