RNG are working properly?

Most don't know in informatic you can have just pseudo random numbers and not true random numbers, now looking at the numbers of posts complaning against systems involved in RNG, are we sure the algorytm is working properly?

For people who want know more about here is explained how work "random" numbers in informatic.

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

 

Please review this official post with responses from Chris & Berek from SOTA as the last 2 entries before it was locked:
https://www.shroudoftheavatar.com/f...dom-number-generator.96824/page-5#post-855113

But long story, short... Yes... :-D

 

In my opinion those posts come from a lack of knowledge in the field of statistics.
Can't blame ppl though. Statistic su**s.

 

I suspect in some/many instances, the RNG is fine, but the percentage that's presented to the player is incorrect, instead of anything being wrong with the calculations themselves.

I remember playing an old game XCOM where I had something like a 133% chance to hit, and still missed.

It's just that people seem to automatically assume that something is up with the backend calculations, instead of the UI presentation. On top of that, humans are terrible at randomness. Failing a 97% mining change 5 times in a row may seem unlikely, but there's hundreds of people mining all over the world day in and day out, over the course of a year. That's a lot of mining instances. The person who "won" that lottery is the one who is going to be on the forums bellowing his/her face off.

 

Just two things here.

1)
There isn't a single computer Backgammon game out there whose forum isn't filled with users accusing the computer player of cheating with the dice. I've seen similar things with monopoly sites as well.
This is human nature and part of our wetware. (Look up "loss aversion" if you wish).

BUT
2)
This isn't a new problem and game designers know this. Hence why most games made for entertainment fudge the dice.

For instance since humans will very much notice a bad streak then one method is to 'save' a roll when the player has a 'too' good streak and re-use that when the player has a 'too' bad streak.
The player won't notice that they didn't get that 10 success streak in a row anyway. While they would have noticed the prevented 10 fails streak in a row.

Some game solves this by adding some power ups to the players where they can earn something which lets them bypass a roll if accumulated unlucky.

Another method is to, after 1h or so, average a user session's important/key rolls and if you are 'too' lucky/unlucky you add a re-roll.

Another is to have a pool for each player with results to pick at random from with bulk refills. Kinda like writing all possible results of the dice on pieces of paper and then drawing random papers out to get the result.

For casino slot machines they have to pay out x money on y bets. (You can find how much in the Info page on the machine - it differs from machine to machine). Thus they have to see that they are not having a too long bad luck streak and need to add some non-random "luck". Smart casinos does this through the jackpot spread on a number of machines thus offsetting individual streaks.

Etc.

But SotA don't fudge the dice.
(Yet...)



Conclusion:
So in my personal opinion this goes into the "if it's not fun then fix it" category. Regardless whether or not the function works as designed or not.

Slow grind actions with long timers and / or high cost with success based on rolls WILL over time result in rage quit inducing 'bad' streaks.

Getting a twelve to fifteen fail streak on a 50+% gathering node simply isn't fun.

Failing ten 50+% masterworks in a row with all the resources which that required simply isn't fun.

So...fix it.
Fudge the dice.

 

I JUST wrote a post on another thread on the matter so I'll just copy/paste it here:

 

No wait, all Garriot games are based on dice rolls, so i'm not complining against RNG system, is like a signature i think, what i mean to say was about 28 items in a row broken to the first or second enchant with 115 enchanting skills and 97 specific If on 30 items i save just 2, ok i can have bad luck, but the negative streak is not a bit longer? And btw considering how much time took gather the mats to do all that it's a bit frustrating. I maked the post just cause in my personal experience i have thing like that, or experience about me Killing a mob 200 times and see nothing drop, another kill the same mob 100 time and drop 8 things. Ok i've no luck at all, but considering the effort at a certain point something can't revert the sort of my RNG dice roll? Cause really if is just based on luck i can stop do anything

 

Now @Wextel , I'm not complaining that you didn't follow the link I posted to the official response that answered your questions & concerns, what I mean to say is here is the information that you didn't seem to get which I wish you would have:
Per @Chris :
****************************
Yes, but the thing is, people like to argue about the closest thing to a perfect random number generator. In reality, the differences are just simply not visible in anything but a statistical proof. You will not see a difference in player perception of them. Also, many of these random numbers are being rolled on the server and there are hundreds of random numbers being generated for other players or events between each of your rolls. In short, if you're arguing over the output of an RNG, you don't really understand how they are used.

