| Author |
Topic |
Haram Haram |
Posted - 2009.12.11 09:58:00 - [ 31]
 Originally by: Beldor
btw:i am doing invetion 1+ year.But i never get failrow as now(last 3 days)
T2 missiles 0/20 success cca 48% chance each
T2 medium rigs (with max +ME decr) 1/15 success cca 53% chance each
maybe is something wrong with CCP chance calculator after patch(or with me )
Ehh, I just did my first invention on a crucifier with a circular logic decrypter(less chance, more runs) with only 1 bpc and was succesfull. I guess its just random, like the lottery. Dont think anything of it just rejoice when you win. |
Roos Stormshadow |
Posted - 2009.12.11 20:41:00 - [ 32]
Originally by: Akita T
Edited by: Akita T on 06/12/2009 18:56:40
The EVE RNG at least SEEMS to be quite "streaky", even if it it might be not (it could still be "too streaky", but we don't really have a sure way to tell).
Considering the number of players doing things that require random numbers, I highly doubt that you've ever received even two consecutive draws from the server's RNG. Far more consecutive draws than two would be required to evaluate the quality of Eve's RNG. |
Qual
Gallente
Cornexant Research
|
Posted - 2009.12.11 20:52:00 - [ 33]
Re: The random number generator.
Guys, when you shoot a gun in eve, the result is a random ammount of dmg. Do you REALLY thing they use a different generator for these numbers? Seriously?
If thier random number generator was broken you would notice like a gazillion PvE and PvP player scream on the forums.
The random number generator is NOT broken. Give it a rest. |
Drevia |
Posted - 2009.12.12 00:26:00 - [ 34]
Edited by: Drevia on 12/12/2009 00:30:53
I've never taken a statistics course, but even I'm certain that most of the examples given here are wrong.
You run job #1, you have a 45% chance of success.
You run job #2, you have a 45% chance of success.
You run job #3, you have a 45% chance of success.
Each incident has a fixed chance. Statistics are not cumulative in a case like this. Just because your first job failed when it had a 45% chance of success doesn't mean that your second run suddenly has a 90% chance of success. It's still just 45%.
They could be if you were talking about pulling out lottery numbers from a single pool of balls. Numbered balls 1 through 50, your odds of pulling out 45 the first ball are 1 in 50. Your first pull is 29. Your source of numbers is reduced, so your odds of getting 45 on the next pull is 1 in 49.
Unfortunately streaks like this are just part of the game. I've done a batch of 100 invention runs and gotten zero successes out of it. I've done batches of 100 and gotten 70+ out of it. |
Akita T
Caldari Navy Volunteer Task Force
|
Posted - 2009.12.12 01:11:00 - [ 35]
Originally by: Drevia
I've never taken a statistics course, but even I'm certain that most of the examples given here are wrong.
Guess you should have probably at least read the basics first then ? Quote:
You run job #1, you have a 45% chance of success.
You run job #2, you have a 45% chance of success.
You run job #3, you have a 45% chance of success.
Each incident has a fixed chance. Statistics are not cumulative in a case like this. Just because your first job failed when it had a 45% chance of success doesn't mean that your second run suddenly has a 90% chance of success. It's still just 45%.
So far so good, correct. BUT... Say you make a total of, oh, 10 tries, at precisely 50% chance of success/failure. You know, something as simple as flipping a coin. There are a grand total of 2^10 = 1024 combinations of success/failure that could happen for that 10-trial run, and each and every one of those sequences of 10 tries has EXACTLY the same chance of happening, namely 1/1024. Now, just how many variations have all successes ? Just one out of 1024. The same as the variation with all failures, actually. So the chance of a 10/10 or 0/10 overall result is 1/1024, or 0.09765625% How about 1 success and 9 failures ? There's 10 of those, actually, same as 1 failure and 9 successes. The overall chance of any of those happening is therefore 10/1024, or 0.9765625% And so on and so forth. Now, mind you, the math does get a little more complicated whenever chances are not exactly 50%, but the math does exist, and is easily available to pretty much everybody who has Excel, for instance. In Excel, there's a function there called "binomdist" that does exactly that, it computes the chance of either exactly S successes or at most S successes out of T tries at chance C% of success (depending on the last argument of the function). Or to return to the simple version : if you flip a coin many times and it keeps showing almost always heads (or almost always tails), do you conclude that you're either very lucky (or very unlucky), or do you conclude that there's something wrong with the coin ? Think about it... Quote:
Unfortunately streaks like this are just part of the game. I've done a batch of 100 invention runs and gotten zero successes out of it. I've done batches of 100 and gotten 70+ out of it.
