Pulling for Captain America - a cautionary tale

Hadronic

Orion said:

To answer the OP's question, LT draws are a binomial distribution with a 15% succees rate (supposedly).

*** Warning - math ahead ***

A binomial distribution has a mean of np, where n = 372 and p = 0.15.  So the mean is 55.8 as the OP stated.

The variance is np(1-p) or 372*0.15*0.85 = 47.43.

The standard deviation is the square root of that, or 6.89.

So 34 5*s out of 372 pulls is (55.8-34)/6.89 = 3.16 standard deviations away from the mean.

To be 3 standard deviations away from the mean happens 0.3% of the time.  But that's on the good AND bad side.  So to be that far off on the bad side happens 0.15% of the time.  But he was slightly worse than that.  So if the % is as D3 says it is, this should happen roughly 1 in 1000 tries.

tl;dr - OP got screwed.

So in my field, a 3 sigma result is enough to make people question the result. Start gathering more data and see if the significance increases or not.

5 sigma is needed to start making claims.

pupazzetto72

I read this thread only now. So I'm not alone. My pulls have been disastrous for 3 months now. A rate of about 1/20, very different from the 1/7 declared. I see around roster of casual players with jj and cap complete, and I have been playing non-stop for 4 years now I find myself with only 4 covers of jj and ZERO, yes ZERO !!!! cover of cap 5 *, I can not in fact play the next pve, being among the essential cap. I wrote many tickets to the d3 but, in addition to responding to the same phrases precompiled equal for everyone (I think the worst customer care of the whole store), they had the nerve to tell me that my percentage is above average. ..  . I understand you then @charliecrocker., I'm really unmotivated and I'm in fact thinking of quitting with this game. Know all that I have evidence, which I attached to the tickets, of SS in which the lords of the customer care bestow 5 * cover as compensation to the "Special" players. 

Hendross

Hadronic said:

Orion said:

To answer the OP's question, LT draws are a binomial distribution with a 15% succees rate (supposedly).

*** Warning - math ahead ***

A binomial distribution has a mean of np, where n = 372 and p = 0.15.  So the mean is 55.8 as the OP stated.

The variance is np(1-p) or 372*0.15*0.85 = 47.43.

The standard deviation is the square root of that, or 6.89.

So 34 5*s out of 372 pulls is (55.8-34)/6.89 = 3.16 standard deviations away from the mean.

To be 3 standard deviations away from the mean happens 0.3% of the time.  But that's on the good AND bad side.  So to be that far off on the bad side happens 0.15% of the time.  But he was slightly worse than that.  So if the % is as D3 says it is, this should happen roughly 1 in 1000 tries.

tl;dr - OP got screwed.

So in my field, a 3 sigma result is enough to make people question the result. Start gathering more data and see if the significance increases or not.

5 sigma is needed to start making claims.

Every answer in business school is, "It Depends".  Tolerance in process variability is a managerial debate and will differ if you're filling sand bags, filling pills, or distributing virtual comics with no intrinsic value.  Regardless I hate it for anyone who grinds and can't progress as fast as someone who is more lucky.

DD-The-Mighty

Hendross said:

Hadronic said:

Orion said:

To answer the OP's question, LT draws are a binomial distribution with a 15% succees rate (supposedly).

*** Warning - math ahead ***

A binomial distribution has a mean of np, where n = 372 and p = 0.15.  So the mean is 55.8 as the OP stated.

The variance is np(1-p) or 372*0.15*0.85 = 47.43.

The standard deviation is the square root of that, or 6.89.

So 34 5*s out of 372 pulls is (55.8-34)/6.89 = 3.16 standard deviations away from the mean.

To be 3 standard deviations away from the mean happens 0.3% of the time.  But that's on the good AND bad side.  So to be that far off on the bad side happens 0.15% of the time.  But he was slightly worse than that.  So if the % is as D3 says it is, this should happen roughly 1 in 1000 tries.

tl;dr - OP got screwed.

So in my field, a 3 sigma result is enough to make people question the result. Start gathering more data and see if the significance increases or not.

5 sigma is needed to start making claims.

Every answer in business school is, "It Depends".  Tolerance in process variability is a managerial debate and will differ if you're filling sand bags, filling pills, or distributing virtual comics with no intrinsic value.  Regardless I hate it for anyone who grinds and can't progress as fast as someone who is more lucky.

It's called gambling. 

Sagetopian

DD-The-Mighty said:

Hendross said:

Hadronic said:

Orion said:

To answer the OP's question, LT draws are a binomial distribution with a 15% succees rate (supposedly).