Feel free to keep this thread going but don't expect me to take part in it and keep it light hearted because there is as much truly applicable science involved as you'll find in the average Tarot card reading. Don't get me wrong, tons of science in PRNG but none of it is applies to this problem or is responsible for the problems players see in games. It is a long tradition for players to blame PRNG for their problems though so keep at it. Usually these are the same people who will flip a coin five times and get heads every time and then think, "Now I'm due for some tails!" I'm glad those people exist though because without them Vegas wouldn't stay in business!

On the other hand, if you see something you believe to be truly statistically improbable, feel free to report that as a bug and we can scrutinize whether the numbers shown to the players matches the math you can't see. That is always going to be the more likely case and will help us track this stuff down. Just don't report that in a thread titled, "Fix your damn RNG" though because it will likely not get looked at. :/

If you truly can't stand randomness in games, I suggest a nice game of chess but even that starts with a random check to see who goes first. If a single random number truly meant success or failure in a combat, I would agree it sucks but a normal fight is usually involves 50-200 RNGs and no one single one will make or break the combat. If you have a fight that pisses you off because you die because you got a bunch of bad rolls back to back or he got a couple lucky crits at the perfect time, just remember how pleased he must be right now. The RNGs also makes it so you likely have to react and change your strategy mid fight. Also, while you might think it would be fun for the most skilled player to win every time, it would get old fast for both sides to have completely predictable results. Again, I believe in a skillful game BUT I don't believe having RNGs makes it less skillful or less fun.
****************************
Again, linked from: https://www.shroudoftheavatar.com/f...dom-number-generator.96824/page-5#post-855113

Chess perhaps?

 

@Arcadeguy It's that very "official" response that frustrates me, though. As much as I have respect for the guy, @Chris 's response argues the symptoms of the issue, not the cause. I can't blame him because he's getting personal heat from the player-base and it's a very engineer type of response: "You say the math is wrong and I'm telling you it's right. Maybe you should go play chess because it doesn't look like you can deal with how RNG works."

That, to me, says that neither parties understand the problem.

 

@Gix I have to respectfully disagree, Chris's response was very much to the cause while trying to avoid the endless discussions regarding what people are declaring as symptoms... i.e. people declaring RNG is broken based on extremely small sample sizes. The players who have done spreadsheets on this with 10,000+ crafts have determined the RNG is accurate within less than 1% variance. This is another post with a person using a 100 to 200 tries at something to extrapolate that over thousands. This isn't a case of both sides not understanding the concept, it's strictly on the player side.

And while you can summarize comments and adjust to your own personal point of view in your posts, please don't make up fake quotes and attribute them to Chris or anyone else for that matter like you did.

Another thread that addresses the depth of RNG/PRNG with a very good discussion which can help more people understand the complexity is here:

https://www.shroudoftheavatar.com/f...rstanding-probability-and-sample-sizes.70213/

 

Uhm????


I'm sorry but that comes through as totally missing the point Gix was making.

The issue isn't the PRNG. It is how the user experiences the features that are using the PRNG.

Now since the wetware we humans have is kinda faulty it optimizes on different things than is logically sound. Then a game that relies on random rolls only, without fudging, for critical stuff will eventually lead to a bad experience. Hence my post above and Gix's point above.

It doesn't matter if it evens out over thousands of rolls. The only thing that matters is that over the course of those thousands of rolls it is deterministic to predict that all players will eventually have a very long fail streak.
That deterministic very long fail streak is one of those rage quitting moments which you really really want to be careful of as a dev.

Now in combat it doesn't matter as much as you roll so many times and each time you roll isn't that critical - so even if you have one of those very long fail streaks you wouldn't normally notice it.

However in crafting the chance is right there on the screen and the outcome is very much black or white with high stakes on the line. With such a mechanic you WILL give each and every one of the dedicated crafters plenty of such very long fail streaks affecting their perception of the game play.
Thus in the forums we regularly hear players who love crafting in other games giving it up in this one because it is too painful when you hit such very long fail streaks. (They normally post it less politely and not necessarily pointing to the correct root cause, but still).


