Pages: 1 2 3 4 5 6 7 8 9 [10] 11 12 13 14 .. 14 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Titus Quintus
|
Posted - 2007.06.04 13:35:00 -
[271]
Originally by: Aykido Invention Formulae (my guess)
This sounds very reasonable. The formula makes all kinds of sense, especially since it uses very round numbers as a base and the stacking of skills is actually very simplistic.
|
Aykido
Gallente Lobster of Babel
|
Posted - 2007.06.04 18:34:00 -
[272]
Edited by: Aykido on 04/06/2007 18:56:06 Ok so in a formula format my above guess looks like this:
Ship = 20% * skill(1+((0,05*level)*(1+(0,1*level))*(1+(0,1*level))) * decryptor * base item (1-(0,05*(meta 4 - actual meta))) = success chance. >>> Example, I try to invent a Zealot with skills at 5/5/4 using a circular logic decryptor. My chance should be 20% * 1,525 * 0,4 * 0,75 (or 0,8 with a built Omen in the job) = 9,15% chance (or 9,76% with the built Omen).
Tuner = 30% * skill(1+((0,05*level)*(1+(0,1*level))*(1+(0,1*level))) * decryptor * base item (1-(0,06*(meta 4 - actual meta))) = success chance. >>> Example, I try to invent a Salvage Tackle 2 with a skill set of 5/4/4 and a sacred manifesto decryptor. My chance should be 30% * 1,49 * 0,8 * 0,7 (or 0,76 with a built salvage tackle I) = 25,032% (or 27,1776% with T1 base item).
Data = 40% * skill(1+((0,05*level)*(1+(0,1*level))*(1+(0,1*level))) * decryptor * base item (1-(0,07*(meta 4 - actual meta))) = success chance. >>> Example, I try to invent a Cov Ops cloak with a skill set of 4/3/2 and an Installation Guide decryptor. My chance should be 40% * 1,364 * 1,3 * 0,65 (or 0,72 with a built Protype cloaking device I) = 46,1032% (or 51,06816% with the protocloak as base item). >>> Example, I try to invent an Invul field II with a skill set of 5/5/5 and Prototype diagram decryptor. My chance should be 40% * 1,5625 * 1,1 * 0,65 (or 0,72 with T1 invul field; or 0,86 with a V-M15 Braced Multispectral as base item) = 44,6875% (or 49,5% with T1; or 59,125% with V-M15).
Is this understandable?
Selling PERFECT PRINTS of 674 of the 730 seeded T1 BPOs: Ships, modules, rigs, drones, ammo, probes, components and capital mods, drones & ammo. (Cap Ships, Cap Components and Outposts 2007/2008) |
Dominique Vasilkovsky
Techmart Industries
|
Posted - 2007.06.05 10:50:00 -
[273]
I like the formula for the skill effect, sounds very plausible.
I'm not entierly sure about the rest though as CCP loves doing everything using formulas and not fixed value lookup. And the only known database values are the reverese engineering and meta values.
Do anyone have a link to a database website that include the reverse enineering numbers? I lost the link I had for one.
Signature approved by Eldo |
Trak Cranker
Feral Tendency Ratel Alliance
|
Posted - 2007.06.05 12:18:00 -
[274]
Where is it I can find the reverse engineering value?
|
sahtila
principle of motion Interstellar Alcohol Conglomerate
|
Posted - 2007.06.05 13:55:00 -
[275]
Originally by: Aykido Edited by: Aykido on 04/06/2007 18:56:06 Ok so in a formula format my above guess looks like this:
Ship = 20% * skill(1+((0,05*level)*(1+(0,1*level))*(1+(0,1*level))) * decryptor * base item (1-(0,05*(meta 4 - actual meta))) = success chance. >>> Example, I try to invent a Zealot with skills at 5/5/4 using a circular logic decryptor. My chance should be 20% * 1,525 * 0,4 * 0,75 (or 0,8 with a built Omen in the job) = 9,15% chance (or 9,76% with the built Omen).
Tuner = 30% * skill(1+((0,05*level)*(1+(0,1*level))*(1+(0,1*level))) * decryptor * base item (1-(0,06*(meta 4 - actual meta))) = success chance. >>> Example, I try to invent a Salvage Tackle 2 with a skill set of 5/4/4 and a sacred manifesto decryptor. My chance should be 30% * 1,49 * 0,8 * 0,7 (or 0,76 with a built salvage tackle I) = 25,032% (or 27,1776% with T1 base item).
Data = 40% * skill(1+((0,05*level)*(1+(0,1*level))*(1+(0,1*level))) * decryptor * base item (1-(0,07*(meta 4 - actual meta))) = success chance. >>> Example, I try to invent a Cov Ops cloak with a skill set of 4/3/2 and an Installation Guide decryptor. My chance should be 40% * 1,364 * 1,3 * 0,65 (or 0,72 with a built Protype cloaking device I) = 46,1032% (or 51,06816% with the protocloak as base item). >>> Example, I try to invent an Invul field II with a skill set of 5/5/5 and Prototype diagram decryptor. My chance should be 40% * 1,5625 * 1,1 * 0,65 (or 0,72 with T1 invul field; or 0,86 with a V-M15 Braced Multispectral as base item) = 44,6875% (or 49,5% with T1; or 59,125% with V-M15).
Is this understandable?
