Pages: 1 [2] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |

Lexmana
Fweddit I Whip My Slaves Back and Forth
592
|
Posted - 2012.06.25 08:17:00 -
[31] - Quote
Dinsdale Pirannha wrote:To the OP: Check in the wormhole forum online. See if anyone else is suddenly having issues with T3 reverse engineering.
But if you were 23/60, and you think you have a 73% expected success rate, I would say there is something definitely odd happening. Don't listen to the dullards. A sample size of 60 IS NOT considered trivial. And if you have results for 120, please post them. I would be very interested in seeing them.
Wow. And you are some kind of professional statistician? You do realize that for most players in game right now everything is working as expected (otherwise they would all whine on the forums). What you see in this thread is the result of publication bias. So out of thousands of players only one is seeing something out of the ordinary on a sample of 60. This is very likely completely within the parameters. |

RubyPorto
SniggWaffe YOUR VOTES DON'T COUNT
2056
|
Posted - 2012.06.25 08:33:00 -
[32] - Quote
Victoria Sin wrote:RubyPorto wrote:Liza Hawkeye wrote:I understand your position, but I'm not bad in statistics (I'm and economist) and I totally see that considering the number of jobs I'm doing - the average chance of getting BPC became much lower. And the chance of getting a good BPC is awfully low now. Thats made me to post that thread to get an answer - did CCP change smth with reverse engineering chance or only I'm suffering from my bad luck or smth. If your a statistician, tell me, would you assume the coin is unfair if you see a run of 40 heads when you flip a coin 400 times? That would be unusual. It's (1/n)^x, where `n' is 2 (2 possible outcomes) and x is the number of coin tosses. For 10 in a row, you get 0.5^10 = 0.0009765625, which I believe is 1 in 1024 sets of 10 coin tosses.
Thanks.
The law of large numbers says that unusual things happen constantly.
So it's been a loong time since I took statistics, so my practical skills are pretty gone, how would you try to figure out how likely the outcome, 23 successes in 60 tries with an 80% chance of success, is? This is EVE. -á Everybody Versus Everybody. |

Lexmana
Fweddit I Whip My Slaves Back and Forth
592
|
Posted - 2012.06.25 09:02:00 -
[33] - Quote
RubyPorto wrote:So it's been a loong time since I took statistics, so my practical skills are pretty gone, how would you try to figure out how likely the outcome, 23 successes in 60 tries with an 80% chance of success, is? With those parameters it is acatually a very low probability for observing Gëñ23 successes (p=2.23E-12). With 50% chance of success the result would be barely significant (p=.046). You can try it yourself with this binomial calculator.
RubyPorto wrote:The law of large numbers says that unusual things happen constantly. That is a likely explanation. I am sure OP did not choose a random sequence but instead took one with the most extreme outcome out of quite a few possible other sequences with less extreme outcomes. We need to account for that and for all the other players that don't experience the same. This would increase the probability of this observation dramatically.
OP may have struck some very bad luck but the next batch of 60 is very likely to be better . |

Crellion
Parental Control
27
|
Posted - 2012.06.25 09:07:00 -
[34] - Quote
Liza Hawkeye wrote:RubyPorto wrote:Liza Hawkeye wrote:I understand your position, but I'm not bad in statistics (I'm and economist) and I totally see that considering the number of jobs I'm doing - the average chance of getting BPC became much lower. And the chance of getting a good BPC is awfully low now. Thats made me to post that thread to get an answer - did CCP change smth with reverse engineering chance or only I'm suffering from my bad luck or smth. If your a statistician, tell me, would you assume the coin is unfair if you see a run of 40 heads when you flip a coin 400 times? Considering that coin can give 3 results: top, bottom and edge, and we don't count physics in it - then the chance of each result is: 100%/3=33.3%. Everybody agree with it. Then if we flip the coin for 400 times - we're waiting for the result of 133.3 tops, 133.3 bottoms and 133.3 edges. If in our experiment results are 40 tops, 200 bottoms and 60 edges then the only thing we got from the result is that we did not enough tries. I understand your position, but I'm still sure (even if it's against logic and is only empirically thoughts) that chances were reduced. Currently I'm having material for 120 more jobs - when I'll finish them (tommorow I think) - I'll present you results.
So scientific knowlledge leads everybody to agree that if you toss a coin a few million times 1/3 of those times it will land on its edge? Amusing 
|

RubyPorto
SniggWaffe YOUR VOTES DON'T COUNT
2056
|
Posted - 2012.06.25 09:10:00 -
[35] - Quote
Lexmana wrote:RubyPorto wrote:So it's been a loong time since I took statistics, so my practical skills are pretty gone, how would you try to figure out how likely the outcome, 23 successes in 60 tries with an 80% chance of success, is? With those parameters it is acatually a very low probability for observing Gëñ23 successes (p=2.23E-12). With 50% chance of success the result would be barely significant (p=.046). You can try it yourself with this binomial calculator. RubyPorto wrote:The law of large numbers says that unusual things happen constantly. That is a likely explanation. I am sure OP did not choose a random sequence but instead took one with the most extreme outcome out of quite a few possible other sequences with less extreme outcomes. We need to account for that and for all the other players that don't experience the same. This would increase the probability of this observation dramatically. OP may have struck some very bad luck but the next batch of 60 is very likely to be better . If OP observes a similar outcome in the next batch of 60 I will reconsider my position or the parameters were wrong to begin with.
That is an awesome tool.
One in 500 billion chance of that result (or worse) with the the 80% parameter. That seems... let's say odd.
Are we sure that the OP's success chance calculations are right? This is EVE. -á Everybody Versus Everybody. |

