|
Author |
Thread Statistics | Show CCP posts - 2 post(s) |
Trak Cranker
Serenity Inc
|
Posted - 2007.06.27 14:54:00 -
[1]
The number of tries don't quite sound large enough to support any conclusion - but something might be up, ofc.
Mary M and others - can't you post your exact number of tries and relevant modifiers?
5 straight fails on manticores does not seem extreemly strange, considering the %'s I have seen on ship invention. Bad yes, but nowhere near impossible. But its still guesswork ofc.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.06.28 12:42:00 -
[2]
Originally by: Zasra
The t2 market was finally stablizing, my guess is bob wasn't making enough money from their bpos so they nuked the success rate. Just another reason to cancel my account I suppose.
Can I have your invention stuff?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.06.28 14:02:00 -
[3]
Originally by: Zasra
Originally by: Trak Cranker
Originally by: Zasra
...guess I'll put off experimenting with invention till some official reply comes.
Can I have your invention stuff?
I guess that part of my post was hard to understand for you? Moron.
I understood that part just as well as I understood the childish reference to a BoB preference in CCP and the even more childish "threat" to cancel your account over it. I take it the last word is your signature?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.06.28 18:18:00 -
[4]
Counterparty, you are officially the mascot of inventors now.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.06.28 18:53:00 -
[5]
Edited by: Trak Cranker on 28/06/2007 18:52:59
Originally by: CCP Chronotis Nothing has changed formula wise and if it did we would tell you.
Very good to know, Chronotis. That is a nice assurance really, and what I have missed most lately.
Any chance we could get a sort of hint in the skill descriptions on chance improvement pr level, as we do in basically all other areas?
Not to make the incention box totally transparent, but to at least make a little bit less gray?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.06.30 18:01:00 -
[6]
Originally by: Ricdic Since discovering a hidden secret last week..
Spill the beans, man. :) Or a hint at least.. :)
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.02 20:41:00 -
[7]
Originally by: Kaaii
Life is strange...
8 x ferox R&D shows 2 vults, 3 nighthawks
Failed both vults, 1 nighthawk..
And got 2 5 run nighthawks using install guide, max skills, and max run bpc...
Kaaii
Whats strange about that?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.03 12:09:00 -
[8]
Edited by: Trak Cranker on 03/07/2007 12:11:27
Originally by: Kaaii
Originally by: Magunus
Originally by: Trak Cranker
Originally by: Kaaii
Life is strange... 8 x ferox R&D shows 2 vults, 3 nighthawks
Failed both vults, 1 nighthawk..
And got 2 5 run nighthawks using install guide, max skills, and max run bpc...
Kaaii
Whats strange about that?
Obviously the 3 missing jobs! 2+3 is 5!!! not 8! ... I think. Right?
Yes, theres 3 more running, finish some time tomorrow, but I had to jump back, big op planned taking M&M's region.....oh, wait!
So I maintain my question. Whats strange about it?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.03 14:30:00 -
[9]
Originally by: TGR Reaper I am doing this propery with max skills and having poor luck since the patch and other are having the same problems. put 2 and 2 togather and hmm something could be wrong?
Part of doing invention properly is knowing the odds and having the patience and wallet to wait out anything but the worst fluctuations in luck.
Did you calculate beforehand what the odds of a wallet breaking bad streak could be?
There is more to this than just having well trained skills.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.03 17:27:00 -
[10]
Edited by: Trak Cranker on 03/07/2007 17:26:03
Originally by: Silent Sam Thoughts?
1: I dont think that holds true. I think you are overanalysing on a limited statistical material.
2: It would require for CCP to hold some kind of state for your invention or invention as a whole. Which would be very unlike anything else.
3: Chronotis, the dev, have said that they would tell us if they change the formula and he had checked back for some complainers, and they had had the chance they should have - implying that the formula is fixed.
|
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.05 18:50:00 -
[11]
Edited by: Trak Cranker on 05/07/2007 18:49:30
Originally by: BillCosby
I HAVE TONS OF STATISTICAL DATA HUNDREDS OF INVENTION OPERATIONS AT A STEADY 40% RATE NOW HAVE COMPLETELY FALLEN OFF THE MAP WITH THE SAME EXACT SKILL SET. I GOT HIT WTH THE MEGA NERF BAT.
In your case I think it can be explained by you using up all your invention power to hold down the caps lock key.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.07 04:11:00 -
[12]
Well in order for us to calculate the chance of this happening, we need to know the actual chance of the invention in question. Which we dont.
