Craft Fake % succes

[NukedGypsy]

@meesterg

Iam aware that there is a better chance of getting a good item when you find a good base with higher tiers already on it. But again THAT WASNT MY POINT iam purerly talking about tier 1,2,3 while crafting on an item without any affixes on it or with 1/2 affixes on it with tier 1/2 or 3.

I showed you how to craft those, and you laughed and replied with this…

@meesterg

Yes you did, that was really determenistic i loved those tier ones Kappa.

Then you say you want more Tier 3s, but then say they are trash…

@meesterg

Tier 3’s i consider trash items but thats just me sorry.

I am confused. I am going to go play the game. It makes more sense in there.

[SrsBznz]

@washa3

I believe you’re confusing aspects of the system. When we say that LE’s crafting is deterministic, we mean that you get to choose what is crafted onto the item, and upon success you get a deterministic result, the affix you chose is now on the item. Many of (but perhaps not all) of PoE’s crafting do not allow this.

I agree with this.

@washa3

Ultimately, barring some sort of true random number generator, it can be said with relative certainty that LE’s crafting is deterministic.

But this is just you being pedantic beyond reason to one-up him, and not your actual opinion, right?

Because from my understanding pseudo-RNG is popular in programming simply because true RNG can not produce the predicted results within a small sample size.

If you craft 10000 items, true randomness will most likely provide results matching the statistical likelihood you’d expect, but in small sample sizes of say, 10 or even 100, it’s more likely that the result is not the statistically expected outcome.

Throw a regular 6 sided die 6 times. Statistically it’s just as likely to end up on one side as any of the other. But you wont get a near perfect split until you’ve thrown it at least a thousand times. That’s why programs use pseudorandomness rather than true.

[jehzir]

In Neverwinter we used table rolled randomness. Meaning that the die could be rolled 10000 times and the result was the highest result of the rolls. Therefore the roll would be differentiated from the result. A python example

import random

count = [0, 0, 0, 0, 0, 0]

for i in range(10000):
count[random.randint(0,5)] += 1

for i in range(6):
print “Value %d happened %d times” % (i + 1, count[i])

then the results of i were matched to the table and returned the value.

[washa3]

@SrsBznz

But this is just you being pedantic beyond reason to one-up him, and not your actual opinion, right?

No, that is my actual opinion. I was just trying to address another line of reasoning that came to mind as I was talking about determinism – If we only consider the surface of the system that doesn’t deal with the algorithm at all, we might say that crafting is deterministic (like I did). However, without other details, that assumption could be wrong based on a number of factors, not least of which is how EHG generate their random numbers for crafting.

@SrsBznz

If you craft 10000 items, true randomness will most likely provide results matching the statistical likelihood you’d expect, but in small sample sizes of say, 10 or even 100, it’s more likely that the result is not the statistically expected outcome.

This is also true of pseudo-randomness, which aims to approximate true randomness, but isn’t quite truly random.

@SrsBznz

Because from my understanding pseudo-RNG is popular in programming simply because true RNG can not produce the predicted results within a small sample size.

This, I believe, is not true. Truly random number generators are hard. Hard to create, hard to validate. Its much easier to have an approximation to true random. I could be wrong about this.

[SrsBznz]

This is far from my area of expertise, but true randomness is generally not used for multiple reasons, such as the aforementioned unpredictability (even at larger sample sizes) and how much more resources would have to be spent to produce a somewhat accurate representation of a statistical likelihood.

@washa3

No, that is my actual opinion.

I called you pedantic beyond reason, I didn’t say that you were technically speaking in the wrong.

Next, are you going to claim that all the loot (besides that which is 100% guaranteed) is deterministic and not random simply because the algorithm behind the item distribution is mostly likely based on pseudorandomness rather than true?

@washa3

This, I believe, is not true. Truly random number generators are hard. Hard to create, hard to validate. Its much easier to have an approximation to true random. I could be wrong about this.

Well yeah, that’s what I’ve gathered as well. Since increasingly high sequences become harder to validate, and short ones will fail represent the statistical expectation.
Again, this is really not something I claim to be well-versed in.

