Understanding Probability and Sample Sizes

no no no, no. Wait a second here. The OP never addressed the issue you're bringing up. If you want to talk about your issue, make your own damn thread.

Plain and simple, this is an informational thread trying to help people understand RNG and probability basics.

 

LOL, I could spend all day reading this stuff! I fucking love you guys, try not to listen to the morons. Thanks for all the input and suggestions, talk to you soon!

 

Yeah well if you want to understand RNG, stop being a crybaby and understand you might go on a cold or hot streak.

You need at least 100,000 examples before anything makes sense statistically... so stop whining and keep on trying!

Or if you want a more dumb downed explanation - your going to break your ****. Go home before it breaks. Maybe you'll get lucky, but your stuff will prob break haha...

 

Actually this thread came across as Drocis being condescending once again. I think the next scheduled terrible thread is one that explains to us peasants what Early Access and Alpha is.

 

That doesn't seem to be the case here. In order to be able to address the issue productively one needs to fully understand what exactly is the issue. While the concept and math of success chance maybe accurate for the large sample, it is not useful to the crafting where people only intend to craft a few quantity at a time.

Although if he's gonna make a post about early access, might as well make some infographic with shiny charts and graphs.

 

Probability is a lot more complex than what he is making it out to be though. He never even addresses things like pseudo-random number generators, or frequency stability. To create a thread which takes the high and mighty approach of what is essentially "stop posting threads about crafting fail streaks - you know nothing" with an introductory video on how probability works is condescending.

And that is before we even get into the debate of whether a truly random system is good or not for a player's game experience.

 

Was that really the question?

I thought we were only asking if the percentage was as advertised in the interface when you craft the item. If it is not exactly accurate I would prefer to know the real statistics before I make my decision, if devs dont want to tell the numbers than leave it blank, but dont have an innacurate percentage.

 

I was just giving an opinion.

I agree the percentage should be accurate. However, it's clear that there are many people confused by what the percentages mean in terms of probability. IF you have an 80% chance of success, that does not mean that every 10 consecutive times you'll only fail twice. I don't believe it's the responsibility of the developers to explain that, I believe it's the responsibility of the players to learn how probability works, and in this case that should only require watching the video in the OP which lasts for a total of 8 minutes.

 

As I said before, the RNG could be broken. Most RNG's are not truly "random" and that's a discussion that we could have in a bubble to our heart's content. Would it make people feel better? Probably, because they don't like the system of random in the first place.

If I sound condescending, maybe it's because we have a number of people that defending their ignorance of statistics REPEATEDLY as if it's "ok" to be completely wrong about facts. Again, you're all welcome to your own opinions, but you're not welcome to your own facts.

At some point we have to agree on what the facts are, in this case it's how probability works, and then we can debate the if the system is fun or not. But we can't reasonably make up our own facts just because we don't like something.

In my case, I actually have few problems with the RNG (assuming it's working correctly, and I have ZERO reason to believe it's not). But I understand that other people may never get over the fact that success is purely random. If they simply made threads saying how they didn't like the randomness, that would be more reasonable, in my opinion, rather than flooding the forums with misinformation and meaningless numbers.

If it's condescending to tell people what the facts are, I'm guilty as charged.

 

If it was my game, honestly I would probably put some mechanics in that essentially gave players a free pass once in a while.

So basically, it would be like, every 5 attempts, you would get an automatic 100% chance within the next 5 attempts afterwards. Not necessarily 5, but some amount that I wouldn't disclose.

It is a game after all, and games are supposed to be fun.

What really grinds my gears here though is that the crafting system, or shroud in general, is candy as hell already. Look at UO, UO:F is a freeshard that spawned almost 4 years ago, and they have in the tens of thousands of uniques logging in annually, more than most if not all of the broadsword servers.

There, if you're trying to make gear, you lose all the ingots on a fail, but not only that, you can get PK'd while you're mining before you even get that far. Hell, you might even get stolen from or PK'd, drylooted, pack horse stolen, all while standing in the crafter area inside town while surrounded by other players who practically can't help you avoid it.

So really, if you're only losing your gear here when it comes to enchanting or mastercrafting, that's the price you pay. Things could be way worse. They almost need to be to keep the game alive for 19 years like UO.