Lexmana
Fweddit I Whip My Slaves Back and Forth
592
|
Posted - 2012.06.25 09:32:00 -
[36] - Quote
RubyPorto wrote: That is an awesome tool.
One in 500 billion chance of that result (or worse) with the the 80% parameter. That seems... let's say odd.
Are we sure that the OP's success chance calculations are right?
Yeah, I am starting to think there is something wrong with the premisses for this calculation. We will know if OP can replicate similar findings in the next 60 trials.
I would like to think that if OP is correct (something changed) we would have many more reprots with similar findings.
The most severe mistake that can be made in these calculations is not including all data. Maybe OP "forgot" to include some jobs that didn't fit the pattern? Also, selecting a sequence might be a problem. If someone runs 1000 of these jobs and want to report the most extreme sequence 60 observations long there are about 940 possible sequences to choose from. If we include a range of sequence length (30-80) it would multiply about 50 times.
|

RubyPorto
SniggWaffe YOUR VOTES DON'T COUNT
2057
|
Posted - 2012.06.25 09:59:00 -
[37] - Quote
Lexmana wrote:RubyPorto wrote: That is an awesome tool.
One in 500 billion chance of that result (or worse) with the the 80% parameter. That seems... let's say odd.
Are we sure that the OP's success chance calculations are right?
Yeah, I am starting to think there is something wrong with the premisses for this calculation. We will know if OP can replicate similar findings in the next 60 trials. I would like to think that if OP is correct (something changed) we would have many more reprots with similar findings. The most severe mistake that can be made in these calculations is not including all data. Maybe OP "forgot" to include some jobs that didn't fit the pattern? Also, selecting a sequence might be a problem. If someone runs 1000 of these jobs and want to report the most extreme sequence 60 observations long there are about 940 possible sequences to choose from. If we include a range of sequence length (30-80) it would multiply about 50 times.
Still, so long as it's a sequence of 60 in a row, only one in 500 billion sequences of 60 will have 23 or fewer successes at an 80% chance of success. Even if there's 50,000 such sequences in 1000 actual runs, there's only a one in Ten Million chance that such a sequence exists in the 1000 run data set.
@OP, are you sure you're using the intact relics every time? This is EVE. -á Everybody Versus Everybody. |

Lexmana
Fweddit I Whip My Slaves Back and Forth
592
|
Posted - 2012.06.25 11:10:00 -
[38] - Quote
RubyPorto wrote: Still, so long as it's a sequence of 60 in a row, only one in 500 billion sequences of 60 will have 23 or fewer successes at an 80% chance of success. Even if there's 50,000 such sequences in 1000 actual runs, there's only a one in Ten Million chance that such a sequence exists in the 1000 run data set.
@OP, are you sure you're using the intact relics every time?
I do agree that the probability is still very low. We have to exclude the possibility of using the wrong relics. With 60% chance of success p=0.000569 it is a very likely outcome even if we assume a very low selection bias. With 50k possible datasets p=1. With 1000 datasets p= 1-(1-0.000567)^1000=.43.
However, observing very small probabilities does happen in a game like EVE. If we assume that such event will always be posted on the forums. If we assume that 50k players are producing dataets with 50k possible sequences 10 times a year we would get a probability of OPs observation of of p=1-(1-2.23E-12)^(50000*50000*10)=.10. About once every decade .
|

RubyPorto
SniggWaffe YOUR VOTES DON'T COUNT
2060
|
Posted - 2012.06.25 11:15:00 -
[39] - Quote
Lexmana wrote:RubyPorto wrote: Still, so long as it's a sequence of 60 in a row, only one in 500 billion sequences of 60 will have 23 or fewer successes at an 80% chance of success. Even if there's 50,000 such sequences in 1000 actual runs, there's only a one in Ten Million chance that such a sequence exists in the 1000 run data set.
@OP, are you sure you're using the intact relics every time?
I do agree that the probability is still very low. We have to exclude the possibility of using the wrong relics. With 60% chance of success p=0.000569 it is a very likely outcome even if we assume a very low selection bias. With 50k possible datasets p=1. With 1000 datasets p= 1-(1-0.000567)^1000=.43. However, observing very small probabilities does happen in a game like EVE. If we assume that such event will always be posted on the forums. If we assume that 50k players are producing dataets with 50k possible sequences 10 times a year we would get a probability of OPs observation of of p=1-(1-2.23E-12)^(50000*50000*10)=.10. About once every decade  .
If only WHs had been out that long.
I think that incorrect relics are the root of the problem. I doubt CCP would stealth nerf reverse engineering out of the blue. This is EVE. -á Everybody Versus Everybody. |

brinelan
RSK. Curbstomp..
5
|
Posted - 2012.06.25 11:21:00 -
[40] - Quote
Liza Hawkeye wrote:
8/23 are not the BPC for 1 exact subsystem but for 2 subsystems. So it's 3+5. and 15 BPC for other 2 subsystem that are unworthy of production.
I'll monitor my future reverses, but I'm not sure that results will be better...
You could always stop re'ing caldari... if you only have a 1/4 change that your successful bpc is worth money, is it even worth bothering with? |
|