[washa3]

@SrsBznz

I called you pedantic beyond reason, I didn’t say that you were technically speaking in the wrong.

This some sort of misunderstanding between you and I here. My apologies if I misunderstood your question.

@SrsBznz

Next, are you going to claim that all the loot (besides that which is 100% guaranteed) is deterministic and not random simply because the algorithm behind the item distribution is mostly likely based on pseudorandomness rather than true?

Programmatically, that could be true. The reason I brought up pseudo-randomness in the first place is that the crafting algorithm itself is in question, and I thought of a way it could be nondeterministic, but is relatively unlikely. What really matters is my first statement

@washa3

I believe you’re confusing aspects of the system. When we say that LE’s crafting is deterministic, we mean that you get to choose what is crafted onto the item, and upon success you get a deterministic result, the affix you chose is now on the item. Many of (but perhaps not all) of PoE’s crafting do not allow this.

[jehzir]

@SrsBznz

This is far from my area of expertise, but true randomness is generally not used for multiple reasons, such as the aforementioned unpredictability (even at larger sample sizes) and how much more resources would have to be spent to produce a somewhat accurate representation of a statistical likelihood.

I plain and simple do not understand your comment. I gave you an example of nested randomness. Which is rolling for a random matched with a predetermined table result based on the highest number of the results. You essentially have to get the 16.67% chance correct based on the set table value.

[SrsBznz]

Because you need to run such a high amount of rolls its very resource intensive, and since you can’t reproduce “the same” true randomness, you also can’t reproduce possible bugs and check how solid the system actually is.

My argument isn’t that true randomness is less accurate overall, but because the methods of achieving something close to true randomness are very inefficient, it’s used less commonly if there’s no purpose to creating a randomness that can’t be “cracked” (such as for security encryption).

[jehzir]

that many rolls is not resource intensive at all, compared to talking in kilobytes verse gigabytes. It is done all the time, maybe without you realizing it. You never should be able to reproduce the same true random, then the bug is confined to the result, not the table. That very logic is what is making the Craft % success appear fake when it is “true to itself” and that is all that matters.

[SrsBznz]

Since there are people more knowledgeable than me that can explain it better, I’ll just shamelessly post the following:

Here’s an interesting question I was pondering last week. The .NET base class library has a method Random.Next(int) which gives you a pseudo-random integer greater than or equal to zero, and less than the argument. By contrast, the rand() method in the standard C library returns a random integer between 0 and RAND_MAX, which is usually 32768. A common technique for generating random numbers in a particular range is to use the remainder operator:

int value = rand() % range;

However, this almost always introduces some bias that causes the distribution to stop being uniform. Do you see why?

Let’s suppose that rand() generates a uniform probability distribution; for any call we have a 1 / 32768 chance of getting any particular number. And let’s suppose that range is 3. What is the distribution of rand() % 3? There are 10923 numbers between 0 and 32767 that produce a remainder of 0, 10923 that produce a remainder of 1 and 10922 that produce a remainder of 2, so 2 is very slightly less likely, and 0 and 1 are very slightly more likely. We’ve introduced a bias towards smaller numbers.

This difference is tiny; so small that you would not notice it without millions of data points. But now let’s suppose that we want to generate a random number between 0 and 19999 using this technique, and see what the bias is. How many ways are there to get 0? We could generate 0 or 20000, so 2. How many ways are there to get 1? We could generate 1 or 20001. And so on, up to 12767. But there is only one way to get 12768 out of rand() % 20000! Every number between 0 and 12767 is twice as likely as every number from 12768 to 19999; this is a massive bias.

How can we quantify how massive or tiny the bias is? One technique is to graph the desired probability distribution[1. Here shown as a continuous function, though of course it is a collection of discrete data points.] and the actual probability distribution, and then take the area of their difference.

Every power of two gives no bias at all; half way between each power of two gives an every-increasing bias area.
Here the red line is the desired probability distribution: an even 1/20000 for each element. The blue line is the probability of getting every element with rand() % 20000, and the shaded area is the difference between the desired and actual distributions. The area under both distribution lines is 100%, as you would expect for a probability distribution. A few quick calculations shows that the shaded area is a massive 28%! By contrast if we did the same exercise for rand() % 3 the shaded area would be 0.004%. (Of course the two shaded areas are equal since the total area under both curves is the same. We could just compute one of the areas and double it.)