Thus even though I love the guy, but Chris' reply is wholly inadequate as a response to such venting/ranting from frustrated players. Which is what Gix point out by paraphrasing what he said. Paraphrasing is not the same thing as fake quotes.

Let me show you by paraphrasing your post: "to enjoy the crafting in the game you must have a spreadsheet". I'm not saying that you said that, I'm simply boiling down the message to whatever rhetorical point I want to make. In this case that we can't expect every crafter to have a spreadsheet on the side so that they don't quit the game just because the human wetware isn't designed to cope with too much bad luck.

So your "it's strictly on the player side" is only valid if we don't mind losing a number of players just because we refuse to do what most games do when faced with similar problems.

 

Thanks @Spoon , I couldn't have articulated that response better if I tried.

 

I understand the feelings about this but the fact is that it was intentionally designed this way. The point is that you're risking what you crafted for a more powerful version. Remember all the "risk vs reward" arguments? Circumventing that risk in any way cheapens the reward.

 

Fudging doesn't reduce or change the risk if done properly. Fudging changes the distribution of streaks by manipulation based on knowing the history of a given individual. Thus changing the experience, not the outcome over time.

 

Manipulation like that serves to move the results of a three-state probability toward the center. Since, in masterworking and enchanting, there is no critical success (there is only success or failure) then any manipulation to reduce streaks of failure would result in more success.

 

I'd just like to add that a third masterwork is not nearly as valuable in desirability as this system makes it end up costing.

 

Then intentionally fix it. The reason why there are discussions on the issue has no bearing on whenever or not the problem was intentional.

"Sonic Boom: Rise of Lyric" is an atrocious game regardless of whenever or not it was "designed" to be that way.

The risks of losing a soldier in X-Com is still great even without having the player losing a 97% hit roll turn after turn.

Besides, the best "risk vs rewards" mechanics are the ones that aren't associated with RNG to begin with.

Manipulation of the RNG output is one thing, designing your game so that the failures aren't insulting the player's trust and time invested is another thing altogether.

If master-working and enchanting has a binary output, another alternative would be to change it.

 

I don't get the reasoning behind this objection.

I said "if designed properly". Why do you take the assumption that the devs would fail to implement this properly as a given outcome?
Above I gave five examples of how other games does things out of which three of them does not result in an increased chance of success. Why do you think that this game would fail to do such where other games have succeeded?

That assumption is demonstrably false since it is only true if one presuppose that there isn't a similar manipulation of streaks of success. Where in three of the examples I gave above from other games does exactly that. (Please note though for these to work properly you don't inform the player that that is what you are doing.)



Edit:
To make a clearer statement, it might be prudent to give the generic idea of it.

If you have chance X and you designed a working RNG then after thousands and thousands of rolls the statistical outcome of that is X over the whole population of players. Without fudging individual players will of course be distributed with <X and >X, where some would end up in the extremes of the bell curve of distribution.
In this design you are going to have unlucky players that when hitting those <X will quit the game long before their own individual statistics would even out.

However with fudging your goal is to have every individual player end up closer to X, while still getting the sum of X for the whole of the population. This is done through manipulation of streaks, or by trimming the bell curve if you will.
In this design there are less lucky/unlucky streaks making the effect of the RNG less impactful over small samples but remain the same over large samples. Thus increasing player retention.

Edit 2
If we are talking about role playing games, then this could be something as simple as the pseudo equivalent of making a d20 system versus making a 3d6 system. They both have an average of 10.5 but their distribution and thus the perception of their rolls is completely different.
Where the d20 system would be more appropriate for things where you roll often and want to have wide variety and that the unexpected is part of the fun - like combat.
While the 3d6 system would be more appropriate for things where you roll seldom and where you want less variety - like crafting.

 

This would require keeping track of success and failure for each avatar. That's fine for a single player or limited multiplayer game but wouldn't scale well for an MMO.

 

Again I don't follow. How could something which other MMOs does not scale well for this MMO?

It is no different than having quest flags, or number of kill lists, or achievements, or any such stat.

Edit:
Since it seems that most of this comes from misconceptions, so I edited the post above and added examples to make it clearer what fudging is and does. I also added a seperate way of looking at random roll distribution through a system design example from pen and paper RPGs.