The World of Layonara
The NWN Persistent World => Bugs => Fixed Bugs => Topic started by: Script Wrecked on August 29, 2008, 04:09:51 AM
-
Description: My character has a 65% chance of success in Alchemy. I generally make 14 attempts at crafting in each session (being a box of ingredients). Theoretically, that should average 65% x 14 = ~9 successes each time. For the past 4 out of 5 sessions (4 x 14 = 56 attempts), I have succeeded 6 times in each session (6 / 14 = ~43%).
This is a significant variation, some 22% of total results, or 33% of all successes.
Given the consistency with which this is happening, potentially, this could indicate something awry with how success is being calculated.
Location: Port Hempstead Enchanting Hall, North Point Crafting Hall, Moraken's Enchanting Hall.
Verified:
Reproducable: 4 out of 5 sessions so far.
-
To have less than or equal to 30 successes out of 56 at a 65% chance each, the probability for that to happen would be: 0.051... which is about 5%. In other words, on par chance as with rolling a 20 on a d20. Reasonable enough, is it not? :)
To have less than or equal to 6 successes out of 14 at 65% chance each, the probability for that event would be: 0.075
I'm sure that if you are tossing in the "successful attempt" and/or all successful attempts you have ever had, the chance of success truly is rather reasoanble. :)
-
But I'm not refering to the probability of any particular outcome. I'm refering to the average of all outcomes. If I have a 65% chance of success, from 56 outcomes, I would expect (in the order of) ~36 successes.
Regards,
Script Wrecked.
-
What I'm pointing out is that your "unlucky" outcome still had a very high chance of happening (a little bit more than 5.1%) and there are no strong reasons to believe that something weird is going on here, especially considering that no one has touched this part of the system for probably a very long time.
If you toss in your "lucky" run, you will likely notice that this outcome had an even greater chance of occuring.
As a side note... the probability of that you would have succeeded 36+ out of 56 would have been ~60.4%.
Because of this whole thing is done in a pseudo-random fashion, we can really only talk about probabilities and compare them to see if it's reasonable or not. Rolling a 1 on a d20... do you consider that reasonable? If so, then that was exactly what happened to you.
Edit: Oh, wait... 24/56? Not 30/56 gotcha... That has a considerable lower chance: 0.0005810... which is one chance in 1721 and makes one raise some greater doubts but still, the code has not been modified to my knowledge. It would still be highly benefical if you could also state the success rate of a greater sampling (i.e. the successful round?).
And also... did you do "bulk creation"? Like creating 2 or 4 items at once?
-
Edit: Oh, wait... 24/56? Not 30/56 gotcha... That has a considerable lower chance: 0.0005810... which is one chance in 1721 and makes one raise some greater doubts but still, the code has not been modified to my knowledge. It would still be highly benefical if you could also state the success rate of a greater sampling (i.e. the successful round?).
The successful round was 11 from 14.
And also... did you do "bulk creation"? Like creating 2 or 4 items at once?
Not with results like that. :)
Regards,
Script Wrecked.
-
I have known other people having he same problems with the same odds. with 65% chances on ten attempts per session only making 2 success. Mathematically if you have 65% chances to make an item, on 10 tries you should get a result of 6. Unless I'm wrong which can also happen 1 on 3 times *grins*
-
I am possibly absorbing your luck with only 15% chance of crafting a cougar skin bag. And I made it too. ;)
-
Sometimes you are lucky, sometimes you are not. :D
One day I succeeded 3 times in a row at 60% and later failled 3 times consecutively ...
-
One day I succeeded 3 times in a row at 60% and later failled 3 times consecutively ...
Which is averaging 50%. :)
The probability of 3 successes (at 60%) followed by three failures is:
[INDENT]0.6^3 x 0.4^3 = 1.38%[/INDENT]
Whereas the probability of 3 successes (at 60%) from 6 attempts is:
[INDENT]C{6,3}(0.6^3)(0.4^3) = 27.65%[/INDENT]
Regards,
Script Wrecked.
-
As I understand it, the system is server-wide and so while one person may be unlucky for a spell, as I have surely encountered, another may be extremely lucky, and another still right on the odds. I have been in all three places, often, and I think it is a bit like tossing a penny. I remember seeing a list of the distribution of crafting rolls server wide the last time this issue came up. It showed pretty conclusively that there was a reasonable distribution of rolls.
I hope this is not a bug, and I hope your luck changes.
Gilshem Ironstone
-
I have had days where even though I have an 80% chance of crafting holy water, I roll a 3 about 5 times in a row. and yes, it remains the same number. It doesn't happen every day though, so it may just be bad luck.
-
Crafting Rolls Distribution (http://orth.ca/cgi-bin/cnr_rolls.pl)
-
Without bulk creation and with that successful round added into the calculation... 35 or less out of 70 gives you odds of 0.7% which IMO is still very unlucky... but cannot truly be explained other than by bad luck. :(
Oh, by the way, was any obvious pattern shown for the 6 successes out of 14? Like 1-4 = failure, 2 = success, 3 - 7 = failure, etc...?
