Thread: What the heck is wrong with crafting?!?!?!

I have two characters with crafting skills at 70% with crit items etc. I have been crafting armor/weapons for new alts and I have, in the last few weeks, had a rash of failures to get criticals! &&&!?!?!?!?!?

I try to get a crit at 70% 15 times and get....3 criticals? And that seems to happen a lot and its starting to wear thin on the nerves. Have they nerfed criticals in crafting?

Nothing wrong with crafting. You are seeing normal random number generator in action.

You flip a penny hundred you are not going to see HTHTHTHT. You are going to see places with HHHHHHHHHH in a row. You are going to remember the 10 heads because they are unusual. It has to occur.

The only way to get of the long strings with 70% chance of FFFFFFF or SSSSSSSSSSS is to remove the random. Go to a system where you have two recipes each with a different material list. Each recipe has a single product. Many of the guild recipes are the alternative recipe that only gives the equivalent of the critical reslt for a one shot recipe with two possible products.

Yes, unfortunately randomness can do those nasty things to you...

However, sometimes I also have the feeling that the (pseudo-) random number generator would not pass many quality cheks, because I've also had my share of unlikely streaks of good/bad luck. Unless Turbine uses special CPU instructions (only available on fairly new CPUs I think), you can only have pseudo-random numbers, that means, a deterministic algorithm that just *appears* to generate random numbers, and there's pretty good but also horribly bad algorithms for that.

If you do the math, statistically the chance to get just 4 "blanks" in a row with 80% crit chance (e.g. +18% tools & crit scroll) is already as low as 0.16% (0.2^4), or 6 in a row with 68% (no scrolls) is 0.1%, yet I've had that happen a litte bit too often for my taste to fit to those chances.
And for some weird reason, one of my characters seems to have bad luck more often than all other toons together...but just from my results it's impossible to determine if PRNG sucks or if I'm just a very unlucky person.

Only Turbine can run the PRNG through various statistical tests to determine it's quality.
In any case it's at least better than this:
http://xkcd.com/221/

You are correct that Turbine uses a Pseudo Random Number Generator (PRNG). No application designer writes their own PRNG. Every software designer including me uses one of the standard PRNGs that are commerically available. Turbine uses Microsoft C++ that is part of Visual Studio. Turbine uses the Microsoft PRNG that is part of the Microsoft libraries. There must be millions of applications that use the Microsoft PRNG. That PRNG is well tested.

PRNGs are one component of Discrete Mathematics. You need one or more doctoral level expoerts to design a good PRNG. You got to test by fire by running trillions of times to make sure it really does generate a Uniform distribution - each number is equally likely. There are no six, 10, 100 number generators. A PRNG has a large number of possible numbers it can generate - 16,000 or more. As a designer you convert the number generated to the smaller range of numbers like 1-100 that you need.

Discrete Mathematics includes PRNGs, encryption, decryption and code breaking. PRNGs always produce the same sequence of numbers when the seed (starting value) is the same. As a user you do not know the seed, you can not know the sequence. In decryption, you have the sequence all you need is the key (PRNGs call the key a seed). With the key you can get back the original data.

The problem that a PRNG has is that they are prone to long strings when you reduce the range of numbers from 16,000 to 100. There is no way to solve this problem because every number depends on the previous number(s). A PRNG is not random - does not generate independent events like a die does.

True random numbers are mathematically construct. A simplifaction of real life. There is no such thing as a truly random number. All events in real life are dependent on what happened in the past. The coupling make be very strong or weak. It is still exists.

Whenever I develop software using a random number generator, I am not looking for random numbers. It is always simulation application. I need a generator that has remembers what happens last time to generate a new number. PRNGs can work resl well. For example, when you are dealing with bit corruption in messages going over the internet. It is very common to get multiple errors in a row. You take a static hit to a cable. The spark does not go away after blowing up one bit. If you lost bit 43, there is a really good chance you are going to lose 44, 45 and 46. A single bit error correction technique isn't going to help. You need to build in a good detection method combined with a retransmission.

Summary, with a PRNG you can be assurred that over a long haul - all the numbers are equally likiely regardless of the range your are using. Odds are you will see excessively long strings when using reduced ranges like 1-100 which can be very frustrating to a single user. One way to combat the long strings is to stop trying after a few failures. Log in another character. Try again later. If you are hot, you might as well keep making stuff.

Originally Posted by Yula_the_Mighty

No application designer writes their own PRNG. Every software designer including me uses one of the standard PRNGs that are commerically available. Turbine uses Microsoft C++ that is part of Visual Studio.

I totally agree on the first part, but sincerely hope you are wrong on the MSVC part... ^^
I've written some monte carlo path-tracing code (to render "photo realistic" images) where you need a good PRNG even if it's just to verify the correctness of your more regular sampling patterns. And while i never tried myself (Linux coder :P), I've seen other people rendering images with the random function of the MS C standard library, just for the laughs. It looked...interesting...too bad the old ompf.org threads seems gone, there were lots of funny experiments and accidents...shoot I should do some more coding again, if only the stuff going on in my head wouldn't regularly keep me from sleeping.

Anyway, I'd rather not trust closed source implementations unless they tell the algorithm and exact variant used, like Mersenne Twister as described in paper XY etc. so I can read up if it really fits my needs.

Originally Posted by Yula_the_Mighty

There is no such thing as a truly random number. All events in real life are dependent on what happened in the past. The coupling make be very strong or weak. It is still exists.