Without naming the expected success rate, those numbers are not very meaningful. Even if the invention chance would have been barely 20%, to get ZERO successes out of 100 tries, that's 1-in-almost-5-billion chances of it happening. If what you say is true, and your expected chance to invent wouldn't have been much lower, that's actually pretty good proof that the system _IS_ indeed broken. |
illusionary beauty
I-Tyranny
|
Posted - 2009.12.14 07:46:00 - [ 36]
Edited by: illusionary beauty on 14/12/2009 07:52:19 Originally by: Beldor
maybe is something wrong with CCP chance calculator after patch(or with me)
After the patch my invention success went down. Maybe i just hit a really really bad streak, but in a year i've never seen that poor a run over a couple days time. Originally by: Qual
Guys, when you shoot a gun in eve, the result is a random ammount of dmg. Do you REALLY thing they use a different generator for these numbers? Seriously?
I heard there was a bug with guns causing the first shot or last shot or something to miss everytime and that they changed it in Dominion. |
Robert Caldera |
Posted - 2009.12.14 12:18:00 - [ 37]
Edited by: Robert Caldera on 14/12/2009 12:18:18 Originally by: illusionary beauty
After the patch my invention success went down. Maybe i just hit a really really bad streak, but in a year i've never seen that poor a run over a couple days time.
confirming the same here. The highest was 5 of 10 post-patch, mostly only 2-4 inventions success. Invention seems got nerfed in Dominion. |
OnoSendai
Caldari
Intergalactic Absurdities Unlimited |
Posted - 2009.12.17 03:13:00 - [ 38]
I would have to agree. It seems as if everything is using the ship success rates of 20% rather than modules and such at 40%. Petition time. At least they can verify the coding is correct. |
CSI Eve
Caldari
atomic comic
|
Posted - 2009.12.17 03:55:00 - [ 39]
invention has been buffed actually. because post-patch i have a success rate of 83% where i should have 49.92%.
come one whiners, use your brains one time |
Zifrian
Deep Space Innovations
|
Posted - 2009.12.17 07:51:00 - [ 40]
Man, so many people don't understand simple statistics.
There is nothing wrong with invention. |
Robert Caldera |
Posted - 2009.12.17 15:04:00 - [ 41]
still getting bad invention rates, mostly 4 of 10, hardly over 5. This has been not that poor before the patch. I understand statistics and I think I'm ok saying invention chance went down after hundred inventions hardly coming over 5/10. The mean value of 1/2 means there should be as many under 5 as over 5. So I'm getting hardly over 5 and the most under 5 (3-4).
This sucks, invention is broken. |
CSI Eve
Caldari
atomic comic
|
Posted - 2009.12.17 16:46:00 - [ 42]
stop invention please |
Robert Caldera |
Posted - 2009.12.17 17:05:00 - [ 43]
Originally by: CSI Eve
stop invention please
stop posting please |
Joel McBeth
Caldari
Involuntary Sub-Saharan Crowdsourcing |
Posted - 2009.12.17 18:19:00 - [ 44]
http://en.wikipedia.org/wiki/Binomial_distribution
http://en.wikipedia.org/wiki/Law_of_large_numbers
P(1, 10, .45) ~= 0.0207
That's a 2% chance of only getting 1 success out of 10 trials.
The important thing to take from this is it is not a 0% chance. |
Forduc |
Posted - 2009.12.17 18:42:00 - [ 45]
You'll also have to remember that if you do more batches of 10 inventions, number of successes from one batch will always be meaningless (in statistical sense). My statistics is little rusty, but I think that you can't get meaningfull(significant) results from just 10 inventions, atleast if success chance is near 50%. |
CSI Eve
Caldari
atomic comic
|
Posted - 2009.12.18 03:50:00 - [ 46]
invent brains, please |
Robert Caldera |
Posted - 2009.12.18 12:07:00 - [ 47]
dudes, dont teach me maths, had enough of them in the university.
