test content
Logo
What is the Arc Client?
Install Arc
Bizarro RNG explained (Not Random!)
zeusom
Member, NW M9 Playtest Posts: 810 Arc User
So for a good while I have paid close attention to and even sometimes carefully tracked most everything RNG related in NW using a handheld notepad (150+ eSoT runs, RP crit patterns, Dragon drops, % item upgrades, Lockbox drops, much much more) and have a good amount of data to draw upon. Through all of this I have noticed one odd but consistently recurring trend with Neverwinter RNG:
LONG-TERM OBSERVATION:
RNG drop events and lack thereof tend to occur in clusters.
Why might that be?
For Example: 150 eSoT runs tracked and 5 Draconic Rings dropped total but 3 of them within 5 runs (runs 122-127). I have seen many many more similar examples of 'RNG clustering' recorded. Clustering observations got to a point well beyond reasonable anomoly - I am certain there is something going on and NW rng is not truly random but why.
Neverwinter and most allpc game RNG is not truly random but is pseudorandom. Although written and applied with best intentions, the problem with using computer code for RNG is that patterns emerge which would be entirely consistent with loot drop clustering as well as long dry spells. The game RNG is determined by algorithms written in binary code. These algorithms are 'good enough' for games and such, but the results are in fact entirely predetermined by the code.
There are true random number generators but they are based on truly random chaotic non-digital processes such as the totally chaotic static of atmospheric noise such as here: http://www.random.org/. With these, no patterns emerge.
Pictures worth 1000 words (from a website about RNG):
BITMAP GENERATED WITH CPU (PSEUDORANDOM) ALGORITHM (Notice emergent patterns)
Notive the clearly emergent patterns!! Same thing happens in Neverwinter RNG.
BITMAP GENERATED WITH TRUE RNG, Atmospheric noise (No pattern)
This is what true randomness 'looks like'.
Wikipedia on Pseudorandom RNG: http://en.wikipedia.org/wiki/Pseudorandom_number_generator
Wikipedia on True vs Random (#2 down the page): http://en.wikipedia.org/wiki/Random_number_generation
Would be nice were Neverwinter to draw on true random number generation to eliminate this unusual behavior and make an unprecedented leap to use (the first?) true RNG in a pc game.
Interesting huh. Cheers!
LONG-TERM OBSERVATION:
RNG drop events and lack thereof tend to occur in clusters.
Why might that be?
For Example: 150 eSoT runs tracked and 5 Draconic Rings dropped total but 3 of them within 5 runs (runs 122-127). I have seen many many more similar examples of 'RNG clustering' recorded. Clustering observations got to a point well beyond reasonable anomoly - I am certain there is something going on and NW rng is not truly random but why.
Neverwinter and most allpc game RNG is not truly random but is pseudorandom. Although written and applied with best intentions, the problem with using computer code for RNG is that patterns emerge which would be entirely consistent with loot drop clustering as well as long dry spells. The game RNG is determined by algorithms written in binary code. These algorithms are 'good enough' for games and such, but the results are in fact entirely predetermined by the code.
There are true random number generators but they are based on truly random chaotic non-digital processes such as the totally chaotic static of atmospheric noise such as here: http://www.random.org/. With these, no patterns emerge.
Pictures worth 1000 words (from a website about RNG):
BITMAP GENERATED WITH CPU (PSEUDORANDOM) ALGORITHM (Notice emergent patterns)
Notive the clearly emergent patterns!! Same thing happens in Neverwinter RNG.
BITMAP GENERATED WITH TRUE RNG, Atmospheric noise (No pattern)
This is what true randomness 'looks like'.
Wikipedia on Pseudorandom RNG: http://en.wikipedia.org/wiki/Pseudorandom_number_generator
Wikipedia on True vs Random (#2 down the page): http://en.wikipedia.org/wiki/Random_number_generation
Would be nice were Neverwinter to draw on true random number generation to eliminate this unusual behavior and make an unprecedented leap to use (the first?) true RNG in a pc game.
Interesting huh. Cheers!
Post edited by zeusom on
0
Comments
There is definitely something wrong. Exactly what is impossible to say without examining their code. I used up 23 preservation wards today trying to upgrade something that had a 50% chance of success. The statistical chance of that legitimately happening is so small that it should never happen. Others have had this happen as well. I also went 8 weeks running tons and tons of eLoL without a belt drop, then on the 9th week I got 9 over 2 days. That just doesn't make any sense at all.
My suspicion is that it's more so related to bugs in their code, as opposed to issues with patterns in the random number generator. Something more along the lines of this: http://asheron.wikia.com/wiki/Wi_Flag.
Far from likely but it's bound to happen.
...and irrelevant. Sure in a pseudorandom generator patterns emerge ---- if you are the only one using it just like your image.
The problem is the RNG is being fetched by thousands of different individuals at completely different times and those timings are completely random. Any "pattern" is thus spread out truly randomly across the entire gamespace and players.
I also would put more emphasis on possible Wi flag type of bug.
Encounter Matrix | Advanced Foundry Topics
PHP rand() on Windows is not cryptographically secure -- they literally say that in the documentation, which is why you are seeing the patterns. Given that Neverwinter is using C++ and presumably can call a better library than PHP (a truly <font color="orange">HAMSTER</font> language if there ever was one), you have no proof that this is the pattern being used. It could even simply be using /dev/urandom, for all you know ( which looks like: http://2.bp.blogspot.com/-8amoAOhDJIw/VJNRaeAxqeI/AAAAAAAABjU/miwzwGTSuXg/s1600/random.dat.bmp )
Second, imagine what happens when your need for RNG exceeds your true entropy source (e.g. more players just logged in). Your RNG call will literally block (to the user, this will appear as a total freeze of his app) until you have more entropy. If you're a programmer for UI or server applications, this is really unacceptable.
