100 shards drop: Myth or Reality?

[pozi]

hace 20 horas, Observer_X dijo:

In my experience, RNG in this game has shown to work differently for different players, and IMHO that is an unfair thing.

I can say the same about national lottery. That's what randomness is all about

hace 9 horas, Kenrae dijo:

RNG, by definition, is random. I prefer games to be less random, but that's a matter of preference, not fairness.

The shards system was an improvement over the previous one, which was much more random. I think the system should be this one: you always get some shards, the same ammount you get now when it triggers, but you need many more shards. 1000 or 10000 or whatever the number you need to have the same average. You'd have the same average but less variance between players, and you'd have something else: a mathematical maximum number of battles to get the girl, which would make people use more kobans actually, because it's easier for anyone to use them for something that will eventually happen instead of something that _might_ happen.
I think it would be a win-win situation for both players and Kinkoid.

That system was already discussed when shards were implemented. It has some flaws: first, some players will never invest in kobans as they know for sure they just need to wait and do a certain number of battles to get all the girls. And other group of players will spend tons of cash now to get all the girls as soon as possible, then wouldn't have any incentive to keep playing the game. The randomness of shard's drops combined with an exact amount of shards (100) needed to get a girl creates a reasonable balance, although variance can piss some players off in some events

Edited June 28, 2020 by pozi
avoid spamming

[Kenrae]

hace 1 hora, pozi dijo:

I can say the same about national lottery. That's what randomness is all about

That system was already discussed when shards were implemented. It has some flaws: first, some players will never invest in kobans as they know for sure they just need to wait and do a certain number of battles to get all the girls. And other group of players will spend tons of cash now to get all the girls as soon as possible, then wouldn't have any incentive to keep playing the game. The randomness of shard's drops combined with an exact amount of shards (100) needed to get a girl creates a reasonable balance, although variance can piss some players off in some events

I've seen it in other games and it works much better in practice, not the way you're describing at all. I fail to see why this game would be different. Those big spenders you're describing already have all the girls anyway, they keep playing because they keep adding new girls and new content.

[Observer_X]

On 6/28/2020 at 8:52 PM, pozi said:

I can say the same about national lottery. That's what randomness is all about

No offense my friend, but if you truly think that the RNG of this game is akin or even comparable to the one of the national lottery, I truly don't understand what is that keeps you playing the game: a similar comparison makes me think you should feel that playing this game would be only an absolute and absurd waste of time.

Mathematical matters apart, just the fact that the national lottery is a single extraction made over long intervals of time, while in this game the RNG is involved in most of the major implementations inside the game, and therefore used by each player a number of times each day, makes the two things deeply different. It is the fact that the RNG in this game is so heavily used that allows people to talk about "average behaviour" regarding a relatively short period of time (anyway shorter that the one of the national lottery). Although I am not an expert, in my school days I have studied statistics enough to know that a normal RNG should have the same average behaviour with every player. It's obvious that heavy koban users have more tries at their side and therefore (always on average) should have obtained more girls with a lucky 100. The unfairness I mentioned is relative to the fact that, taking as a reference the 25k tries of a free player, with that number of tries made in almost the same time, any member of any number of players should have more or less obtained the same number of girls. As this is on average, for example (just making numbers here) if the majority of those players have gained 6 girls in that period with that number of tries, it is quite normal that there are players with 5 or 7 girls, but things start to get towards zero probability when one examines that group of players in search of players with only 2-3 girls or even 9-10 girls. Now, it's obviously possible that many free players weren't interested enough to post their lucky 100 drops in this thread, but at first glance, the results I have seen here tell me what I had already said above: this game RNG don't act the same for every player.

[_shal_]

On 7/3/2020 at 7:42 PM, Observer_X said:

The unfairness I mentioned is relative to the fact that, taking as a reference the 25k tries of a free player, with that number of tries made in almost the same time, any member of any number of players should have more or less obtained the same number of girls.

25,000 rolls aren't nearly enough for that with an apparent probability of 1/5000 or even less.

[pozi]

En 3/7/2020 a las 19:42, Observer_X dijo:

No offense my friend, but if you truly think that the RNG of this game is akin or even comparable to the one of the national lottery, I truly don't understand what is that keeps you playing the game: a similar comparison makes me think you should feel that playing this game would be only an absolute and absurd waste of time.

