Bonus stage is so bad

**YoshiEnVerde** · 07-15-2015

[MENTION=101]ChibiKika[/MENTION], you make a good point.
Problem is that every verifiable use case you add to a thesis/experiment its complexity grows exponentially.

Just making general prob. math for this will need a dozen of us loading data for a month.
Taking different timezones, access levels, and player dedication would need for us to triple, or more, the amount of data gathering, and at least double de data collation part.

I have the same 8hs for sleep, and 9hs for work, with an extra 2hs of home<>office daily travel. And that doesn't take into account time for eating, cleaning the house, etc (I'm single).
The point about this kind of game is that you don't need to be logged in 24/7 (unless you're fighting for a high place in an event). You can log in every 2~4hs (depending on your level) and just burn the stamina.

That's another way I try to keep my data gathering the most sterile as possible: If I log in every couple hours and burn all the stamina regenerated, the data will be spread amongst as many spin-configs as possible.

**Opalia24** · 07-16-2015

I find it hard to find any clear patterns in my data:
Here I show a sample of my data. It is my data from area 5-3. You can see if you can find pattern I did not see.

Normally when I do not have holidays, I lock in at 6:00, 7:40, 17:30, 20:00, 23:30 so data from many time periods.

My data from area 5-3 summarized.
C 133 times
P 65 times
G 130 times
E 113 times
N 806 times
L 2 times
Sum 1249 spins

C 10,7%
P 5,2%
G 10,4%
E 9,1%
N 64,6%

Spot layout:

1 2 3
4 5 6

I always picked girl in spot 2. It have been best spot in other test.

Number of N,R and SR-girls seen in the 6 spots:

Spot1: N-Girls: 107 R-Girls: 26 SR-Girls: 0
Spot2: N-Girls 97 R-Girls 33 SR-Girls: 3
Spot3: N-Girls 104 R-Girls: 20 SR-Girls: 9
Spot4: N-Girls 103 R-Girls: 26 SR-Girls: 4
Spot5: N-Girls 99 R-Girls: 30 SR-Girls: 4
Spot6: N-Girls 99 R-Girls: 30 SR-Girls: 4
Sum N-Girls 609 R-Girls: 165 SR-Girls:24

Procentage of girls seen that are: N, R and SR:
N-Girls 76,3% R-Girls: 20,7% SR-Girls: 3,0%

(In N-Gacha around 1.5% of girls are SR. 3x Chance seems to have around double SR chance as a N-Gacha)

N=Nothing
E=3x Exp
P=3x Pero
G=3x Gift
C=3x Chance(R spots that have rare girls, RS: spots that have SR-girl)
C(R:5,SR:2) = Rare girl in spot 5, SR-girl in spot 2.
L=3x Level

C(R:5,SR:0) = Means photos shown but no SR-Girl. Happened once.
If there are a SR-girl photos are always shown.

Raw data:

->Open Spoiler<-

- - - Updated - - -

20 More spins at bonus stage:
1x triple Exp
1x triple Gift
1x triple Chance
17x Nothing

This data fit perfectly with the data I already got.

At this speen it will take 18 * 20 =360 spins = 3600 Stamina just to finish all Chance quests

Players get in 24 hours just 1440 stamina. So players need tons of luck to finish all Chance quests without using items at bonus stage.
Unless data sudently change Bonus Stage Suck big time.

- - - Updated - - -

22 More spins at bonus stage:
1x triple Exp
2x triple Gift
3x triple Chance
16x Nothing

Better this time but only good enogth to finish 2nd Chance mission before reset at 0:01 AM.

- - - Updated - - -

I looked at the data from many1000s of spins I have recorded so far.
I cannot find any proof that chance change over time from large data sets.
In the end it is just pure luck. From time to we all got a good day.
No spots are better for picking than others long run.
Players get proximately the SR-girl 1 in 6 when one of the 6 cards is a SR-Girl.
Human seem to forget (ignore) loose easily but remember wins.
This might help to explain why so many loose so on much gambling – They feel they really win-ing but they are in fact losing. People look too closely will find patterns.
I found them many times – each time I made a theory and the started to collect new data to test it.
So far all my theory about some spots to be better or game to be anything but just random have failed.

I might be wrong but that is what data tells me so far.
I would be happy if anyone could prove me wrong.

18 More spins at bonus stage:
0x triple Exp
1x triple Gift
0x triple Chance
1x triple Pero
16x Nothing

3 hours of spins and not a single triple chance.!!!

22 More spins at bonus stage:
3x triple Exp
0x triple Gift
0x triple Chance
2x triple Pero
15x Nothing

**ChibiKika** · 07-16-2015

@YoshiEnVerde I do realize how much complexity the consideration of an RNG alteration by time period would add to the equation itself but unfortunately I have to emphasize that if such a thing were in fact programmed into the system and those gathering data chose to ignore it, the conclusion from rendered data would undoubtedly be inaccurate if not completely false. For that matter, it would render most of the statistical data gathered thus far worthless.

That being said, in positive hopes that is not the case, it would be desirable to hope otherwise and only come to that conclusion once enough data has been gathered to determine that there is in fact such an anomaly not accounted for. All I'll ask is to set it aside but take it into consideration if something seems off down the road.

And admittedly yes, I'm not very meticulous when it comes to testing statistics (was more of a calculus guy) so everything I say is more or less unproven hypothesis, in some cases one that's like religion in that it can neither be proved nor disproved. Ultimately it will become little more than a plausibility to consider if all other statistical conclusions seem to fall short somewhere.