I'm doing hundreds of inventions a week and I would rather assume a nerf than a really really really fukkin bad luck streak getting only 3-4 inventions out of 10 all the time. |
Xious
Caldari
Sarz'na Khumatari
Ushra'Khan |
Posted - 2009.12.19 00:51:00 - [ 48]
Curious, why are computer made rng's not 100% random?
Not questioning the statement, just wondering why. |
Bia Bri |
Posted - 2009.12.19 08:09:00 - [ 49]
Originally by: Xious
Curious, why are computer made rng's not 100% random?
Not questioning the statement, just wondering why.
computers are only capable of following algorithms with predefined initial conditions often called the 'seed.' If you start a pseudo-rng with a particular seed it will always produce the same sequence of 'random' numbers. In practice this is of little consequence. Seeds are often taken by noting the internal clock time (down to the millisecond or finer) of when the computer was turned on or the first request for a pseudo-rng was requested and gives a new seed each time. Any decent algorithm gives a sequence of numbers that is essentially indistinguishable from a 'true' random sequence. (it would requires billions of billions of numbers to be able to detect differences- far more than an average consumer will ever even approach). |
Lord Fitz
Project Amargosa |
Posted - 2009.12.19 10:21:00 - [ 50]
Originally by: Bia Bri
Originally by: Xious
Curious, why are computer made rng's not 100% random?
Not questioning the statement, just wondering why.
computers are only capable of following algorithms with predefined initial conditions often called the 'seed.' If you start a pseudo-rng with a particular seed it will always produce the same sequence of 'random' numbers.
In practice this is of little consequence. Seeds are often taken by noting the internal clock time (down to the millisecond or finer) of when the computer was turned on or the first request for a pseudo-rng was requested and gives a new seed each time.
Any decent algorithm gives a sequence of numbers that is essentially indistinguishable from a 'true' random sequence. (it would requires billions of billions of numbers to be able to detect differences- far more than an average consumer will ever even approach).
Not to mention in order to see this not 100% random you have to be the only person getting any numbers generated, in a game like eve where there are hundreds of requests per second to any nodes RNG, the randomness is actually guaranteed by humans. Also because there's a large number of people doing large number of invention runs, you're guaranteed to see 'fairly' rare outcomes very often. |
fuer0n |
Posted - 2009.12.19 12:59:00 - [ 51]
Originally by: Akita T
Edited by: Akita T on 06/12/2009 19:11:11
Originally by: Lord Helghast
wow so basically your lucky as hell to even get 1 success but if u do u might get a few in a row.
Actually, more like 99.75% chance to get AT LEAST one success.
0 success out of 10 tries with 45% success chance is 0.25% (or 1 out of 395 )
1 success out of 10 tries with 45% success chance is 2.07% (or 1 out of 48 )
2 success out of 10 tries with 45% success chance is 7.63% (or 1 out of 13 )
3 success out of 10 tries with 45% success chance is 16.65% (or 1 out of 6 )
4 success out of 10 tries with 45% success chance is 23.84% (or 1 out of 4 )
5 success out of 10 tries with 45% success chance is 23.40% (or 1 out of 4 )
6 success out of 10 tries with 45% success chance is 15.96% (or 1 out of 6 )
7 success out of 10 tries with 45% success chance is 7.46% (or 1 out of 13 )
8 success out of 10 tries with 45% success chance is 2.29% (or 1 out of 44 )
9 success out of 10 tries with 45% success chance is 0.42% (or 1 out of 240 )
10 success out of 10 tries with 45% success chance is 0.03% (or 1 out of 2,937 )
P.S. This is not "EVE math", it's "real-world math".
P.P.S. Notice how the total chance to get either 3,4,5 or 6 successes out of 10 tries is almost 80%.
Still, roughly 1 in 5 people doing a 10-slot run will get (2 or less) or (7 or more) even under completely normal circumstances.
It's not a sham, it's just basic statistics.
by your maths if you play roulete (red or black) and start betting with 1x currency and tripple the bet every time you lose you're on a sure winner. the fact is you have a 50/50 chance at the start of every bet and no ammount of math you calculate from the previous bets makes any difference at all. not factoring in the zero. |
Aargh |
Posted - 2009.12.19 16:05:00 - [ 52]
Originally by: fuer0n
Originally by: Akita T
Edited by: Akita T on 06/12/2009 19:11:11
Originally by: Lord Helghast
wow so basically your lucky as hell to even get 1 success but if u do u might get a few in a row.