Mathematical matters apart, just the fact that the national lottery is a single extraction made over long intervals of time, while in this game the RNG is involved in most of the major implementations inside the game, and therefore used by each player a number of times each day, makes the two things deeply different. It is the fact that the RNG in this game is so heavily used that allows people to talk about "average behaviour" regarding a relatively short period of time (anyway shorter that the one of the national lottery). Although I am not an expert, in my school days I have studied statistics enough to know that a normal RNG should have the same average behaviour with every player. It's obvious that heavy koban users have more tries at their side and therefore (always on average) should have obtained more girls with a lucky 100. The unfairness I mentioned is relative to the fact that, taking as a reference the 25k tries of a free player, with that number of tries made in almost the same time, any member of any number of players should have more or less obtained the same number of girls. As this is on average, for example (just making numbers here) if the majority of those players have gained 6 girls in that period with that number of tries, it is quite normal that there are players with 5 or 7 girls, but things start to get towards zero probability when one examines that group of players in search of players with only 2-3 girls or even 9-10 girls. Now, it's obviously possible that many free players weren't interested enough to post their lucky 100 drops in this thread, but at first glance, the results I have seen here tell me what I had already said above: this game RNG don't act the same for every player.

Again, RNG is the same for every player but it won't offer the very same results to all of them. That's RNG, and averages are just that, averages. We don't know what's the variance.

Even in a normal distribution there will players on both extremes of the Gauss bell, what we can call outliers (and it seems you are one of those, bud luck my friend). The probably of those outliers wanting to express here in this forum either their frustration or their happiness with the results of their tries is, I think, quite high (especilly for the 'offended' ones). I mean, I started this very same thread trying to understand if there were some hidden mechanism behind the shards functionality because I had the feeling there was something odd on how the first 100 shards drop were distributed. It was made clear quite soon that there isn't, just pure RNG. Anyway, don't take any info given in this forum as something directly extrapolable to the whole base of players. Only a few of them (no more than 50, I might say) keeps posting here on a regular basis so there's too much data unavailable to us, even with the great effort some of them try to make sharing their own experience.

So no offense my friend, but if you truly think RNG is not acting the same for every player, and assuming you feel you are in the 'wrong' side of it,  I truly don't understand what is that keeps you playing the game (and I mean it, because I've seen several posts from you in other threads complaining about how awful and unfair this game is for free players - which I also am, by the way,, and I'm not having that bad feeling you seem to have). 

 

Edited July 5, 2020 by pozi

[Magic1986]

I got Clerk Delicia yesterday. I already had 91 Shards and got a 100 shard drop.

[FinderKeeper]

[Zteev]

On 7/3/2020 at 1:42 PM, Observer_X said:

No offense my friend, but if you truly think that the RNG of this game is akin or even comparable to the one of the national lottery, I truly don't understand what is that keeps you playing the game: a similar comparison makes me think you should feel that playing this game would be only an absolute and absurd waste of time.

 

It is honestly comparable, because it's also (generally) a random number distribution. (Note: There are ways to fix a large lottery, and it can be done physically by doing things with the balls, among other ways.) The big difference is that in any single lottery result, all players are using the same exact random number that is generated, while in this game players using the same random result is rare, limited to things like filling contest groups, and even that may be not exactly how they do it. 

On 7/3/2020 at 1:42 PM, Observer_X said:

Although I am not an expert, in my school days I have studied statistics enough to know that a normal RNG should have the same average behaviour with every player.

That's generally true but...

1. It won't be the exact same average, it will be clustered in a range. If you believe that it will be exactly the same it's time for you to roll groups of 120 6 six-sided dice as a trial 10,000 times, and see how close you get to the same behavior.

2. This is true only in large enough sample sizes, and it truly depends on how low the probabilities involved are. Once you get into very low probabilities of events, the number of events needed to generate a single instance of something occurring increases, and that increases the number of trials needed massively.