Now I'd never dare to claim that, quantum mechanics to my knowledge says differently, things like radioactive decay are simply not predictable, apparently there is no known reason why some atom decays now and not 5 minutes later or earlier...but then again, what do I know...

Originally Posted by Lynx3d

Anyway, I'd rather not trust closed source implementations unless they tell the algorithm and exact variant used, like Mersenne Twister as described in paper XY etc. so I can read up if it really fits my needs.

I have not checked lately. Microsoft used to give the source for their PRNGs. All the PRNGs that I've worked with are very simple math formulas. It is easy to convert the executable code into human readable form.

Originally Posted by Lynx3d

Now I'd never dare to claim that, quantum mechanics to my knowledge says differently, things like radioactive decay are simply not predictable, apparently there is no known reason why some atom decays now and not 5 minutes later or earlier...but then again, what do I know...

What you mean is that we as humans do not understand radioactive decay well enough to predict it. You examine the progress of human knowledge over the years. Many things that were totally incomprehensible are understood today. Many things that could not be predicted can be predicted today.

For example, if you had asked someone 400 years ago if it would be possible to figure out whether a baby would develop certain diseases via genetic testing. They would have laughed at the Bunny. They probably laughed at the concept of DNA, genes, genome sequencing and genetic testing. When I was in engineering school, I did not think it would be possible to sequence a human genome.

Just because it not possible in 2012 to predict the decay of an atom, does not mean that it is a random event. The only thing we can say for such is that we do not know.

Even standard hole drilled die do not generate a uniform distribution. Some numbers are more likely to come up than others due to way they are drilled. You add in the people factor of how you roll a die to further skew the results.

Food for thought. Think a little bit about the following explanation. A standard die has six holes on one side - the opposite side has one hole. The die is imbalanced. The side with one hole is heavier than all the other sides. It more likely (not by much) that side with the one hole will end up on the bottom. That means that the number six is the most likely number to be generated.

I would suggest reading about:
http://en.wikipedia.org/wiki/Chaos_theory or as it is sometimes called the "Butterfly Effect". Essentially this area of mathematics is the study of systems where the "seed" (initial condition) make massive changes in how the system evolves over time.

It is also covers systems what should be a minor input to a running system results in a massive change in behavior. Lorenz wrote a paper on prediction. Does a flap of Buttery's wing in Brazil set off a tornado in Texas?

Originally Posted by Yula_the_Mighty

What you mean is that we as humans do not understand radioactive decay well enough to predict it. You examine the progress of human knowledge over the years. Many things that were totally incomprehensible are understood today. Many things that could not be predicted can be predicted today.

For example, if you had asked someone 400 years ago...

No. That line of argumentation has been tried by Einstein, Podolsky and Rosen about 80 years ago. It's been proven to be wrong. You might want to read http://en.wikipedia.org/wiki/EPR_paradox, especially the paragraphs about hidden variables and Bell's inequality.

From all what we know quantum mechanics implies "true" randomness.

Originally Posted by Beornson

No. That line of argumentation has been tried by Einstein, Podolsky and Rosen about 80 years ago. It's been proven to be wrong. You might want to read http://en.wikipedia.org/wiki/EPR_paradox, especially the paragraphs about hidden variables and Bell's inequality.

From all what we know quantum mechanics implies "true" randomness.

Why are you using a wiki page to cite anything? Any college instructor will tell you that is the last thing to use to cite anything credible.

At any rate, ever since the beginning of Lotro I have always hated the critical success ratio in crafting. It is far too random and doesn't weigh in the percentage of success accurately.

I like this thread. So wanted to contribute my experiences:

I have a level 17 jeweler with level 13 crit tools, just yesterday was doing some crafting for level 38 items and got crits 9 out of 10 tries. Switched to level 54 items and got 2 crits out of 6 tries, then switched back to level 38 for 3 out of 3 crits, then went to level 35 for 1 out of 4 for crit. This one toon I have noticed before that certain level items crit better than others, even when crafting different level items in one crafting session. My level 57 guard is the same way with level 55 crit tools. So 55% for level 13 crafting seems as good as 80% crafting at level 55. The rest of my crafters more or less do good over all and seem random but with a lean towards whatever the crit is expected, but the jeweler and metalsmith tend to depend on what level items I'm crafting.

Go figure?

Originally Posted by Borrhavan

At any rate, ever since the beginning of Lotro I have always hated the critical success ratio in crafting. It is far too random and doesn't weigh in the percentage of success accurately.

There have been many threads about the randomness of crafting and crits. All of the ones that I have read where the poster actually put effort into recording their crits over a long period have come to the conclusion that the percentage of observed crits very closely mirrors the expected percentage. In almost every case where someone is claiming that they do not, they talk about very short runs of crafting or just go by what they remember.

There's nothing wrong with crafting. It's working as intended.

See sig's first link, Lorebook entry. The experiments have been done.

Of course players are more than welcome to contribute their own 1000-3000 recorded samples. (Before anyone scoffs at that number, see the 'important lessons' section at the very top of the Lorebook entry. )
____________________
Crit crafting & the RNG | Dumb noob things we've done in LotRO... | LotRO Content Graph by zone/level
"Be not so eager to advance that you fail to experience the moment, in life and in all things." -Deluros, Rivendell
legendary: adj.
1. Gilded (having a pleasing or showy appearance that conceals something of little worth)
2. Having an appearance of grandeur that obfuscates by complexity and arbitrariness
3. Disposable