Actually, more like 99.75% chance to get AT LEAST one success.
0 success out of 10 tries with 45% success chance is 0.25% (or 1 out of 395 )
1 success out of 10 tries with 45% success chance is 2.07% (or 1 out of 48 )
2 success out of 10 tries with 45% success chance is 7.63% (or 1 out of 13 )
3 success out of 10 tries with 45% success chance is 16.65% (or 1 out of 6 )
4 success out of 10 tries with 45% success chance is 23.84% (or 1 out of 4 )
5 success out of 10 tries with 45% success chance is 23.40% (or 1 out of 4 )
6 success out of 10 tries with 45% success chance is 15.96% (or 1 out of 6 )
7 success out of 10 tries with 45% success chance is 7.46% (or 1 out of 13 )
8 success out of 10 tries with 45% success chance is 2.29% (or 1 out of 44 )
9 success out of 10 tries with 45% success chance is 0.42% (or 1 out of 240 )
10 success out of 10 tries with 45% success chance is 0.03% (or 1 out of 2,937 )
P.S. This is not "EVE math", it's "real-world math".
P.P.S. Notice how the total chance to get either 3,4,5 or 6 successes out of 10 tries is almost 80%.
Still, roughly 1 in 5 people doing a 10-slot run will get (2 or less) or (7 or more) even under completely normal circumstances.
It's not a sham, it's just basic statistics.
by your maths if you play roulete (red or black) and start betting with 1x currency and tripple the bet every time you lose you're on a sure winner. the fact is you have a 50/50 chance at the start of every bet and no ammount of math you calculate from the previous bets makes any difference at all. not factoring in the zero.
"not factoring in the zero". Or indeed house limits. LOL. |
kit amor |
Posted - 2009.12.20 02:27:00 - [ 53]
Originally by: FroschForscher
come on. there´s nothing worng with the rng. its just that only the bad streak people post on the forums post patch
Ok I'm quite new to this am inventing ship hulls rather than mods. I've maxed my datacore skills and the encryption is at 4. I don't "batch" invention - so far 1 Bpc to try and make a 1- or 3- run t2 bpc of a specific ship type. I DO always use decryptors. So far, from 5 attempts, I've had 4 successes, producing an Ishtar bpc, 3-run Helios, 3-run Ishkur and a Kronos bpc. (The Kronos is the one that failed once.) Maybe it's a streak and I'm about to have lots of failures but so far, touch wood, all is very well.  So the patch is fine... |
kingjames488 |
Posted - 2009.12.20 03:22:00 - [ 54]
Originally by: mechtech
I'm just going to put this out there... The eve developers are incredibly talented, having put together arguably the most complex game ever made.
To imply that they couldn't put together a random number generator (there are thousands of public algorithms for efficient, solid, RNGs out there as well) is a very unbelievable claim to make.
The RNG works, and you got unlucky. Just to confirm people's doubts though, it would be nice to have a line in some patch note that states that a dev went over the RNG and found it to work as intended.
eve its self actually seems uite simple to me... a large area where people interact as they choose. but thats my opinion... I've just gotten into invention and so far had a decent rate of sucsess. so ive done prolly 10 inventions total and only a couple have failed, tho i know thats not much at all, goes to show how luck can play out when working with small numbers. |
Svedge Greywolf |
Posted - 2009.12.20 11:32:00 - [ 55]
I'm having invention problems too, getting a success rate of around 25-33% on modules after Dominion, far from 47% which is should be. The success rate was quite stable before Dominion..
|
Lord Fitz
Project Amargosa |
Posted - 2009.12.20 12:08:00 - [ 56]
Originally by: Svedge Greywolf
I'm having invention problems too, getting a success rate of around 25-33% on modules after Dominion, far from 47% which is should be. The success rate was quite stable before Dominion..
It should be 20% for some people and 70% for others. You're just one of the people that you should be getting lower. Congratulations, you've discovered why some people lose all their money and others get rich when gambling. Any statistical accuracy needs a far larger sample. |
Mikelangelo
Gallente |
Posted - 2009.12.21 20:28:00 - [ 57]
I'm not really surprised that happened to you.