But lets say its 50%. Then the odds are 1/10000. Which bad, but not impossible. It will prob. happen to someone over time - and that could be you. :)
If its 30%m then it becomes close to 1%.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.07 04:16:00 -
[13]
Originally by: Hemma Roids
And to the GM who so kindly graced us with his presence my personal opinion is ya a lying sack of S**T, easy to say oh theres nothing wrong and offer zero zip and nada as proof, you want us to have some faith in the devs and gm's, (notice lower caps, they don't honestly deserve better) then offer up a formuli, we(the inventors) deserve to know what the hell we are pouring our cash into
Lying sack of <something> is quite the accusation. And you do know that you have the option of not pouring your cash into something you know nothing about? I like think inventors are the kind of people that go about this in a scientific way. And dont just pour stuff into a black box machine and then cry foul when they do not like the result.
There is plenty of options of doing ground work on this, before commiting any isk.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.07 11:40:00 -
[14]
Originally by: Death Ryder matey, wake up a smell the devs but ya smooching... its your line of thinking that allows the devs to keep 1 and all in the dark as to the formuli behind invention.
The only thing stopping you from ascertaining the input/success relations is a lack of effort. I like that they don't come out with the formula. It punishes people that treat this like the T1 market and the ones that do no research.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.07 13:16:00 -
[15]
Originally by: Sokudo
SOMETHING has changed - with maxed or near-max skills, we should be finding failures to be rare.
Says who?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.08 00:02:00 -
[16]
Originally by: Dal Thrax
Randomizer has to be borked.
Dal
Or you have to realise that the ones most prone to post on the forums, are the ones that are experiencing the more noticable effects of it having randomized results.
People having average run-of-the-mill relatively satisfying results are not going to come running to the forums.
Its like saying that the lotto drawings must be borked, because you only see winners in the papers.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.08 13:43:00 -
[17]
Originally by: Ricdic
Realistically there is no number of tries that can base a real argument on. If I done 50,000 attempts today and got a 50% success rate, I could try 50,000 attempts tomorrow and have a 10% success rate. Without knowing the formula's all we can do is show different environments and try to theorise as to possible reasoning, forumale and solutions.
Statistically you can say that with 50000 tries at 50% succes, there is 99,999% certainty that the number is within 0,9877159% points of the right one. Thats prob good enough to base an argument on.
At 50 tries that certainty gives a spread of 31% points.
So yeah, statistically some number of tries are better to base an argument on than others. 50 is def. not it, though.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.08 23:44:00 -
[18]
Originally by: TGR Reaper
Originally by: Tunak
Originally by: TGR Reaper interesting no ships 1 run and you are having way better success than i am. thats kinda weird that i put everything into it and get nothing even with maxed skills!
In no way is this interesting. This is perfectly normal. Please learn math.
Yeah I just finished college math statisitics and algerba...
If you did, you should know that the numbers you have posted in no way can support any kind of conclusion in terms of a change in probability.
Tunak; I would probably use the word "possible" instead of "normal", which could indicate something being the most common. But in essence you are right. (But in The Forge you might not be. )
Anyways, I would love to hear if someone is testing the corp run invention tries. But if that was a 100% fail, I think we would have heard from a bit more people. And I find it hard to believe the the fact that they are run via corp installations would make the chance calculation change. Not impossible, but hard to believe.
Its obvious that those REALLY unlucky lately will come here and post. And that those that are REALLY lucky lately will be less inclined to do so. So the combined numbers posted here will be heavily skewed towards things being worse than before.
And even so, the numbers are not plentiful enough to support any kind of usable conclusion.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.09 08:54:00 -
[19]
Originally by: Mnengli Noiliffe
however when people do hundreds of jobs and their rates fall, this is no longer probability, it is STATISTICS. which definitely means that chances have changed.
Maybe you need to brush up on your statistics class. You fail statistics if you use words as "definitely".
All you can do with 100's of jobs (or any number of jobs really) is to say that something MIGHT have changed, with X percentage confidence. And given that the most commonly used 95% confidence gives a spread of 5% at a sample size of 1000, I would say you need be very careful about making any conclusions after 100's.
And remember, most importantly, we are not seeing random sampling in this thread. There is good cause to believe that we are seeing mostly reports from people that have had bad runs.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.09 15:13:00 -
[20]
Edited by: Trak Cranker on 09/07/2007 15:14:24
|
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.09 15:15:00 -
[21]
Edited by: Trak Cranker on 09/07/2007 15:15:16
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.10 00:22:00 -
[22]
Originally by: Lab Technician071548
That conclusion is of the form, "Either this result, based on samples, is true (or false, depending on test result) of the population or this sample is unusual."
You are not allowed any conclusion unless it is accompanied by a confidence factor and margin of error. And even given those, you cannot apply anything as final as a true or false value.
