# Statistics-related questions

I recently asked this question (10K only. Screenshot for others.) which has gotten a lot of negative feedback, mostly due to it's statistical nature. I have deleted it as the downvotes seem to answer my question, so I'm not sure if the link will break.

However, the fact that the question is crafted out of a limited sample should be irrelevant. If a user comes here to ask a question based on their experiences, then it should not matter that the probability of it happening to them would be.

For example, if a user always got a critical hit on every 5th attack then they might assume that's how the game works. If they ask a question based off of that, it should not be considered a bad question because of an assumption from statistically improbable observations.

The user notices something from their limited sample and draws a hypothesis. They then ask on that hypothesis. The answer can easily be that they were extremely (un)lucky, but the statistics of it happening should not factor into how good/bad a question is. While most of these types of questions can be solved through a quick search, some may not be as obvious.

So, what makes a statistical question like this a good question (given it has a +8 as of now means the community thinks it is good)? The only difference is that my question had personal observation and the other one does not include it.

• I'm not sure why you deleted the question, honestly. I could get an okay answer to your question relatively quickly
Commented Jul 20, 2013 at 19:01
• @badp Partially for the `Peer Pressure` badge and partially from actual Peer Pressure. Commented Jul 20, 2013 at 19:02
• I'm not sure why people decided to answer the question in the comments area honestly.
Commented Jul 20, 2013 at 19:04
• @badp I think when a user wants to answer the equivalent to "No" they tend to be hesitant without sources to back them up. Commented Jul 20, 2013 at 19:06
• Personally, I was explaining in the comments why the premise of the question was incorrect, and I was writing up an answer with greater detail and math. And I didn't downvote Commented Jul 20, 2013 at 19:10
• That's fine, I appreciate the comments. I just usually place comments and downvotes in the same line of thinking such that if there are downvotes and commented, the comments likely (try to) explain reasoning behind downvotes. Commented Jul 20, 2013 at 19:20
• Could you sum up your question for non 10k users? Commented Jul 20, 2013 at 19:40
• I downvoted the question not because it was about statistics, but because it operated from a deeply flawed premise. If you're going to ask a question about statistics without showing even the most basic understanding thereof in your question, that smacks of poor research effort. Commented Jul 20, 2013 at 19:41
• @CruelCow Here you go. Commented Jul 20, 2013 at 19:42
• The problem is that these are not statistically improbable events. It is actually quite probable that a random distribution will produce similar results. For example, here are twenty random results, as generated by random.org/integers : 5 8 8 2 5 9 8 3 9 4 4 3 2 10 6 1 10 4 10 1 -- note that there are no 7s in those results. In the next twenty results, there's a 7 but no 3s or 10s (five 9s though). Random numbers have streaks. With only twenty picks, it would be statistically improbable for you to get at least one of every card (assuming there are ten possible cards). Commented Jul 20, 2013 at 21:18
• @Brythan Again, it is not about getting one of every card but multiples. In my case 25% (5/20) was the save card. Then 20% was Skyrim and 20% was Prison architect. This means that 65% of 20 cards were the same 3 (given 10 total). Furthermore, this is irrelevant to the question. Based on the observation, I asked how the system worked. If it truly is pure statistics, then that is the answer. Commented Jul 20, 2013 at 21:29
• And note that in my second run, I also got 5/20 of one number (9). I didn't mention it, but I also got three of three other numbers (I think 1, 4, and 8). Interestingly, I also got two streaks of three (1 and 8?). Yes, that's just how randomness works. If you hadn't deleted the question, someone might have gotten a reversal badge from it. Commented Jul 20, 2013 at 21:47
• @Brythan Which is fine, but given that I had 5/20 of one card, it raised the question of "Am I more likely to get this card over others" which is a valid question given the data. This gets reinforced when I noticed my friend had 4 Dead Island Riptides out of around 15 cards (~27%). The question remains valid. All I'm getting now are comments in the form of answers to the question. Commented Jul 20, 2013 at 22:54

The entire premise of your question is flawed. You're hypothesizing that the drop rate varies on a per user basis, and there's no way to reasonably test that.

For anything approaching accuracy, you'd need anywhere from 100 to 1000 or more cards, per user. You might be able to get one or two users that get that many cards, but in order to answer this, they'd have to report their results.

In order to provide an authoritative answer to this, you'd need enough baseline values to actually determine if there's enough deviation from expected values to answer this. Without that, this can't be answered. Even if this exists, what value does it provide, beyond pure curiosity? How does knowing you have a better chance of getting a KSP card help your enjoyment of playing anything?

In the end, it's definitely game related, but I don't see how it's A) reasonably answerable, or B) provides any value, even if answered.

• In response to B, If you are more likely to get certain cards, then you know you will more then likely need to trade/buy cards to finish the badge. This is mainly about the current Summer Sale, but it could potentially be for any game that gives cards based on purchases. Commented Jul 20, 2013 at 21:10
• In response to A, you can see any user's cards by using the following format: steamcommunity.com/id/[ID#]/badges. This means you theoretically have access to see every user's cards, giving you a more then sufficient sample size. Commented Jul 20, 2013 at 21:22
• I think there's probably at least a privacy concern in trolling through private profiles (if that's even possible), and you're not refuting B at all. The only way you can know is after the fact. After your personal sample size is large enough, you know which cards to get because you don't have them. Commented Jul 20, 2013 at 21:43
• @fbueckert - I'd imagine anyone who set their profile to 'public' would be fair game however, at least for non commercial interests. I do still believe its a speculative question, as only the devs could answer it authoritively
– Robotnik Mod
Commented Jul 21, 2013 at 15:41
• @Robotnik Well, with a big enough sample size, we at least have enough data to make some rather accurate guesses, but I still question the value of figuring this out, authoritive or not. Commented Jul 21, 2013 at 16:41

As hinted at in the comments here and on the question itself, there are two problems with your question:

1. It's unanswerable by anyone except a handful of folks at Valve.
2. If we make unwarranted assumptions (due to #1) about the probabilities of receiving cards and answer based on them, then it's purely a statistics question. It has nothing to do with gaming or Steam.

That's just not the kind of question this site is for.

• @badp Steam Market numbers are not all generated cards. To do a proper analysis, a partial snapshot doesn't give an accurate picture. Commented Jul 20, 2013 at 20:36
• @fbueckert That's incorrect as I try to explain in my answer, but that's not relevant. His question is about whether or not users get put into bins with different card dropping probabilities. My question is about whether or not cards drop equally as likely for everybody. They're different questions and I'm afraid Matthew is correct about the lack of data. I don't know if that's a good enough reason to drive a question off.