test content
Logo
What is the Arc Client?
Install Arc
Will the % chance to Fuse ever work correctly? As well as skill kits ?
abombination247
Member Posts: 1,279 Bounty Hunter
All the % chances to complete anything is completely off as many have noticed. I wonder if there are any plans to fix this so it is actually the % that you want it to be.
First skill nodes. Its a 75% chance which is pretty good in our favor then it fails 5 times in a row. This just isn't possible with this % no matter how you want RNG to be a part.
Next fusing. On the 95% chance to fuse lower ranks they fail way to much its 95%. Recently I have been doing a large amount of fusing I am trying to be like well if I fuse 100 items maybe it balances out in the end. WRONG. Just now doing a rank5 to 6 which is a 40% chance meaning almost half the time it will fuse. It failed 8 times in a row. Then its not like it compensates and the next 4 or 5 fuse. That one fused then the next 4 fail. So the % is just way off.
All I am asking is if you want the actual %'s that you have listed then please fix this cause a lot of players are really frustrated.
Or just put what the current % is. and drop all the displayed ones by half and say sometimes its buggy and will fail 8 times in a row. good luck.
First skill nodes. Its a 75% chance which is pretty good in our favor then it fails 5 times in a row. This just isn't possible with this % no matter how you want RNG to be a part.
Next fusing. On the 95% chance to fuse lower ranks they fail way to much its 95%. Recently I have been doing a large amount of fusing I am trying to be like well if I fuse 100 items maybe it balances out in the end. WRONG. Just now doing a rank5 to 6 which is a 40% chance meaning almost half the time it will fuse. It failed 8 times in a row. Then its not like it compensates and the next 4 or 5 fuse. That one fused then the next 4 fail. So the % is just way off.
All I am asking is if you want the actual %'s that you have listed then please fix this cause a lot of players are really frustrated.
Or just put what the current % is. and drop all the displayed ones by half and say sometimes its buggy and will fail 8 times in a row. good luck.
Post edited by abombination247 on
0
Comments
But even after several people posted supporting evidence, rather than posting their own results, the usual Cryptic fanboyzs just chants the mantra of "RNG is RNG"......
Sample size is way to small in these examples to be trying to analyze %'s from RNG.
That would be accurate if a true RNG was used. But it is not
Those examples of 5/6 times, sure, but people have tried this with several hundred, and the percentages are still off. That is way past the number needed for statistical significance.
Having said that though, statements like this just aren't true.
"First skill nodes. Its a 75% chance which is pretty good in our favor then it fails 5 times in a row. This just isn't possible with this % no matter how you want RNG to be a part. "
If failing 5 times in a row wasn't possible the RNG would be broken. This should happen approximately one out of a thousand times, and considering the many thousands of players running around looting nodes it should happen quite often (throughout the game).
My experience is this:
- from 99 rank 4 you get 19-22 rank 5. 20 successes and 20 fails means the chance is in the vicinity of 50%.
- from 16 rank 5 to 1 rank 7 I average 10 green wards (been 3, been 23, I count with 10 and usually end up with less and then once in a while a terrible streak consumes the formerly spared ones). 5 success 10 fail is 33% chance.
So I think they aren't actually off that much with the fuse chances.
Making a 100 rank 7s from rank 4s is around 18.000 Fuse button clicks. It seems much, but what I know about statistics tells me that number still isn't a large enough sample pool to eliminate RNG, so there may be people with both worse and higher numbers.
I'm sorry that I can't show logs about this, but since making profit in the AH depends on whether I know these numbers correctly, you can pretty much believe me when I tell you, I even told you worse numbers then what the reality is, because I'm being pessimistic when assessing the risk/gain ratio before investing.
Although with the profession nodes, I also see too much 3 fails in a row up to 5 fails in a row. I didn't count those, but I agree it _feels_ like it isn't 75%.
I can't really tell why there's a fail chance in the first place actually, double the kit price and go from 75% to 100% if you ask me.
75% is just a nuisance.
It's not the only thing I do for profit and I don't always do it because it's not always profitable.
But when I do play the fusion game, I don't just blindly gamble. I calculate the profit margins using current market prices and the percentages given by the fuse window. If the percentages were significantly off, I wouldn't be making any profit. I'd be losing AD to the process. But I don't lose. I gain. A lot. About as much as the math tells me I should. Sometimes more, sometimes less.
I can't say much about skill nodes because I never bothered to keep a record of them... but the fusion chances seem to check out just fine across large enough sample sizes.
The pseudo RNG used in this game does seem to be "sticky" though. You rarely get anything close to an even distribution of successes and failures. It tends to come in streaks of failures and streaks of successes. Even when it's not streaking outright, it still heavily weighs in one direction or the other.
Anyway, long story short... I think the percentages are true. It's just that the RNG is very streaky.
which means u can do 4 times... and the next fuse still has the same chance to fail/succeed!