Every power of two gives no bias at all; half way between each power of two gives an every-increasing bias area.

The moral of the story is: when using the remainder technique, you’re introducing a bias towards small numbers, and that bias can get quite large if the range is large. Even a range of a few thousand is already introducing a 5% bias towards smaller numbers.

How much bias is introduced by the remainder technique?

[SrsBznz]

@washa3

Programmatically, that could be true. The reason I brought up pseudo-randomness in the first place is that the crafting algorithm itself is in question, and I thought of a way it could be nondeterministic, but is relatively unlikely. What really matters is my first statement

Well again, I do agree with your first statement and understand that you were probably just trying to prove a point in your argument against meesterg claims of PoE’s crafting system being less based on RNG and more in the player’s control (which is just silly, imo).

[jehzir]

SRS it was my job to calculate these figures IN MY HEAD for probs and diff calcs. Now you are getting into the split number line methodology which 0 to 10 is 5.5 not 5 as customary. This is resolved through true int and truncation. The traditional way to resolve that is to round up the even and drop the odd above the 5.

Python Random Number Function and Syntax.

The scholarly work behind it is the basis of how I implemented the system mentioned earlier.
Downey Random and Float points

[Shrukn]

@meesterg

for the Exalted stat.

but it…has exalted stat and survive 2 annuls, 2 stability and Guardian to T4. thats me giving an example of where rng was in my favor instead of just crying about the fractures at low %

doesnt change the fact this games crafting is too easy and complaining about it makes you seem kinda spoilt

I found a 14% crit Sovnya last night with T5 melee and guess what it received a Damaging fracture as I went to put T5 minion melee damage and lowered the Crit chance so it wasnt even an exalted item anymore

I guess I should make a rage post about how this game isnt fair etc

[jehzir]

Nah bro, its cool. Those things happen. No risk it, no biscuit.

You made it further than many and that should be celebrated, not looked at with jealousy.

My only complaint was that the number of 3x T3 affixes breaking on applying the 4th T1 new stat was pretty awful. For a 90% success rate it bricked a full tab of Oil flask relics. At one point I had 70 but ran out of room. I asked Mike to give it a look, he did.

Mike gave a sufficient reply, that the number itself was true, but the system itself has a lot of room to be improved upon.

I think the player base wants a balanced system that has both elements of RNG and deterministic without such harsh penalties since the cost are near priceless to replace due to rarity of some affixes.

I think the term instability is not actually correct word to use. It should be impurity (as in smelting iron or steel). As you add or remove components from the kiln the mixture improves or degrades the forging process.

[talkinkorean]

You can’t have deterministic crafting without heavy penalties. I mean you can, but at that point might as well play an MMO that rewards you for the time spent completing objectives. I can’t speak for everyone, but at least for me I enjoy ARPGs for the loot hunt. If I can continuously craft the exact item I want without setbacks then I’ll eventually get bored since I can easily make T20 items with enough time spent. Then it’s all of a sudden a huge jump to get exalted items. The curve is off.

The current system is fundamentally flawed. Straight up. POE does some deterministic crafting but they put it behind so many loops that it doesn’t feel like it. LE straight up does it from the get go. You can easily get a T9 - T12 item that has all the stats you want and I find that boring. I personally think the crafting system needs to take more of a step back and put more into drops being important. Exalted items being dropped only is a step in the right direction, but it’s not enough.

[Shrukn]

@talkinkorean

The current system is fundamentally flawed. Straight up. POE does some deterministic crafting but they put it behind so many loops that it doesn’t feel like it. LE straight up does it from the get go. You can easily get a T9 - T12 item that has all the stats you want and I find that boring. I personally think the crafting system needs to take more of a step back and put more into drops being important. Exalted items being dropped only is a step in the right direction, but it’s not enough.

