RNG are working properly?

So, you're saying that keeping track of every random number generated for every avatar is the same as keeping track of quest flags or stats?

 

No. But that is not how you do it though so that isn't really relevant.
Depending on the example above you'd either store some sums, some streaks or just whether if you are hot/cold.

Again, if other games have done this (for decades) and there are several that are doing it right now, including MMOs, then why do you suppose that it is not feasable? I don't get it.


I'm feeling that I'm unecessarily stepping on some sensitive toes here, since there must be some emotional stake I'm missing which makes you argue against something which already exist in practical reality?
So I feel like apologizing, but I don't know for what.


Please feel free to ask for details or explanations on any of the things I've outlined above and I would happily do that if it helps.

 

I believe what @Spoon is saying is that you wouldn't keep the individual random numbers, you'd keep the fail/success rates.

In another thread (boy, parallel threads are a joy to deal with... /not), I talk about:

... so all you'd need to do is track a single value per player. So you're allowed to fail, but you're also allowed a mulligan if you fail too often. This could be limited to "slow" rolls that don't get repeated in rapid succession... like crafting or looting, as opposed to combat.

Which is probably what he's suggesting (or something similar).

 

Your not stepping on any toes. Ultimately, if something could be implemented that doesn't impact performance, removes the same number of items over time, improves the perception of the system and can't be exploited then I'd be for it.

 

It should be unexploitable as the rolls are done on server side and not client side. If it was client side rolling (which is the bane of any well designed MMO) then its exploitable as all get out (ArcheAge had this issue when I first got into it and it really screwed those who were honest vs those who found how to break the client).

 

So, once I figure out the mulligan and I succeed a curtain number of times then I should switch to a "practice" product because it's very likely that I'll fail soon, right?

 

Sort of. Depending on how it was structured if you are succeeding too often then it may give a negative mulligan as well. Kinda like counting cards in black jack (not that you are lucky enough to get a free winning hand, but that the numbers could float positive or negative based on streaks). It just depends on the tolerance level that was programmed in for how likely you are to get it.

*Edit* However, if I were designing it I would assume that the level of tolerance would be tied to the skill and potentially (if you are code heavy which we are not) the type of item/enchantment that is being done. This could be done specifically to try to stop players from gaming the system in that way.

 

IF designed correctly, then check check and check. However it does introduce some amount of complexity and thus a vector for bugs.

It would have a miniscule performance effect, but since we only want it implemented for critical rolls then it is mostly just crafting - and in crafting the actions are far apart anyway due to the action wait timer.
It would be a mistake to implement it for non-critical rolls which repeat all the time - like combat.

It would remove the same number of items, it would produce the same number of high end goods and the sum of the success/fail rate would be identical in the total population.

It would definately improve the perception and the user experience.

It could only be exploited if you had a dev contact on the inside, at which point it wouldn't matter anyway.

That would be an inproper implementation.

Instead what you do is having different tiers.
So that you cannot affect the tier above/below by rolling in a different tier.

Edit
In @Moiseyev Trueden 's card counting example. That only works if you can raise or lower the bet or switch table.
Here we control the casino server side, so you would always end up at the same table if the bet is of a certain value.

 

Anytime you manipulate random results, you're providing an opportunity for people to game the system. You can define the system to account for that but then you're making it unfair for those who just want to play and not count cards.

 

Again, that isn't how it works.

Lets boil it down to basics and say that we roll 1d6. Rolls of 1 and 6 are always miss/success.

In system A we simply roll and the result is given by the roll. This will over time result in long streaks of 1s and long streaks of 6s. Where too long streaks of 1s will make some players quit the game.

To remedy that, in system B we introduce two new rules, if you just rolled a 1 and then roll another 1 we reroll and that reroll sticks even if it was another 1, then the same for rolls of 6. As simple as that we have reduced the streaks in both ends giving the same end result, but we have reduced the number of streaks and the length of the streaks by a factor of 2.

The only way to manipulate that system is to stop on a roll of 6. In our game that would mean deleting your character and start over at the top of a hot streak. That would never be worth it even for a botter with full knowledge of the rolls.
....and that most importantly presupposes that you know that you have just rolled a 6 - where in our game that is hidden information server side.

Edit:
And rolls of 2-5 are unadjusted. Those still result in fail/success but since they are in the middle they are not as interesting in reducing the bell curve.

Also - that is just an example of a simple system - a real system would of course be different, like the tiers I mentioned above.

 

As I said it is only kind of like that. Its still the simple breakdown of the number goes up or down depending on streak. Like @Spoon pointed out, it wouldn't benefit you because you can't change your bet (in this case trying the throw away "practice" item would be attempting to change the bet) and you can't walk away from the table (that explains so much about crafting... stuck at the table all day long). All tables (item tier or enchanting value or whatnot) would have their own table with it's card count that is not tied to the others. But that's stretching the analogy a bit to the breaking point (as the intention was not to show how it could be gamed but merely how the idea of the floating +- value is applied to other systems).

Also, this wouldn't be how it would be implemented in actual practice.

 

So I should work on cheap "practice" items until I fail then switch to my good item because, at that moment, I'll have a much higher chance of success.

 

And this is the vital bit to stop gaming the system is that players are not only ignorant of the roll, but that it must be server side (client side can be hacked to reveal or edit these results).

 

Different tier items shouldn't affect the design of the count. So the practice item would still only benefit or negatively impact that same tier item.

i.e. table leg should only affect other similar wood based crafting tier items and never benefit the ring of +20 nerf busting.

 

I didn't go through the posts on this page, but I have a simple little design for tracking the states.

Every time you fail, you decrement an int counter. When the counter hits the streak limit, lets say 5 in a row is considered unlucky, you get a success.

When you succeed the counter is set back to 0, a middle state.

Every time you succeed, you increment the counter. If you get 2 in a row you are considered really lucky and you are reset to 0. Or maybe some value higher or lower.

The point is the PRNG value isn't being manipulated so you don't have to track it at all. Instead a simple counter with guaranteed success and failure values is used to track the streak, much like a card counter would do in Blackjack to figure out what states he's in.

 

Nope.

 

The problem with that approach is that you'd have to basically have to store a value for each % chance and combined with each end value. Making it kinda impractical. Hence why most systems which still want the same output use rerolls.

 

So now we're talking about tracking state for masterworking and enchanting for each item type for every avatar. What if my "practice" items use basic materials and my "good" items use special materials but they are the same type?

 

In the example I gave, you can still fail a mulligan roll. All it does is reduce the rate of failures on highly improbable rolls.

I personally don't care what the numbers are so long as what they do aren't insulting a player's valuable time and energy.

If the success and failures are binary, I'd fix that first and worry about the RNG after.

 

Would also like to parrot the "It's not fun" sentiment. I will say I have had more *streaks* of consecutive failures than I have had in any other game. This game requires an obscene amount of solo-misery in a mine if you've any aspiration to enchant/MW and every hour or less I'll manage to see at least one 3-strikes-I'm-Out fail streak with 93% chance of success. Either I'm the most unlucky SOB around (wouldn't doubt it) or there's something seriously buggered somewhere outside the code Chris linked. Statistically speaking I fail consecutively FAR more often than I should.