Exactly. I remember the same arguments about the "15% chance to proc extra potions" ability in WoW..... "how can I fail 10 times in a row!??!?! Broken!" Well, no - if you throw it in a spreadsheet, you see that it's quite possible to get that many failures at a 15% chance. And as you say.... even when something is 1000-to-1 odds of happening, there's still many thousands of attempts occurring daily.
Some people should learn how to write (and spell) properly too....
Again, it is not a true RNG.
I don't see how a pseudo RNG versus true RNG argument is relevant to the post you quoted?
Pseudo RNGs aren't necessarily designed to produce smooth distributions. In fact, it's something they typically try to avoid in an attempt to more closely mimic true RNG and be less predictable.
Actually pseudo-RNGs ARE necessarily designed to produce smooth distributions over the (very) long-term. Reputable systems also try very hard to avoid extreme streaky behaviour in the shorter term.
Over large sample sizes, of course... if number generator, for example, generated even numbers more often than odd numbers, it wouldn't make for a very good random number generator.
On the other hand, if, across a sample size of 10 (as proposed by the originally quoted post), it always generated exactly 5 even numbers and 5 odd numbers, it would be just as useless of a random number generator... if not even more useless. After all, that would allow you to predict the properties of the next number it would generate, given a history of generated numbers.
Unless specifically designed for a purpose that requires even distribution, modern PRNGs will avoid both of those extreme scenarios. Because, for most purposes, you don't want the user of the PRNG to be able to predict what the next number will be.
As to the overall 75% success rate, compare all the times you've broken 5 in a row with all the times it works on the first try. What's that? You don't remember all the times you succeeded on the first try? Time for a crash course in basic psychology.
http://en.wikipedia.org/wiki/Negativity_bias
Neverwinter Census 2017
All posts pending disapproval by Cecilia
Actually I've found the 75% to be fairly accurate, with a margin for error just under 1 in 4 kits break on the first attempt, 1 in 16 on the second and one in 64 on the third, in fact I've only had one node take more than 4 tries since open beta started.
We have no idea if they're accurate or not. If an item reads that there is a 25% chance we have to hope that's really the case. A bug or bad code could mean it's really 15% instead. Then people will complain the proc rate isn't right via testing, while others will say random is random, and nobody will ever know for sure.
@headlesshorseman
I have had multiples case where it has been over 10, a few times over 20, and a couple where I just gave up. Oddly, the longer streaks seems to be mainly in Neverwinter Graveyard.
If you've actually had a 75% success rate trial fail 20 times in a row, well, that's a one in a trillion event. If it actually did happen, it would support the idea that the success rate wasn't actually 75%. However, Ockham's Razor says you're exaggerating or misremembering. Perform a controlled test, and replicate these results. Record it in video, or with enough screenshots to make a good case, or I'm just calling it BS. If you can get solid evidence beyond yours and others' anecdotes, send it to the devs and I'm sure they'd look into it when they've got time.
I fused. 400 enchants. that is too small ?
PS. Buy my rank 8's I made on the AH since I spend 500k on preservation wards.
Neverwinter Census 2017
All posts pending disapproval by Cecilia
That begin my point - at the very least some nodes are bugged or the RNG is insanely streaky.
In any case since I doubt you will find anyone who bothers recording every minute of their game-play, I think you are pretty safe intoning RNG is RNG.....
"Sample size is way to small in these examples to be trying to analyze %'s from RNG."
Central limit theorem.
As a statistician, I take an interest in all the probabilities within an MMO...
If I try and have 4 successes in a row...does that mean the rate is 100%. There are simply not enough observations to draw that conclusion. The same goes for flipping a coin. If I have two heads in a row, is that 100% chance to flip heads every time?
/boggle
Yes, yes it is.
I work with RNG's; some of them have been shown to have problems. It's a very rare event. Excepting amateurs' (generally bad) "homebrew" RNG's, finding a previously unknown and provable serial correlation in one could get an author published in a professional journal.
Pseudo-RNG's can be investigated by rigorous statistical methodologies; if you want to assert that one isn't working right, that's the way to do it. If you can't or won't, then don't be surprised if anecdotal assertions are ignored.
Other things can go wrong in code, even if the RNG is working perfectly. The threshold for comparison against the RNG draw might be wrong. There can be quite a bit of logic between the actual RNG and the test for success that people claim isn't working right. However, the developers have certainly given that code review, testing, and QC checks on the way to the game going live. If you think there's a bug in something this basic, you have to give them good reason to look into it.
Again, capture some screen shots or get some video. You don't need every second of your day, just get before and after screenshots of the skill node attempts. Make sure the files have good date stamps on them. Host them somewhere and send a summary and a link to one of the Cryptic reps here. I'd suggest log files, but I don't know if NW's log files record skill node or fusing attempt success/failures. And even if they did, text files are far too easy to fake up. If any of the readers here are certain a skill node is bugged or that the RNG in the game isn't working well, then put the effort into gathering evidence.