Originally by: Lab Technician071548
For example, if you take 1,000 samples from a population in the millions or billions, and each sample is composed of 100 units, some percent of those samples will give a false positive. That is, some of those samples will lead you to believe that the question you are testing is resolved by the statistical test that simply is not true.
I would say that most of those samples would not be _true_, in terms of the question you are asking. If the population mean is that 75,6 of them like chocolate, no sample of a 100 people will be true to that fact.
Talking of false positives that way in terms of statistical sampling is a slightly horrible misrepresentation of the matter. As almost all samples are ultimately false positives.
For a sample to be a false positive in the statistical sense, it would have to be assumed positive at some point. And to do that you need to introduce the confidence level, because statistics is not in the business of ruling anything out.
Originally by: Lab Technician071548
The guy next to you could pull out 10 white balls in a row, a much less probable event, and conclude that all the balls are white or, within a certain probability boundary, his sample was unusual.
You should apply that term a little more. Because thats front and center of the matter.
Originally by: Lab Technician071548
We have a lot of people pulling balls out of bags, but they don't know how many are black and how many are white. They also don't know how many balls are in the bag. In some cases, such as command ship inventing, some balls are black, some are green, and some are blue because there are two positive outcomes.
That is only true in terms of profit calculation. In terms of succes determination there are only black and white balls in the bag.
Originally by: Lab Technician071548
Some people say you need thousands of samples to get an idea.
No, they say you need thousands to have a certain degree of confidence in your numbers.
Originally by: Lab Technician071548
...the statistical tools available to you.
If you by statistical tools, mean more data, then yes, of course.
Originally by: Lab Technician071548
You can make A statistical analysis with 10 data points but you can often get quite good results with only a hundred. Your comfort level with sample size will depend on your training. My training is multidisciplinary. The economist in me says I want at least 10,000 data points. The sociologist in me is comfortable with using categorical data. The psychologist in me says that's all crap and we can do a nice little analysis with 20 data points even though more are desirable. A lot depends on what you want to get out of your data.
That there is pure crap in terms of statistiscs. Numbers are numbers.
Whether you feel confident in making assumptions based on them has NOTHING to do with statistics, but purely to do with how big a chance you are willling to take on it being right. Or how dumb or cynical you are.
Hence the the guy with the money wants many data points, the guy with low skills, but a lot to say, will work with less, and the astrologist will work with close to nothing.
Who do you want to go with? The one that will be held accountable or the one just letting out hot air just to say something?
For a guy actually teaching this stuff, that was sure a lot of empty words, with almost catastrophical misdirections.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.10 23:08:00 -
[23]
Edited by: Trak Cranker on 10/07/2007 23:09:34
Originally by: Lab Technician071548
A confidence interval (based on a probability, not a "factor") tells you how "good" your "answer" is or, in other words, how "close" "to" "the" "truth" "it" "is", with the truth often being unknowable in absolute terms, which is why we resort to statistics in the first place.
The confidence interval does not tell us how close to the truth our result is.
It tells us how big a _chance_ there is for our result to be _within a certain distance_ of the truth. But we can never be sure that our result does not represent the remaining chance of it falling outside that distance. Even way outside that distance.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.11 11:32:00 -
[24]
Originally by: Tunak
Oh and to get back on subject with the invention stuff. People are now looking at corp/personal/POS differences. Did any of these things factor into success before the patch? For these to matter CCP would have to intentionally code it in. Even if they did code it in they would have had to change something which they've already said they haven't.
There _could_ be some general code, that would miss any decryptors and such attached to a POS job or the likes. Stuff that is assumed to work, but isn't on POS's and corp jobs. But there is little reason to think that changed in Rev 2 - and there are enough people working in corp labs and corp POS labs - that I think we would have far more complainers, if the chance had decreased like some here would like us to believe.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.08.06 22:47:00 -
[25]
Originally by: Servus Roscoe its easy
just run a t-test or some other simple stats math on the data here and we can say that for some unlucky souls there has been a significant (p < 0.01) change. Plenty of the examples here were significant at <0.05 which is a scientific standard.
I am struggling to find sense in the above. "Plenty of the examples here were significant at <0.05" - What on earth do you mean by that?
I don't think I have seen any examples here that would support any conclusion whatsoever. Remember we are not seeing a random sampling of the total distribution in this thread or threads like it. So you can't run any meaningful tests on it.
What we are seeing is the natives blaming anything that is not as ordinary as they themselves are, when things go against them. Mental savages that cannot abide the unfortunate turns of events that has befallen them and therefore looks for unfamiliarity/malevolence/incompetence in something thats is but the order and nature of the life they have undertaken.
|
|
|
|