Understanding Probability and Sample Sizes

To be honest we have to even say we don't know who is actually on these forums if anyone at all (except for ourselves and anybody who we speak to in person who indicates they read the forums by discussion it with us).

That said these forums could be generated by IMB's Watson. and all the people who know wjat has been said we don't know if they wrote anything. they might have read it but ???

Statistics polling pretty much anything that we observe like this might just be a show or it might be very rral but this conversation might mean nothing if Im right.

 

Many folks fail 3 times out of 20 attempts at 95% and call foul. 95% of 20 is 19 so getting only 17 stinks. However, the odds of that happening are around 6%, enough to be uncommon but nothing suggesting that something is wrong.

Personally, I only watch for 1 in 10,000 events. Failing consecutively 4 mining attempts at 90% or 3 skinning attempts at 96%, I mean I could track all my interactions for statistical anomalies, but Port isn't paying me enough for that.

The problem is 3 fold:

1. Luck is always against the player. A player I'd far more likely to fail an important 83% check than to pass a 17% check ... because very few players would ever take a risk on a 17% check, that'd be throwing away time and effort.
2. The odds calculation is hidden. Never mind the possible hidden values, how much does skill count, what is the difficulty of a given task? Improving transparency would give player control, or at least the illusion thereof, over the process.
3. Crafting's primary design goal is to be a gold sink. Pretty it up all you like, if the system's intent is to eat player time and many, don't be horribly surprised if any perceived flaw is magnified.

Fix some of those and the complaints will shrink to a much more manageable level.

 

It was strange, over a period of enchanting i was only getting certain choices, cant remember what they were now, but they were all similar, not what i wanted at the time. But it made me think at the time, do various enhancements have more chance of appearing at certain times/stars in sky/moon/weather etc. I haven't done much since but i was going to take note of what enhancements i was given the option of at what times. I wouldn't put it past them to have something like this hidden in game for us to find out, they sneaky like that, and i like it...

 

1) Implies that Portalarium would intentionally lie to its playerbase out of sheer malevolence.
2) How much more transparency can you have than a % chance? Flip a coin, you have a 50% chance of heads, 50% chance of tails. How much more transparent can it get? Again, the only opacity here is if Portalarium would intentionally lie to players and hide variables, which is frankly unlikely given their behaviors so far.

 

The biggest bias is that we remember failures a lot more than good streaks. The other day I succeeded a 23% roll 6 times in a row (0,00015% chance of happening, so roughly 3/20000). Yet how many people come on the forums to complain about rigged stats then?

Imagine the reverse, failing a 77% chance of success roll 6 times in a row. I bet you'd hear a lot more complaining about rigged stats then. Yet, the chance of either happening is the same. That's simply how randomness works. Sometimes you win, sometimes you lose.

 

Certainly possible and there are numerous tests that Portalarium can do in their server logs, I posted NIST tests in the prior thread. The simplest being turning the RNG into a bitmap and seeing if it looks random lacking patterns.

The point is that players themselves cannot do it, all anyone anecdote's prove is the RNG is working as intended because hot good luck and cold bad luck is a feature of randomness, like it or not. Vegas certainly likes it because the profit of that fact that many do not understand randomness.

 

So this is where we get into "seeing all the data". We don't like random events, so even if we saw all the data there would very likely be people saying the data was "rigged". (word used purposely with contemporary events in mind)

At the end of the day, Portalrium has all the control and we really just have to trust that they're using their power for good and not (for example) buying up COTO's to maintain or increase the demand.

I personally believe that Portalarium is good people and if something is broken they'll fix it.

 

See my above post.

"The biggest bias is that we remember failures a lot more than good streaks. The other day I succeeded a 23% roll 6 times in a row (0,00015% chance of happening, so roughly 3/20000). Yet how many people come on the forums to complain about rigged stats then?

Imagine the reverse, failing a 77% chance of success roll 6 times in a row. I bet you'd hear a lot more complaining about rigged stats then. Yet, the chance of either happening is the same. That's simply how randomness works. Sometimes you win, sometimes you lose."


TO BE FAIR, it's not impossible that the Portalarium math is wrong. The point is that you would need MASSIVE data for that to be determined, and only Portalarium has that. So it's pretty much a moot argument at this point IMO.

 

The seed had some issues = the problem.

They need to generate a new one and run a couple 'flattening' passes if there's huge clumps of highs/lows

Currently I can game those clumps because I recognize it's flaws

Others have seen the exact same patterns, so it's very real.

Would be nice to have them recheck, reseed and verify the %s for exceptional chance at L100, prosperity tools and founder tables. Because that is completely wrong too

 

The only way to do that is to roll, then tell if you will succeed or fail before you craft. What use is a crafting system that tells you WILL fail, of course you will cancel and reroll. Then the system becomes biased to only successes, it become predictable. Crafting no longer functions as a gold sink. Multi account players working the game own the crafting economy.

With the current system the multiaccount players can fail hitting the bad luck streak, while the casual player can succeed hitting the good luck streak.

Just like the housing lottery. which most people have been saying is a fair why to give access to deeds. Without randomness no casual player could ever acquire a deed, either by winning the lottery, or buying the deed from someone who won what they did not want.

 

Oh exactly. Something is amiss. Either it's the RNG, the numbers it's rolling against, the calculations to determine those numbers, the UI reporting back the numbers or, quite simply, the design of the system. Or, it could be working 100% as designed. In that case it still needs something done to make it less frustrating and depressing. You can still have failures, they just need to kick us in the gut a bit less.

 

That certainly is a reasonable crafting design that uses RNG, makes rares expensive and commons cheap by degrading quality on failures rather than breaking. It still has the same amount of grind and luck to achieve a rare.