Pseudorandom algorithms were invented for a reason. Unless a task absolutely need trues randomness (i.e. high grade cryptography), a cryptographically secure pRNG is good enough.
If you think the RNG isn't random enough, post a large-N, statistically significant test result.
Once I failed 27 times in a row trying to upgrade R6. Even started a thread about it XD
Between both games I've opened in excess of 10,000+ lock boxes (easily). Something to try if you're into that sort of thing.
You assert: "Same thing happens in Neverwinter RNG" but fail to provide a Neverwinter Bitmap for us to compare with the above bitmap which you downloaded from random.org.
If you never generated a comparable bitmap from Neverwinter how can you make the assertion that Neverwinter produces the same bitmap?
Once I talked to my friend that played Battle of the Immortals a few years ago, and he said that they always went to the least populated instances/areas to have the highest chances.
WTB Class Reroll please
All seven of my characters are working their way through the jewelcrafting profession. They're all in sync with their tasks as they all started at the same time. Yesterday the first three characters all got a green Jewelcrafter instead of a white Assistant Jewelcrafter when performing the upgrade Gem Setter task.
I'd be fascinated to know whether this was a weird bug in my favour or working as intended. It's odd either way.
Pseudorandom numbers are perfectly fine, BUT they can certainly be messed up. If Cryptic rolled their own RNG it could be less random than you'd hope. Even if they have a good RNG it could be that they are seeding it in a manner that messes things up.
So there could well be something wrong, but you are not going to figure it out with even hundreds of observed results. After all, you don't know what numbers are being used behind the scenes (odds and loot tables and all that).
Fact: Computer-generated "randomness" cannot be true randomness because it's all based on a mathematical calculation rather than chaos (yes, I am actually familiar with it and how to do it from my long-time-ago computer programming stint). So it comes down to the method and odds chosen my those writing the random-generation code. There are some here who are coders and I'm sure they can confirm what I'm saying.
Hence: OP is correct but it should not be any surprise to anyone familiar with how computer randomness-generation works.
I just tried to upgrade an artifact from epic to legendary. It was supposed to have a 10% success rate but it took me 43 green wards to finally got it done. Interestingly, I also had similar "bad luck" with upgrading another artifact 2 double RP weekends ago... and also wards sale... Hmmm, the morale is I will not use my wards in next double RP weekends or when a ward sale is in progress.
Also another classic pattern rng fail, in Tuern skirmish, 2 months ago. Did 4 runs, I opened chest, got book, next 2 chests got dragon gem, artifact, last chest got belt. This happened TWICE in 2 weeks in almost same order, book, artifact, belt.
Thats a poorly implemented rng method, for a game that is literally based on rng. How can someone expect from people to trust them with their money, when clearly there is rng code failure in almost all of game's aspects, is beyond understanding.
While constructive feedback is our only way of helping, you cant possibly expect from people to refrain from bashing at everything if this feedback goes mostly unnoticed.
I also remember people theorizing in CO (same game engine, I think) that they were seeing drops in succession.
Nice to see someone fiddling with numbers though 😃
Most likely just random
Fox Stevenson - Sandblast
Oh Wonder - Without You
Do not go gentle into that good night.
Rage, rage against the dying of the light.
- Dylan Thomas
It doesn't really matter. The odds as were posted to get 23 fails in a row for a 50% task is 1 in 10M. It can happen, but it's so rare that most players should never see it. It's happened to me more than once. I could care less about "self reported" data. I have my own data and I know these types of failures should NOT be occurring with the frequency that they do. That tells me something is wrong with their code.
More worrying though is the secret modifiers that cryptic uses to "enhance" the numbers that do get returned by the lookup tables. It's too computationally expensive to use modifiers directly within the RNG so they are tacked on at the end, sometimes with results that are much too predictable. The modifiers themselves are likely very complex, taking many bits of data to create them (such as age of account, how often the user plays and for how long, have they ever spent real money, what their current character stats are etc, along with a seed number that is supposed to provide an additional layer of randomness but likely is responsible for those long strings of 50% chance failures in a row that are statistically highly unlikely, but are seen in game on a regular basis.
It's unfortunate that more powerful random number solutions are not available (or too expensive) to use on the scale such as used by a video game entirely based on chance. Cryptic can't really be blamed for using shortcuts for RNG, but they should be aware of the pitfalls and do more to put in safety nets to keep things like the lockbox bug (stuck lookup and modifier caused by rapid polling when hundreds of boxes are opened in a short period) and the chest bug (low priority RNG producing no randomness at all delivering exact same contents for extended periods). Things that affect their bottom line get attention (lockbox bug). Things that affect YOUR bottom line (chest bug, ridiculous strings of failures on a high chance roll) likely don't get more than a slightly adjusted modifier or nothing at all.
I also notice that if i go into a new zone I will get some refinement stone drops at the beginning and then it will stop completely. Maybe they put some code in there to stop farming bots or something.
It reminds me of the soft caps on some of our character stats. It makes the game feel quite rigged at times.
After reading responses I agree a more likely cause than simple pseudorandom rng patterns is something code-related, such as happened with the Wi flag case example in second post. Maybe someday some developer will discover the source of the strange RNG patterns it as happened with Wi flags.
Sopi SW Youtube channel pvp brickabrack
myles08807 said, "Back in my day, we didn't have any of this fancy Mulhorand gear while we were leveling . . . we walked uphill both ways while dying once every five seconds while leveling, and we liked it fine!" . . . Now, get off my lawn, you kids!"
pointsman said, "I don't rue the game. In fact I don't feel any regret for the game at all."
looomis said, "I don't like people changing to alts and then bragging about their mains like schizophrenic role players."