*** Warning - math ahead ***

A binomial distribution has a mean of np, where n = 372 and p = 0.15.  So the mean is 55.8 as the OP stated.

The variance is np(1-p) or 372*0.15*0.85 = 47.43.

The standard deviation is the square root of that, or 6.89.

So 34 5*s out of 372 pulls is (55.8-34)/6.89 = 3.16 standard deviations away from the mean.

To be 3 standard deviations away from the mean happens 0.3% of the time.  But that's on the good AND bad side.  So to be that far off on the bad side happens 0.15% of the time.  But he was slightly worse than that.  So if the % is as D3 says it is, this should happen roughly 1 in 1000 tries.

tl;dr - OP got screwed.

So in my field, a 3 sigma result is enough to make people question the result. Start gathering more data and see if the significance increases or not.

5 sigma is needed to start making claims.

Every answer in business school is, "It Depends".  Tolerance in process variability is a managerial debate and will differ if you're filling sand bags, filling pills, or distributing virtual comics with no intrinsic value.  Regardless I hate it for anyone who grinds and can't progress as fast as someone who is more lucky.

It's called gambling. 

This is really a crux of the whole problem. There is an over reliance in the game on the idea that the act of rolling the dice on covers will entice continued play. They may have found a limit for a lot of people on how much they can take.
For example I have gotten 17 Professor X Master Plan covers and he is still 5/3/3. The last 3 have come from latest pulls. Am I supposed to notify customer service repetively to get the color changed? Wasn't the entire point of champion levels to give incremental increases? Why gate that behind something other than the need for the Iso to do it? 
The true issue is that the game becomes so centered around this in 4 to 5 land that they have to have a special customer service activity to compensate. The popularity of this post and of hoarding shows that there is a disconnect between how the game wants you to play and how a player is compelled to play. The changes needed will be easier once development of features and content is aligned around bringing the two together.

grunth13

Took 40 pulls to get 1 cap america.  I did get 4 jj and 2 robbie, but didn't want the robbie covers.

RTA3000

 So decided to blow my measly hoard of 32 pulls. Got:
1 GRRR
8 BP RED ( already 3 2 5)
6 Shuri blue (already 5 1 4)
5 Nico purple ( already 2 1 5)
3 purple Elektra (Already 5 4 2)
4 lockjaw blue And green ( already 2 5 5)
rest was champ level on other 4*

26 wasted covers...

ZeroKarma

RTA3000 said:

 So decided to blow my measly hoard of 32 pulls. Got:
1 GRRR
8 BP RED ( already 3 2 5)
6 Shuri blue (already 5 1 4)
5 Nico purple ( already 2 1 5)
3 purple Elektra (Already 5 4 2)
4 lockjaw blue And green ( already 2 5 5)
rest was champ level on other 4*

26 wasted covers...

Looks completely random....

BlackWidower

I don't hoard LTs, but I have hoarded over 600 CP, that I spent
And also opened countless LTs, before I finally got a cover today, just after I finished my clears. 

It really did feel like this one was really not going to drop.

My biggest problem (and it doesn't take away from the people that sacrifice to hoard tokens and all the math and percentages and stuff), was that they made Steve the mandatory 5* such a short time after being available.

And there is already a 5* Steve Rogers, which could have just been re-vamped
(Although, I really do love slaughtering CA FA in PVP, so there's that).

CharlieCroker

Just pulled another 20 Latest and went 1/20.  So that's now 35/392 in total for <9%.  How many more do I need to pull for the numbers to correct themselves, 100, 392, 1000 or yet more?

As other posters have suggested, there should be some form of protection in place to stop this happening.  It's horrendous enough for me as an established 5* player, imagine how even more devastating this would be for a 4* player trying to make the jump who has hoarded and/or bought for 6+ months and after nearly 400 pulls still wouldn't have even one 5* champ.

@Brigby

HoundofShadow

I think you need to wait for 6* to appear to get guaranteed 5*. The thing with probability is that each pull is independent of the previous or next pull. The ratio is not going to go down with each pull like:
1:7, 1:6, 1:5... 1:1. But in the long run, probably 10,000 or 1000 pulls, it would even out.

Or you can wait for Cap. America (IW)'s feeder.

The fastest 5* to get a feeder is Spiderman (Peter Parker). He was released on 7 Jul 2017 and got his feeder on 20 Nov 2017. 137 days or 4 months 14 days passed between the release of Spiderman (Peter Parker) and introduction of his feeder. 