Andy DelGardo
68
|
Posted - 2012.06.25 11:21:00 -
[41] - Quote
So we can verdict that Lisa is indeed CURSED! So my offer still holds and i will contact this blessing dude from a other thread to purge her curse, if u sell her to me.
bye Andy
PS: On a serious note, i just finished my batch last night and nothing odd at all. From 30 intact tries i got 5 accl. bay BPC, with a overall success rate of 25 out of 30 tries, so i'm even over the estimated 80%. So we should stop "wondering" about a odd numerical occurrence, if we cant reproduce the actual result. |

Graic Gabtar
The Lemon Party
297
|
Posted - 2012.06.25 11:24:00 -
[42] - Quote
Liza Hawkeye wrote:I'm trying to get what happened with reverse engineering in the last couple of weeks. For example I got only 3 Accelerated Ejection Bay BPCs from 60 reverse engineering works made with skills 5-5-4 from Intact Ancient Relics and it caused me to lose money on this according to total cost of each BPC. As for now - my chance on reverse engineering is is like 40% and it makes me cry, because I'm just loosing money. I'm wondering - will CCP 'repair' tech III reserch/production or it will be left unprofitable/low-profitable thing?
Yours sincere, Liza Hawkeye Don't we have an S&I forum for this kind of thing? EVE shall be purged by fire - please Gods let them ALL burn in Jita. |

Liza Hawkeye
Discontent Innovations
1
|
Posted - 2012.06.25 12:28:00 -
[43] - Quote
Andy DelGardo wrote:So we can verdict that Lisa is indeed CURSED! So my offer still holds and i will contact this blessing dude from a other thread to purge her curse, if u sell her to me.
bye Andy
PS: On a serious note, i just finished my batch last night and nothing odd at all. From 30 intact tries i got 5 accl. bay BPC, with a overall success rate of 25 out of 30 tries, so i'm even over the estimated 80%. So we should stop "wondering" about a odd numerical occurrence, if we cant reproduce the actual result.
Sorry Andy, I won't sell you Liza. Because of 3 things: 1. It's my Jita trader alt and nothing more 2. You don't like it's skill 3. I don't wanna sell a toon with such cool name 
Anyway, congratulations with nice RE jobs results, wish I was you  |

Andy DelGardo
68
|
Posted - 2012.06.25 12:35:00 -
[44] - Quote
@Liza Hawkeye
btw don't u "normally" deal mainly in defensive subs, which are much more stable, since cheap RE artifacts and 2-3 "good" sub per category. The only reason i do the accl. bays is simply because my trading partners like them, i hate those since getting hold of those ever camped intact artifacts sucks, i really hate to 0.01ISK over them in jita.
PS: RE market is kinda crazy in the last weeks :) couple of market manipulation, btw are u associated with "Boss" ? |

Thorn Galen
Bene Gesserit ChapterHouse Sanctuary Pact
663
|
Posted - 2012.06.25 13:01:00 -
[45] - Quote
I'm not involved in Indy. What is this reverse-engineering you speak of ? The universe is an ancient desert, a vast wasteland with only occasional habitable planets as oases. We Fremen, comfortable with deserts, shall now venture into another. - STILGAR, From the Sietch to the Stars. |

Andy DelGardo
68
|
Posted - 2012.06.25 13:05:00 -
[46] - Quote
Thorn Galen wrote:I'm not involved in Indy. What is this reverse-engineering you speak of ?
The process at studying and poking those WH Artifacts, until we have a "Heureka " moment to produce T3 Subsystems. The only strange thing to me seem, that even after 1000 of RE jobs, my stupid scientists always forget there research, so we start at square one over and over. |

Liza Hawkeye
Discontent Innovations
1
|
Posted - 2012.06.30 03:18:00 -
[47] - Quote
Okay, finally I finished these 120 jobs - took so long because was busy IRL. Results: Jobs run: 120 Jobs success: 85 Success ratio: 70.83% (Estimated: 79.8%) Valuable BPCs: 18/85 Valuable BPC ratio to other BPCs: 21.176% (Estimated: 25%)
So, to crown it all I can say that on that batch my deviation from the estimated values is normal. Not perfect, but normal. I suppose that my previous batch was just 'cursed' lol. Everything looks like Reverse Engineering is okay. |
|
|
|
Pages: 1 [2] :: one page |
First page | Previous page | Next page | Last page |