The problem with this idea is they would have to flag failed items from renchantment otherwise you can avoid the grind and just rework the same thing over and over.

 

Not sure if this has been stated, but you can get valid statistical conclusions with reduced samples if you have an initial hypothesis using Statistical Hypothesis Testing.

https://en.wikipedia.org/wiki/Statistical_hypothesis_testing

I agree with some comments i've seen skimming the thread; whether the math is right or wrong, the current crafting path is just grindy and repetitive, which regarded as boring or unattractive to some. Too much relying on randomness make the player feels he's not in control, and that personally makes the game less fun to me. Same thing happens having high tier fights... I depend on critical strikes to kill the mobs because standard attacks don't do enough damage to get em down before they kill me.

 

No malice implied, it's not even a statistical problem really. Set up booths on opposite sides of a city, hand people 5 bucks and give them the chance to risk that 5 bucks for a chance to win some unspecified prize worth more than 5 bucks on a die roll. The catch is that players at one booth win only on a 6 and the other booth's players only lose on a 1. You'll find that more people lose at the booth with good odds than win at the booth with bad odds, not because the game is rigged, but because folks walk away with their 5 bucks.

Play a game like Xcom, you get to see that your base to hit is 60%, plus 10% for your scope, plus 30% for flanking, minus 2% for the target's defense for a total to hit of 98%, which you will still miss. That's why in the sequel there was such an uproar when they introduced a dodge mechanic that wasn't factored into the to hit and which couldn't be managed by the player.

 

Yes, flag/tag the item as "Finished" and make sure that the mastercraft/enchanting/whatever systems allow no other modifications. Maybe even go to the extreme and only allow repair kits to be used on it; disallow COTOs to repair. That would guarantee that the player would have to buy another weapon eventually, but depending on the price of COTOs at the time, perhaps that would be a limiting factor on it's own, not to mention people naturally wanting to get better things anyway.

This would go a long way toward making the effects of failed crafting be less depressing, and would reduce people's snap conclusion that the RNG is broken.

Anyway, Browncoat Jayson started a thread for this kind of discussion here: https://www.shroudoftheavatar.com/f.../mid-tier-masterworking-and-enchantment.70355

 

A test for that is Fourier Analysis, which shows the frequency of the highs and lows. This is one of the NIST tests that Portalarium could run. However you are making assumptions about random numbers that are not valid. Randomness means that the signal has a flat spectral response, meaning both of these sequences must exist in equal amounts. Flattening the distributions out is the same as running a high pass filter on the results, it removes the hot and cold streaks.

1010101010
1111100000

The game then becomes predictable in the short term rather than the long term, which means the retired multiacount player can work the odds because the sample sizes required to achieve probabilities with confidence become much lower. That is not good for the economy if crafting is predictable.

Certainly if the same seed is being used and sequences are the same for the same player then it can be gamed. That could only happen if the RNG was a client number rather than a server number, surely they are not that stupid at security?!

 

Why not just remove the % and say something like this seems like an easy work, or this seems like a moderate to easy work, this seems like a moderately difficult work, this will truly be a test of your skill, and ultimately this is about as hard as successfully navigating an asteroid field (but this is fantasy so it isn't impossible, may the force be with you.


******
However I can see why thry use the numbers because it shows progression. if you're a 3rd level foghter and you go up t level 4 and you don't see any change in your character (like additional hitpoints) then gaining that level is just for show and tell. it is meaningless.

Ultimately this is all meaningful of cource but perhaps there is a way to strike the ballamce since some people are going to wonder why they gained skill and their % hasn't changed much or whatever, and if they don't see percentages there needs to be something to see that means something.


Although I suppose they can just keep it as it is and do what they plan on doing since that is probably well thought out.

 

You are correct, and when you have a system with a less confident prior-state, statistical variance becomes more of a flag. A highly confident system (e.g. a coin flip) that shows improbable pattern (10 heads in a row) is less worthy of scrutiny than one which is less confident (e.g. a computer game in development). That was your earlier point that people didn't understand statistics and that 400 was not meaningful. But I suggest that 400, if controlled for, is, and your comments on statistical relevance were misleading (accurate to what the video states by not to the point you were making). It's a strawman to argue the 3 10% failures in a row is all people are arguing about. Hence, why I posted what I did. And yes, I think your approach was incredibly condescending. You used a straw man to represent an entire group of people raising issues, and then talked down to them in a manner which implied you were more knowledgeable than them and needed to educate them on why they were wrong.

 

It's in the Unity Engine. Unsure how/where. Could have a hashed local copy to fix latency issues

When it takes a day to farm and process , and a couple seconds to roll/destroy, flattening would improve the user experience.

Sample size is always going to be extremely low for rare mat crafting. Therein lies one glaring issue

I could care or less about gathering strikes a fizzle stretch.

My rate is under 20% when making 20 wands, but if I use streaks and patterns and do components in between, it's more like 40%. Perhaps I just have limited precognition

 

It's not a strawman to say they were arguing about it - it's a verifiable fact. I don't think I said or implied that's ALL they were arguing about. I think I did a good job showing that what the actual problem was is that they don't like the RNG system if it works or if it doesn't work as designed. Essentially "most" of the arguments come back to that. Wouldn't you agree?

Sorry you feel that way. Perhaps you would feel differently if we were having a face to face conversation and tone was more accessible. I'm intending to be direct. In some cases I am no doubt more educated than some of the people posting here. In other cases I'm not. That was far from the point of the OP however.

But what's more important to me for our conversation is this. Are you saying that as a learned person you have doubts that the RNG is working as designed? If so, what do you point to that brings you to that conclusion? If not, why are you focused on that part of the conversation?