2nd: DareDevil
Release date: 24 Aug 2017
Feeder date: 17 Jan 2018
Time elapsed: 147 days or 4 months 25 days

3rd: Star-Lord
Release date: 27 Apr 17
Feeder date: 14 Mar 18
Time elapsed: 322 days or 10 months 16 days

The rockett

HoundofShadow said:

I think you need to wait for 6* to appear. The thing with probability is that each pull is independent of the previous or next pull. The ratio is not going to go down with each pull like:
1:7, 1:6, 1:5... 1:1. But in the long run, probably 10,000 or 1000 pulls, it would even out.

They already said no 6*.  Confirmed from Will. 

Quebbster

The rockett said:

HoundofShadow said:

I think you need to wait for 6* to appear. The thing with probability is that each pull is independent of the previous or next pull. The ratio is not going to go down with each pull like:
1:7, 1:6, 1:5... 1:1. But in the long run, probably 10,000 or 1000 pulls, it would even out.

They already said no 6*.  Confirmed from Will. 

Yeah, we got Supports to chase instead of 6* characters...

heybub

You guys have probably seen this, but someone wrote a simulator for pulls...if we believe the simulator (I have no reason not to, but haven't really looked into it), 315 gives you a 50% chance of getting a 3 latest 5*s.  So, those aren't great odds.

https://forums.d3go.com/discussion/57271/latest-legends-pulling-simulator#latest

This doesn't address your question of how impossible a draw this is, as others have above, but 3sigma is far from impossible.  In particle physics, if we discover a new particle at 3sigma from the null hypothesis, we never claim discovery, only that the data suggests there is a new particle.  We usually wait until 5 or 6 sigma, before we stick our necks out on the line to claim a new particle.  Of course, in those experiments there is quite a bit of systemic errors, and any underestimation of those could have a big impact on the results.

I've seen posts where people claim they pulled like 6 exact cards in a row.  There the probability is far less than 1/1000.  D3 has claimed those experiences are purely statistical, so I'm sure they'd say the same thing here. 

Neuromancer

CharlieCroker said:

Just pulled another 20 Latest and went 1/20.

Was it Ghost Rider?

CharlieCroker

A final update on this thread as I'm done pulling Latest until Okoye enters packs.  Finished 38/412 for 9.22%.  That's got me GR champed and up to level 452, HoboCap champed and JJ still stuck at 11/13.

Postitive - for the 2 weeks of his boost I got to use Cap in PvP and he was fun to use, if not particularly strong.  But it meant I didn't use Gambit at all which is a win in my book.  Didn't have any problem beating anything except Gambolt teams which were still dicy despite my running someone with 100k health.

Negative - well 9.22% says it all really.  As expected no comment from anyone from the devs here and the predictable canned response from CS saying that 15% is not guaranteed and my outcome was one of the many possibilities.  There's also the issue that HoboCap will now see zero use outside PVE (due to Gambit) but that's something I knew when I opened my hoard.

As to the future, now that HoboCap's boost is over I have no desire to return to the Gambit meta so once the season is over I'll be taking yet another step back from PvP.  My VIP expired last week and for the first time ever it has not been renewed.  I don't think I'll even bother spending £2 to keep intercepts rolling as the game is going in a different direction to me and whilst intercepts are great, I have no use for iso or HP going forwards.

The rockett

This is very sad to see but a VERY predictable outcome of years of CS/Devs not addressing one of the biggest concerns in the game.  Now another long time supporter of this game and I know has spent more money than one would like to admit, is going to be stepping away.  I just do not understand what direction this game is going anymore.  

It was great playing with you bud over many many seasons/years.  

optimus2861

RTA3000 said:

 So decided to blow my measly hoard of 32 pulls. Got:
1 GRRR
8 BP RED ( already 3 2 5)
6 Shuri blue (already 5 1 4)
5 Nico purple ( already 2 1 5)
3 purple Elektra (Already 5 4 2)
4 lockjaw blue And green ( already 2 5 5)
rest was champ level on other 4*

26 wasted covers...

That's mind-blowing. Almost as if the server got trapped in some kind of loop and kept spitting out the same cover sequence(s) over & over again each time you pulled, even though it depleted a token each time. I know the odds are next to zero that CS will help you, but get a screen grab of your pulls and send in a ticket. 

Over half the pulls toasted on three specific character + colour combinations. That's gotta be deep in the 1-in-ten-thousands range. 

broll

About a week back I had the reverse luck. On a whim I pulled 20. I got 5 5*s (25%) all of them were Cap and all of them usable to add up to 13. I also got 1 5* BH.