I have known other people having he same problems with the same odds. with 65% chances on ten attempts per session only making 2 success. Mathematically if you have 65% chances to make an item, on 10 tries you should get a result of 6. Unless I'm wrong which can also happen 1 on 3 times *grins*
When probability is involved, there are really no "should". Logic is lucky completely in black and white. "Should" means "must" which implies "0 or 100%". :)
-
Unfortunately, from Orth's link, I only get a grey screen with a box titled "Distribution of Layonara Crafting Dice Rolls".
Without bulk creation and with that successful round added into the calculation... 35 or less out of 70 gives you odds of 0.7% which IMO is still very unlucky... but cannot truly be explained other than by bad luck. :(
*cough* Or... something awry in the algorithm. :)
Oh, by the way, was any obvious pattern shown for the 6 successes out of 14? Like 1-4 = failure, 2 = success, 3 - 7 = failure, etc...?
[INDENT]1) S, 5, S, S, 3, S, 5, 4, 5, 5, S, S, 5, 7
2) 1, S, S, 3, 6, 2, S, 5, S, 4, 7, 2, S, S
3) S, 6, 2, 3, S, S, 6, 7, S, S, 2, 5, S, 3
4) S, 7, S, S, S, 7, S, 3, 4, 1, 6, S, 6, S[/INDENT]
Regards
Script Wrecked.
-
The NWN RNG is faulty, and forces "apparent randomness," rather than generating truly random numbers.
What this means is, for every person that gets a string of good numbers, someone else gets a string of bad ones. It seems less apparent in a single-player game, but when it's trying to balance it all server-wide, you get these anomalies.
This is a Bioware Bug, and cannot be fixed.
-
just curious....is this the same way the roll for soul mother works?
-
just curious....is this the same way the roll for soul mother works?
Yes, it is.
-
The NWN RNG is faulty, and forces "apparent randomness," rather than generating truly random numbers.
What this means is, for every person that gets a string of good numbers, someone else gets a string of bad ones. It seems less apparent in a single-player game, but when it's trying to balance it all server-wide, you get these anomalies.
This is a Bioware Bug, and cannot be fixed.
There are no proof of that but rather being a wild guess. Few other games explicitly give you the rolls which could explain why there are very few remarks about them.
What is heavily suspected though, is that the RNG is seeded in a faulty way. First of all, the RNG seems to be seeded per area, and not server wide, which would negate your statement of that someone else on the server would be affected by "your" rolls.
The seed seem to consists of the state of the area (static ones like module placeables and creatures and PCs) with time not (?) being involved. A way to check that? Well... create a module of one's own, save it, perform a roll of some kind, reload and retry. If Bioware has not changed it since... um... 1.64 or something... then the result will be exactly the same everytime.
The game does not keep track of which numbers it "already has given out" and try to balance the number output...
-
you can view the rolls while crafting if you set your dice bag to be broadcast locally. I've done that by mistake once and I was seeing the rolls in the server window.
-
NWN loves low numbers is a saying that goes way back. It's amazing how fast we forget our running successes, and remember, etched into your minds, our running failures. Been there-done that.
-
Actually, I remember my running success well, a whole bunch (fourteen?) of Crystal Rods I at something daft like 25% success.
Regards,
Script Wrecked.
-
Unfortunately, from Orth's link, I only get a grey screen with a box titled "Distribution of Layonara Crafting Dice Rolls".
1 | 45953
2 | 45591
3 | 45469
4 | 45877
5 | 45970
6 | 45742
7 | 45544
8 | 45665
9 | 45901
10 | 45926
11 | 45771
12 | 46076
13 | 45560
14 | 46142
15 | 46425
16 | 45714
17 | 45787
18 | 45810
19 | 45765
20 | 45859
So from above is seems rather well distributed.
Guess someone else is getting the high rolls.
I have failed 5 times in a row on 80% success, it happens :)
-
I have failed 5 times in a row on 80% success, it happens :)
Failing 5 times in a row is fine, as long as you got the commensurate 20 successes as well. :)
Regards,
Script Wrecked.
-
The only sad thing in this is that, the rolls gets distributed across the whole server.
1 it makes the % that you see of your chances of success skewed.
2 it doesn't reflect a very balanced system.
When someone makes a painting, his chance of making it right is not affected by 20 other people painting at the same time around the world. I know that unfortunately it's probably some coded thing in the core system of the game. But it is still sad.
-
The only sad thing in this is that, the rolls gets distributed across the whole server.
1 it makes the % that you see of your chances of success skewed.
2 it doesn't reflect a very balanced system.
