Since my last post:
2 x 3*
1 x 4*
Personal totals
62x 3* = 67%
21 x 4* = 23%
9 x 5* = 10%
Since my last post
3x 3*
1x 4*
Personal totals
32x 3* (80%)
8x 4* (20%)
0x 5* (0%) these are unicorns they don’t exist
2 Likes
Today I got a 3* and a 4*, I consider it still as my badluck streak, since the 4* is a duplicate 
3* 730 75%
4* 185. 19%
5* 58 6%
Since my last post:
2 x 3*
1 x 4*
Personal totals:
64 x 3* = 67%
22 x 4* = 23%
9 x 5* = 9%
2 x 3*
That’s my 2 trainings thus far 
A 3* and a 4* today…
3* 737. 75%
4* 190 19%
5* 58 6%
1 Like
Since my last post
1x 3*
2x 4*
Personal Total
33x 3*
10x 4*
0x 5* unicorns
First TC20 post:
3* = 0
4* = 1
5* = 0
(I got a duplicate Cyprian)
1 Like
Everyone here please concentrate, gather your energies and pray for me to pull a Grimm from my TC 20 next, thank you.
1 Like
And while we are all concentrating, I would like a Grimm too, let’s all train some Grimms…
Wouldn’t mind a Caedmon, Joon or Vivica either
3 Likes
Since my last post:
2 x 3*
1 x 4*
Personal totals:
66 x 3* = 67%
23 x 4* = 23%
9 x 5* = 9%
We didn’t concentrate hard enough, got 2 more 3* today.
Almost at 1000. Although the 5*s are at 6%, I am going to guess that if we take it to 5000 the numbers turn out to be 75, 20 and 5. Just makes more sense than 75, 19 and 6
If our sample statistic is 6%, then that should be our point estimate. 75, 20, 5 might be quite an appealing little group of numbers, but it’s just as easy to code 75, 19, 6. (And it looks like right now, our sample statistic is bang on 6%, not hovering between 5 and 6.)
Would someone like to work out a 95% confidence interval when we get to a sample size of 1000? I mean, someone that’s not me?
[Crickets.]
I’m going to have do dig out my old notes, aren’t I?
Yes, well, I agree, but I think the sample size is too small. It started out at 5%, peaked above 6 and is creeping back towards 5%. Sure, programming wise 19 and 6% are just as easy, but the developers themself are not computers and humans tend to go for round numbers. So, that is why I said I thought 75, 20 and 5 make more sense.
But I really think the sample size should be bigger to be sure. But I am good either way, don’t care if it is 5 or 6, I think those odds are both fine.
Not only we need a greater sample, but we need much more people posting the results.
As we are now, always the same people post their results, and we already see from the summon gate that some people are incredibly lucky and some other may be incredibly unlucky, and we don’t know if we have a good mix of that or more of a kind.
We never be sure, that’s already a good data as it is.
That’s why I think a confidence interval would be useful - it will tell us what level of precision our sample size gives us. With 6% being such a small sample stat, I suspect that you are right and we will end up with a wide CI. But I’m rusty on this stuff. Maybe 1000 trials is plenty.
Using @Talisax’s last numbers, not updated for later posts:
-
Using the Wilson Method, 95% Confidence Interval for the chance of training a 5* hero at TC 20 = 4.6%-7.6%
-
Using the Adjusted Wald Method, 95% Confidence Interval for the chance of training a 5* hero at TC 20 = 4.6%-7.5%
(I didn’t bother using the unadjusted Wald Method, because I think our sample p is too small for this. I might be wrong. So sue me.)
Personally, I don’t think this tells us much we didn’t know, though it does confirm that our sample size is still too small to distinguish between a 5% chance and a 6% chance, or even a 7% chance.
Edit: our point estimate, btw, is 5.9%.
2 Likes
I start play 1 month ago and spend 2 euro…I have x4 …4* and some 3*…