[GeorgeMTO]

(╯°□°）╯︵ ┻━┻

[Observer_X]

On 7/5/2020 at 6:36 PM, pozi said:

So no offense my friend, but if you truly think RNG is not acting the same for every player, and assuming you feel you are in the 'wrong' side of it,  I truly don't understand what is that keeps you playing the game (and I mean it, because I've seen several posts from you in other threads complaining about how awful and unfair this game is for free players - which I also am, by the way,, and I'm not having that bad feeling you seem to have). 

The answer to this is in this post.

[bane]

timing could have been better

 

[LuxFF]

Free money. (or affection)
I got a Lola while farming for event girl drops, 0-100.

[Master Edward]

Just today began to fight for the Chamelea. Collected 18 shards. The another battle, I press the button and see the “5000 xp” notification, in a couple of milliseconds I understand that it turns out to be 100 shards 😀. Last time I was so lucky in January: I was able to get two girls the same way.

[_shal_]

My 10th full drop.

[GeorgeMTO]

😐

[ChyldeRoland]

[blaa]

my first girl this event, I am glad, I hadnt had to fight her for 400 times and got just over with it, and honestly, that's kinda anniversarly for me. I proudly present you:

I

If my record is correct, it's my 3rd out of 5 100 drop on fredy - you know, what this means, right?

[Тёмный Властелин]

45 minutes ago, blaa said:

If my record is correct, it's my 3rd out of 5 100 drop on fredy - you know, what this means, right?

Your destiny is to be a teacher? 😀

[jelom]

A perfect gift for an anniversary. I had a lucky x100 with just three battles. 

My ninth girl dropped with x100: Evelin, Sara Jay, Ombresse, Ria, Arianne, Dickachoo, Athena, Shade and now Winter Sophie.

[blaa]

and another 100 drop, after ~70 more refills belinda dropped with 50 shards (unfortunately i couldnt save the drop animation, just this bad screen)

keep going 100 drops, I really like you

Edited July 17, 2020 by blaa

[blaa]

mh, this time I only got the animation, after 167 fights I got another 100 drop, I am now on 7 100 drops in total

again: keep going 100 drops, one girl left

[Cartageno]

On 7/3/2020 at 7:42 PM, Observer_X said:

Although I am not an expert, in my school days I have studied statistics enough to know that a normal RNG should have the same average behaviour with every player. It's obvious that heavy koban users have more tries at their side and therefore (always on average) should have obtained more girls with a lucky 100. The unfairness I mentioned is relative to the fact that, taking as a reference the 25k tries of a free player, with that number of tries made in almost the same time, any member of any number of players should have more or less obtained the same number of girls. As this is on average, for example (just making numbers here) if the majority of those players have gained 6 girls in that period with that number of tries, it is quite normal that there are players with 5 or 7 girls, but things start to get towards zero probability when one examines that group of players in search of players with only 2-3 girls or even 9-10 girls. Now, it's obviously possible that many free players weren't interested enough to post their lucky 100 drops in this thread, but at first glance, the results I have seen here tell me what I had already said above: this game RNG don't act the same for every player.

I won't comment on the RNG, I have no reason to believe it is anything but random but I haven't run an experiment to be adequately sure. However, if I understand correctly what you wrote, there is a common misunderstanding about the law of large numbers included.

Given a perfect randomness, the average will indeed hone in on the expected value. Given 100 fair coin tosses, averages will range roughly (2 sigma equalling 95% confidence) from .4 to .6 times head on average. With 10,000, the same confidence interval ranges from .49 to .51.

However, in absolute terms this means 40 to 60 in the first case, a difference of 20; but 4900 to 5100, a difference of 200, in the second, so in absolute terms, the ones doing best will be further away from the unlucky ones.

Both aspects are bound by the square root of the number of tries: From 100 tries to 10,000 tries I have 100 times more, square root of 100 is 10, so the relative deviation will shrink to 1/10th of its former size, the absolute deviation however will be 10fold of what it was before.

Actually quite logical, when you think about it: with only one coin toss, best and worst cannot be apart more than 1, while with a billion tosses a difference of, say, a dozen wouldn't raise anybody's eyebrows.

In terms of this game it means: the more battles you have, the more the unlucky ones will fall behind in terms of 100 shard drops. (They don't have to be the same unlucky ones as before though, that can, and almost certainly will, change, although not in a "from last to first" way)

Edited July 22, 2020 by Cartageno
grammar

[Cartageno]

On 7/6/2020 at 12:18 PM, Zteev said:

Once you get into very low probabilities of events, the number of events needed to generate a single instance of something occurring increases, and that increases the number of trials needed massively.