**Opalia24** · 07-16-2015

The "Happy hour theory": Some hours or periods have bigger success chance than others.
Example: At some periods, triple Chance might be 5% and other 15% with an average 10%.
How to test this? My suggestion look at waiting times. Number of spins from between one triple chance to the next.
If chance of triple "Chance" is, constant waiting times will follow an exponential distribution.
If happy hour theory is correct we could get more very short and very long waiting time.

I also look at how often did SR come in same spot 2 time i a row:
4 out of 24 times or just 1:6. Not much better than the 1:6 chance you get by picking at random.

**YoshiEnVerde** · 07-16-2015

Originally Posted by ChibiKika

@YoshiEnVerde I do realize how much complexity the consideration of an RNG alteration by time period would add to the equation itself but unfortunately I have to emphasize that if such a thing were in fact programmed into the system and those gathering data chose to ignore it, the conclusion from rendered data would undoubtedly be inaccurate if not completely false. For that matter, it would render most of the statistical data gathered thus far worthless.

That being said, in positive hopes that is not the case, it would be desirable to hope otherwise and only come to that conclusion once enough data has been gathered to determine that there is in fact such an anomaly not accounted for. All I'll ask is to set it aside but take it into consideration if something seems off down the road.

And admittedly yes, I'm not very meticulous when it comes to testing statistics (was more of a calculus guy) so everything I say is more or less unproven hypothesis, in some cases one that's like religion in that it can neither be proved nor disproved. Ultimately it will become little more than a plausibility to consider if all other statistical conclusions seem to fall short somewhere.

Actually, it does not become innacurate (or false) as long as you sample a broad enough data set.
That's the point of doing the data gathering for a long time, at as many different hours as possible, by as many people as possible: If you were to get an hypthetical data set for each and every "timezone" of the algorithm, then you'd be working with "perfect" statistical equations (averages and such).
It is a true fact from statistical math that once the data universe you're starting from becomes big enough you could never gather a trully perfect representation in your sample.That's why you try to gather from as representative a set you can.

- - - Updated - - -

Interesting values for my first two days of data gathering.
I'll post everything together after the first week of gathering.

What I can see, however, still shows BA to not be that bad as was originally implied...

->Open Spoiler<-

As I said above, I won't post any definitive thoughts on this until next week, when I'll have more solid data (I could as well get less than 5 Chance Time spins next BA run through).
Then, I'll collate the data loaded by other people too.

**Opalia24** · 07-16-2015

Now i got 171 spin at gatheres at 8 different times:

3x Chance 9 times (5.3%)
3x Pero 10 times (5.7%)
3x Gift 8 times (4.7%)
3x Exp 7 times (4.1%)
Nothing 137 times (80.1%)
Total 171 spins.

My estemate yesteday was 5% for each of the 4 triple results so even though my data tripled no big changes.

I can only collect 22 spins each time then I have to wait 3 hours.

Last 4 times I colleted 20-22 same poor result. Data collected 4 different times with 3 hours between all show 80% nothing spins. Waiting 3 hours don't seem to change proberbilities.

Spin cost should at bonus stage should be reduced to atleast 5 stamina or maybe even to 3 stamina.
Even with a spin cost of 3 stamina area 5-3 is still better.
Spin cost of 2 stamina is too low imo and would make it overpoweres.

**YoshiEnVerde** · 07-16-2015

Originally Posted by Opalia24

Now i got 171 spin at gatheres at 8 different times:

3x Chance 9 times (5.3%)
3x Pero 10 times (5.7%)
3x Gift 8 times (4.7%)
3x Exp 7 times (4.1%)
Nothing 137 times (80.1%)
Total 171 spins.

My estemate yesteday was 5% for each of the 4 triple results so even though my data tripled no big changes.

I can only collect 22 spins each time then I have to wait 3 hours.

Last 4 times I colleted 20-22 same poor result. Data collected 4 different times with 3 hours between all show 80% nothing spins. Waiting 3 hours don't seem to change proberbilities.

Spin cost should at bonus stage should be reduced to atleast 5 stamina or maybe even to 3 stamina.
Even with a spin cost of 3 stamina area 5-3 is still better.
Spin cost of 2 stamina is too low imo and would make it overpoweres.

I'm online pretty much 24/7, so I run my stamina burns in 50~100sta bursts every hour or so. The different times do change numbers.
Instead of a 5% triple-spin overall, and 80% failure, like you got, mine are around 12% Chance, 5% The rest, 75% failure.

I can tell you that most of those CHance triple-spins actually rolled in two different bursts, at a rate of 5~6 Chances in 10 or less spins.

**Tenhou** · 07-16-2015

Okay, i have to say that i am getting somewhat frustrated at explore at the moment. Currently i am sitting in area 7-3 waiting for that lucky time that i get the SR before i move on. However, the SR can be gotten from the thursday daily as well, and with a higher chance to get the chance triplet. Currently pondering whether i should stay in 7-3 until i get it or just move on and get to GHQ asap because this will be the second day in a row where i haven't been able to finish the 10x chance daily, something i never fail with in the GHQ area whenever i can do it. Severely hurts my Kurito attempts...

So any suggestions? Blast away or stay in 7-3 for now?

**chillinfar** · 07-16-2015

A little point. I saw that during chance time the rarest card of the deck (between N cards could be one that you don't own) drops 1-2 places for next roll (between gacha and explore runs). The only problem? It does in both directions (that's the random fact).

Well, is my sight.

**Skulkraken** · 07-17-2015

Tenhou, just move on and get her from the GHQ later. The GHQ doesn't just seem to have a higher chance of hitting Chance Time, it seems to hit on longer pseudorolls than normal exploration, which raises the rarity of the cards that show up, step ups or not.

Thread: Bonus stage is so bad

PeroPero

PeroPero Seduction Integration

Thread Tools

Rate This Thread

Display

Posting Permissions