As ive said in tihs thread, its too easy but thats also being conditioned from 8 years of PoE rng where I just ragequit SSF because the game wouldnt give me a 6link chest and I just quit Ultimatum league for good

I spent a week farming for a chest piece, spent 80% of my materials rolling ‘+1 gems’ basically and then spent ALL my farmed fusings from that week on literally nothing, instead of being motivated to farm more I just ended that little journey

In comparison it would be like farming 100’s of Shards and wasting them on absolutely nothing. you feel robbed

[NukedGypsy]

@Shrukn

it received a Damaging fracture as I went to put T5 minion melee damage and lowered the Crit chance so it wasnt even an exalted item anymore

I had no idea an exalted could lose the exalted affix from a damaging fracture. Thanks for pointing this out. In the long run, I feel this will add even more value to these items in multiplayer just for overcoming that obstacle to max craft an exalted.

[Llama8]

There’s never going to be a fully deterministic crafting systam in a game, there’s always going to be some form of RNG touching it somewhere.

@meesterg

Do i need to make a picture of the poe benchcraft and the harvest crafts to make clear that there is 0 RNG involved on those crafts? Yes you got annulments etc, just like LE has now.

Thats the thing i really dont get.
All the things that you find in delve and from the syndicate members are determenestic. Non RNG based. You find the craft you literally walk to your bench, click your craft done. No 80% brick chance etc.

From memory, what options & when they become available re Harvest, Syndicate & Delve crafting are RNG, so following your logic, they aren’t deterministic. The only fully deterministic bits are the basic benchcrafts that you unlock during the story because they’re always in the same place & you will always encounter them at the same time. Everything has a pinch of RNG thrown in to mix things up.

PoE’s crafting (bench notwithstanding) is a blended system leaning towards the RNG because when you hit that craft button you aren’t guaranteed to get a specific affix (but as I said, there are ways to skew the odds in favour of a particular set of affixes).

LE’s crafting is also a blended system leaning more towards the middle or deterministic side (certainly compared to PoE) 'cause you get to choose what affix you want.

Its a pretty big list compaired to the tier 1’s that you can craft with 100% like NukedGypsy showed. Which you can only get with the glyph scroll (so even an extra cost compaired to POE) .

You’re also very limited in the number of those bench crafts you can use on a single item unless you have specific currency that lets you do more (and that dropping is RNG-based)

Harvest crafting was still quite RNG-y, though on the side of what seeds/mobs crafts you got.

How is the LE system considered determenestic compaired to POE’s system? While LE clearly only has an RNG based system. That keeps baffeling me. Thats just strictly not true.

Because when you use that transmute/alchemy/chaos/exalt/alt/delve currency what you get is fully random. Even if you use delve currency to make it more likely that you get a specific set of affixes, what you get is still random it’s just choosing from a smaller pool.

From memory, this forum has seen this exact discussion as to what counts as random with the rune of removal. Just because a thing is choosing from a different set of results, doesn’t make it not random. And just because you get to choose a specific thing doesn’t mean there was no randomness involved in what options were presented to you.

You can craft really early while leveling in POE and the only thing you pay are the costs.

You can in LE as well. I start crafting at lvl 1-2 & the only thing I’m paying is the costs…

[NukedGypsy]

@talkinkorean

I can’t speak for everyone, but at least for me I enjoy ARPGs for the loot hunt. If I can continuously craft the exact item I want without setbacks then I’ll eventually get bored since I can easily make T20 items with enough time spent.

You are speaking my language. I love the hunt as well. And we all know that a hunter is nothing without the prey. It’s BECAUSE I get so many fractures, that I keep crafting. If I hit max on everything all the time, I wouldn’t craft anymore. It would be very boring.

[talkinkorean]

There’s a spectrum where on one end you want it to be relatively easy to get (6 link in POE is pretty mandatory) and the other end really difficult (all T7 perfect stats would be an extreme example). POE botches that by making 6 links fucking god awful to get.

This discussion is probably going to be moot hopefully by the next major patch. They already acknowledged the current system is problematic. I just hope they move away from or lessen the ability to pick and choose your stats as well as incrementally increasing stat tiers. Do away with bricking items and add diversity to the item hunt (i.e. valuable base items). It’s too stale and monotone currently.