It will average out though. I had 4 cruiser inventions fail, and then 1/4 frigages, then 4 battleships at the same time succeed, with no decryptors.
Then after that 0/4 on ammo. Totally weird, but it eventually averages out given enough sample size. |
Joel McBeth
Caldari
Involuntary Sub-Saharan Crowdsourcing |
Posted - 2009.12.21 22:15:00 - [ 58]
Originally by: fuer0n
Originally by: Akita T
Edited by: Akita T on 06/12/2009 19:11:11
Originally by: Lord Helghast
wow so basically your lucky as hell to even get 1 success but if u do u might get a few in a row.
Actually, more like 99.75% chance to get AT LEAST one success.
0 success out of 10 tries with 45% success chance is 0.25% (or 1 out of 395 )
1 success out of 10 tries with 45% success chance is 2.07% (or 1 out of 48 )
2 success out of 10 tries with 45% success chance is 7.63% (or 1 out of 13 )
3 success out of 10 tries with 45% success chance is 16.65% (or 1 out of 6 )
4 success out of 10 tries with 45% success chance is 23.84% (or 1 out of 4 )
5 success out of 10 tries with 45% success chance is 23.40% (or 1 out of 4 )
6 success out of 10 tries with 45% success chance is 15.96% (or 1 out of 6 )
7 success out of 10 tries with 45% success chance is 7.46% (or 1 out of 13 )
8 success out of 10 tries with 45% success chance is 2.29% (or 1 out of 44 )
9 success out of 10 tries with 45% success chance is 0.42% (or 1 out of 240 )
10 success out of 10 tries with 45% success chance is 0.03% (or 1 out of 2,937 )
P.S. This is not "EVE math", it's "real-world math".
P.P.S. Notice how the total chance to get either 3,4,5 or 6 successes out of 10 tries is almost 80%.
Still, roughly 1 in 5 people doing a 10-slot run will get (2 or less) or (7 or more) even under completely normal circumstances.
It's not a sham, it's just basic statistics.
by your maths if you play roulete (red or black) and start betting with 1x currency and tripple the bet every time you lose you're on a sure winner. the fact is you have a 50/50 chance at the start of every bet and no ammount of math you calculate from the previous bets makes any difference at all. not factoring in the zero.
His math is correct and considers that each invention is an independent trial. Your strategy doesn't work because you don't have an infinite amount of money. Look up the bionomial distribution and the law of large numbers, as i've posted above. |
Luaren Avidius |
Posted - 2009.12.22 01:29:00 - [ 59]
The main thing that needs to be said with stuff that involves chance of success and chance of failure is that you need an increasingly large number of samples to get close to what you 'should' be getting.
The general rule with stuff like this is that the observed chance of success will approach the statistical chance of success as the number of events grows. That means that with 10 tries at something, your actual chance of success will be all over the place. With 100 its going to start to settle down a bit, but still wont be quite where it 'should' be. With 1,000 tries it 'should' be almost there, and as you pass 10,000 tries, 100,000 tries, and 1,000,000 tries, your actual chance should get closer and closer to the statistical chance.
As for random number generators...eve seems to have a pretty decent one, as no one has found a way to exploit or be exploited by lucky streaks. Other games have their own random number generators, which may or may not be so good. The original diablos random number generator was based, I think, on what time the users clock read when the game started. This was exploitable, as people could find out which seeds caused certain items to appear in the game that would be generated, and they could thus make sure that they used one of those seeds when starting a game.
As for the OP's run of 10 straight failures with a 45% chance...well, you reverse-won a lottery with those numbers. That is, the random number generator handed you an extremely rare complete failure. Lucky you!
|
Akita T
Caldari Navy Volunteer Task Force
|
Posted - 2009.12.22 01:35:00 - [ 60]
Originally by: fuer0n
by your maths if you play roulete (red or black) and start betting with 1x currency and tripple the bet every time you lose you're on a sure winner. the fact is you have a 50/50 chance at the start of every bet and no ammount of math you calculate from the previous bets makes any difference at all. not factoring in the zero.
The zeroes and the table limits exist for the precise purpose of foiling that strategy, which could otherwise work if you had a sufficiently large purse. |
All Channels
Science and Industry