Correct. By the way, the same holds for very high probabilities (since they have very low probabilities of "not happening"). There actually a formula for that, and with a rule of thumb accredited to mathematician Leibniz, to get results approaching the "normal distribution" and thus being calculable with common tools, the number of tries you need should exceed

n>9/(p^2*q^2) (more than 9 divided by both p squared and q squared, where p are the odds of something happening and q are the odds of it not happening. Obviously p+q=100%, since things will always either happen or not, and 100%=1, since "percent" is nothing more than a fancy way of saying "out of a hundred/hundreth", and 100/100 is 1. Note that since both p and q are between 0 and 1, dividing by them or their squares will actually increase the number you started with instead of the usual decrease we see in everyday life when divvying stuff.)

So given the above estimate of 1 in 5000, that'd be a whopping 225 million and change tries you need to do before you can properly analyze it. However, I think that number is too pessimistic, but even at a 1/500 chance we'd still need 2.26 million. (Bonus: note how an roughly 100fold increase in tries allows us to check for a 10times lower probability - square roots again, though not exactly, since here also the odds of not getting the shards are included, which are similar, but not exactly the same).

As a conclusion, we can also see that the number mentioned further above in the thread, 25,000 tries, will roughly be enough to check for odds of 1/50 (or better). For worse odds, there is no practical way to check the RNG with our results, even if everybody wrote theirs down and posted them.

Edited July 22, 2020 by Cartageno
missing verb added

[Zteev]

15 hours ago, Cartageno said:

Correct. By the way, the same holds for very high probabilities (since they have very low probabilities of "not happening"). There actually a formula for that, and with a rule of thumb accredited to mathematician Leibniz, to get results approaching the "normal distribution" and thus being calculable with common tools, the number of tries you need should exceed

n>9/(p^2*q^2) (more than 9 divided by both p squared and q squared, where p are the odds of something happening and q are the odds of it not happening. Obviously p+q=100%, since things will always either happen or not, and 100%=1, since "percent" is nothing more than a fancy way of saying "out of a hundred/hundreth", and 100/100 is 1. Note that since both p and q are between 0 and 1, dividing by them or their squares will actually increase the number you started with instead of the usual decrease we see in everyday life when divvying stuff.)

So given the above estimate of 1 in 5000, that'd be a whopping 225 million and change tries you need to do before you can properly analyze it. However, I think that number is too pessimistic, but even at a 1/500 chance we'd still need 2.26 million. (Bonus: note how an roughly 100fold increase in tries allows us to check for a 10times lower probability - square roots again, though not exactly, since here also the odds of not getting the shards included, which are similar, but not exactly the same).

As a conclusion, we can also see that the number mentioned further above in the thread, 25,000 tries, will roughly be enough to check for odds of 1/50 (or better). For worse odds, there is no practical way to check the RNG with our results, even if everybody wrote theirs down and posted them.

I wanted to spare the mathematical details, but thanks for following up with more detailed math. I probably should've gone there in my original post.

I do think that if we had enough players doing it, we could get to the numbers of samples we need, but I suspect that the game doesn't have enough players in total to make that feasible in any reasonably short period - and during that the generation method could change. There are games where there are enough players generating results that one can generate very solid data analysis, though.

[Cartageno]

6 hours ago, Zteev said:

I wanted to spare the mathematical details, but thanks for following up with more detailed math. I probably should've gone there in my original post.

I do think that if we had enough players doing it, we could get to the numbers of samples we need, but I suspect that the game doesn't have enough players in total to make that feasible in any reasonably short period - and during that the generation method could change. There are games where there are enough players generating results that one can generate very solid data analysis, though.

This is generally true, and I play MMOs where players pool their data to get odds from lockboxes or similar an we have pretty solid ideas of non-published probabilities. However, since the assumption in question was that odds will differ depending on payment in the store, the only odds we really could use are those of f2p players who will have the same amount spent (0), while I guess only very few players have a complete list of how much they paid, possibly often not even knowing a good ball park figure, and it would be unlikely for them to be the same or at least so close we can assume them to be practically the same.