When someone makes a painting, his chance of making it right is not affected by 20 other people painting at the same time around the world. I know that unfortunately it's probably some coded thing in the core system of the game. But it is still sad.
For a normal RNG, it has no bearing if you have "your own RNG" or "you are sharing RNG output with others". Just because the previous roll was a 20 (for someone else) does not mean that the chance for the roll after (for you) to also be a 20 is far less than the chance of it being a 4.
The NWN RNG is very unlikely to be cryptographically secure as it is "half proven" to have a somewhat faulty seeding mechanism. But that is not related to what you just mentioned.
So, in summary, do you have any proof at all that the rolls of others (or your own previous rolls) have any meaningful impact on the "future" rolls? In this case, meaningful means that the rolls outputted would be less pseudo random than they otherwise would have been (i.e. compared to if you had your "own" RNG).
-
There's two ways to look at this. The first would be a statistical analysis of all results over a long period of time. orth's statistics link does a pretty good job of illustrating to the rolls shake out.
1 | 45953 - 5.013%
2 | 45591 - 4.974%
3 | 45469 - 4.961%
4 | 45877 - 5.005%
5 | 45970 - 5.016%
6 | 45742 - 4.991%
7 | 45544 - 4.969%
8 | 45665 - 4.982%
9 | 45901 - 5.008%
10 | 45926 - 5.011%
11 | 45771 - 4.994%
12 | 46076 - 5.027%
13 | 45560 - 4.971%
14 | 46142 - 5.034%
15 | 46425 - 5.065%
16 | 45714 - 4.988%
17 | 45787 - 4.996%
18 | 45810 - 4.998%
19 | 45765 - 4.993%
20 | 45859 - 5.003%
(NOTE: Percentages added by me. Results rounded to last decimal point. Total sample size: 916547 crafting attempts)
As this data shows (and it is live data taken over the last few years), the distribution of numbers is fairly even across the board.
Any success or failure calculation is nothing more than a straight comparison of numbers. It's a DC vs. a d20. That's it.
As with most "random" number generators in a computer, the result is only "pseudo-random" and as such may occasionally be subject to apparent non-random statistical perturbations.
Taken over a large number of results, statistically speaking, they're hovering right around an even distribution.
Second:
But I'm not refering to the probability of any particular outcome. I'm refering to the average of all outcomes.
The problem is, each result, each roll of the dice, is not a statistical element but a result on its own. There is no statistical leveling or any other sort of mechanism to ensure that for each d20 rolled that in 20 rolls each possible result will have showed up once...or even in 100 rolls that each result will have showed up 5 times...or whatever the calculated probability should be.
Really, when the crafting system tells you that you have a 65% chance of success, it means for that particular attempt, not all attempts.
From personal experience, I've had both more and less success in any given session than the crafting system has told me I should statistically have.
-
For a normal RNG, it has no bearing if you have "your own RNG" or "you are sharing RNG output with others". Just because the previous roll was a 20 (for someone else) does not mean that the chance for the roll after (for you) to also be a 20 is far less than the chance of it being a 4.
The NWN RNG is very unlikely to be cryptographically secure as it is "half proven" to have a somewhat faulty seeding mechanism. But that is not related to what you just mentioned.
So, in summary, do you have any proof at all that the rolls of others (or your own previous rolls) have any meaningful impact on the "future" rolls? In this case, meaningful means that the rolls outputted would be less pseudo random than they otherwise would have been (i.e. compared to if you had your "own" RNG).
well, that's what i keep hearing in these very thread.
-
if you took a D20 and rolled it say 500,000 times you would end up with a very close even distribution. So think of it this way your rolls are evened out over the course of your play time.
-
Perhaps I should have done this initially. Here are my numbers(1):
% TRY EXPECT ACT DIFF
--- --- ------ --- ------
15 6 0.90 2 1.10
20 5 1.00 2 1.00
25 91 22.75 20 -2.75
30 49 14.70 13 -1.70
35 564 197.40 200 2.60
40 167 66.80 74 7.20
45 608 273.60 306 32.40
50 332 166.00 165 -1.00
55 799 439.45 444 4.55
60 367 220.20 224 3.80
[COLOR=Blue] 65 745 484.25 470 -14.25[/COLOR]
70 118 82.60 76 -6.60
75 632 474.00 472 -2.00
80 7 5.60 6 0.40
85 288 244.80 240 -4.80
90 46 41.40 44 2.60
95 158 150.10 147 -3.10
100 604 604.00 603 -1.00
Overall, not as bad as those four runs at 65% would suggest.
Regards,
Script Wrecked.
(1) assuming I can accurately pull these numbers out of the logs.
-
A quick calculation gives your average diff as 1.025, which, if I understand your math, skews things slightly in your favor.
Though I'm not sure how you got this result:
100 604 604.00 603 -1.00
At 100% success, you always succeed.
Though if we can mostly trust these numbers, it looks like you should attempt things at 45% ;)