 

Well, 400 is far from handful. The point is you cannot be selective about your samples, just start measure the probability and do 100 of runs to see the approximatte probability. For probabilities at edges (below 5% and above 95%) you might need over 100 of runs to see if they are really broken, but you should get pretty close very soon with 95% probability or so. If you think that probability is broken after that then you need to run test on larger sample. My point is you do not need extremely large sample if the preobability seems to be reasonably close at small sample.

I just say that truly random system can be extremly punishing in some cases and not fun at all. That is why I think that each failure should increase your chance of succeeding next time a little bit, while success should reset the probability again. Otherwise people end up getting those perfectly expectable failure chains.

 

It shouldn't matter. If Eve Online taught us anything, it's allowing players to geek out and turn a game into a spreadsheet simulator does nothing except encourage endless arguments that can never be won. The thread will devolve into name calling, like most threads on this subject do, and will get locked.

 

I would be interested in to know how true that is. I don't think it's true.

I think the sample size will only be meaningful if it's relative to the total population of outcomes. In this case, having an outcome from 1 to 100 seems like it would need a very large sample size. Afterall, if you watched the video you needed perhaps 1 million (he actually said "infinite") number of coin flips to achieve 50%. That's just with 2 possible outcomes! (! put for emphasis, I'm not yelling)

 

Yeah he wound need infinite flips to reach 50%, but he would get very close quite quickly with 99.99% probability or so, I really wish I recalled more from my statistic classes, but my point is if the probability is 50%, you willl get quite quickly in the range 45%-55% and it will go closer and closer to 50% with more throws. It will get more and more unlikely to get chain of heads or tails longer than number of throws you had so far...

I think it is about confidence interval (I hope I have the term right). The point is you get range of value (say 48 % - 52 %) and you are 95% or 99% or 99.99% confident that the probability is within that interval. You then play with number of observations (higher number either increasese confidence of interval or makes the interval smaller...) or you make larger interval to increase yout confidence...

The thing is to say the RNG is broken you really should observe 20% probability while the UI says you should expect 60% or something like that. It should not matter whether you are experienceing 48% instead of 51% imho, those are hard to prove.

 

Yes, I see what you're saying.

I'm just not sure that 400 gets us to that confidence interval (I think you have the term right, but I'm also relying on memory from classes I took 10+ years ago). I thought about trying to find my statistics books and figuring it out, but then realized I'd much rather grind away levels.

 

Well, I am lazy to do the math, luckily there are tools on the web already!

https://www.mccallum-layton.co.uk/t...nfidence-interval-for-proportions-calculator/

So you can play with the sample size, confidence level and observed percentage, it will show you how big or small expected probability interval is. And yes for high level of confidence you need large sample size.

 

The age of science has passed us unfortunately. I assume there are more people who think the world is flat today than there were 50 years ago. Decisions, predictions, explanations these days are based on emotion instead of logic/science to a greater extent than at any time I can remember in my lifetime. What the future holds is anyone's guess. That being said, statistics are and have been a very difficult subject for people to understand. The loss of time, effort and mats at the roll of the dice is such an emotional event for some, that the truth of statistical chance experience is thrown out the window in hope of some emotional retribution. You see, it is just not fair.

 

You pretty much have it right, and 400 gets you pretty close in this scenario. The asymptotic curve actually levels out very quickly, it's one reason statistical sampling polls are so accurate based on seemingly small numbers (e.g. 1300 people representing 50,000,000) giving only a +/- of 3% with 95% confidence.

 

Cool video, it show's, very well, how a "game of Crap's" works, and how the "House", always has the advantage because it's already been proven that coin flipping, in reality, isn't a 50% chance... At best, it's 60/40, although you DO have a 50/50 chance of deciding which it will be... Chris S. just happens to be a "Gambler"... Funny how that works...

 

So I have a question...

If we are playing a black jack game on the computer how do we know if the same odds and probabilities exist for a real game of blackjack, and for the one we are playing on the computer?

A person can count cards and determine the likelihood of getting the cards we need. however for all we know the webbased card game we play on the computer might just choose cards at random as if the decks were shuffled before each and every time the players are dealt new cards making counting cards irrelevant.

Same could be said about voting, nobody watches people count, and computerized voting machines may not count anything . .... or there could be a Micah from Heroes season one who manipulates the data resulting in a landslide for Patrelli, and then there is Zenga!!! oh don't get me started with Zenga poker!