I still believe that the base percentage depends on how many datacores you need, hac invention seems more difficult than af invention.
|
Titus Quintus
|
Posted - 2007.06.05 14:16:00 -
[276]
Edited by: Titus Quintus on 05/06/2007 14:15:58
Originally by: sahtila
I still believe that the base percentage depends on how many datacores you need, hac invention seems more difficult than af invention.
I am currently running tests to look into this.
Don't expect an answer soon... For statistical significance I need at least 200 invention runs per meta-level and datacore needs. I am actually building upon chruker's samples since they are so well documentated. This means 1/1/2 skills.
|
Chief Unshiki
|
Posted - 2007.06.08 13:52:00 -
[277]
Just performed my first invention runs today, and was pleasantly surprised by the results, i'd hoped for 4, maybe 5 successes and wound up with 7 out of 9. Probably just beginner's luck though
Skill level (Encryption/DC1/DC2): 4/4/4 Interface (Ship/Tuner/Data): Data Number of datacores from one group: 2 Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): 1.3x Item meta level (0/1/2/3/4): 0 Attempts: 9 Successes: 7
I just started 9 more runs
The product is Heavy Missile Launcher
|
Chief Unshiki
|
Posted - 2007.06.08 19:55:00 -
[278]
the 9 more attempt yielded the same results actually. so a grand total of 14 of 18 successes. i'm pretty pleased with that
Skill level (Encryption/DC1/DC2): 4/4/4 Interface (Ship/Tuner/Data): Data Number of datacores from one group: 2 Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): 1.3x Item meta level (0/1/2/3/4): 0 Attempts: 18 Successes: 14
|
Ranis Garr
Amarr GalSpan Inc
|
Posted - 2007.06.08 19:57:00 -
[279]
Edited by: Ranis Garr on 08/06/2007 19:56:21 This is deep... lol --------------------- "Dont try and be a great man, just be a man. Let history judge who you are." ~William Riker
<IMG SRC="http://www.hazardous-gfx.com/forums/uploads/imagehost/U7-1181327778.gif" |
Vladimir Tinakin
Caldari Hadean Drive Yards Archaean Cooperative
|
Posted - 2007.06.08 22:34:00 -
[280]
Yep. Very hard to reverse-eng a formula when you don't know all the variables, and so you have to be methodical as all hell to get to the meat of the matter.
----------------------------------------------- Adm Vladimir Tinakin CFO Hadean Drive Yards |
|
Velius Donegol
|
Posted - 2007.06.09 12:06:00 -
[281]
Edited by: Velius Donegol on 09/06/2007 12:08:57 Edited by: Velius Donegol on 09/06/2007 12:05:13 Man, invention is frustrating. I had 17/21 success on module invention with 4/4/4 skills, not using any decryptor or meta item. Then In the next 12 runs I had 2 successes. I just wanted 20 BPCs to start with, damnit!
|
Rotti
|
Posted - 2007.06.11 05:59:00 -
[282]
Aykido I have placed your formula into a spreadsheet and get the same results for ships with the 0.4 decryptor. However with a 1.3 decryptor in the sheet I get 30% success. Is this right cause this would mean that it is better to use this decryptor in ship invention 100 attempt would produce 30 5 run bpc as opposed to 9 10 run ones.
|
Trak Cranker
Feral Tendency Ratel Alliance
|
Posted - 2007.06.11 06:34:00 -
[283]
Originally by: Rotti Aykido I have placed your formula into a spreadsheet and get the same results for ships with the 0.4 decryptor. However with a 1.3 decryptor in the sheet I get 30% success. Is this right cause this would mean that it is better to use this decryptor in ship invention 100 attempt would produce 30 5 run bpc as opposed to 9 10 run ones.
Remember it also depends on what the decryptors cost. Although given your numbers, the 1.3 would have to increase the cost pr bpc run by 67% over the 0.4 one for them to be equal. And then there is the cost of any meta items to factor in too, if you use such. That cost is smaller pr runs on a try on the 0.4 one.
Formula is still up in the air ofc. Although it seems it holds up pretty well.
|
Mighty Baz
HUSARIA Curatores Veritatis Alliance
|
Posted - 2007.06.11 07:42:00 -
[284]
Originally by: Aykido Invention Formulae (my guess)
Base chance: Ship 20% Rig 30% module 40%
Skill modifier (Add as multiplier to base chance). Encryption skill 5% per level Relevant science skills increase the effect of the encryption skill by 10% per level. skill combo of 4/4/4 gives 39,2% increase in chance. skill combo of 5/5/5 gives 56,25% increase in chance.
Decryptors (only "known" variable) range of -60% to +30% chance modifier. Unless coldfront figures have been tweaked since the last data dump.
Base item works as negating the penalty of not having one. penalty is -25% for ships, -30% for rigs and -35% for modules. Meta level 0 gives penalies of -20%, -24% and -28%. Higher meta levels are only for modules. Meta level 1 gives penalties of -21% . Meta level 2 gives penalty of -14%, Meta level 3 gives penalties of -7% and meta level 4 incures no base item penalty.
With this formula, a max skilled character inventing a module with best decryptor (+30%) and best named item (-0%) will have a 81,25% chance of success. (4 out of 5).
This would explain the failure rate we have on strip miners and cov-op cloaks. Giving a max success rate of 63,47% (3 out of 5). Strips with Test reports and 4/4/4 skill combo would have 47,85% success rate.
Inventing a ship with a +9 run decryptor and max skills would give 10% max chance on invention (10,41% using a built T1 ship as base item).
Inventing a ship with a +30%/+4run decryptor and max skills would result in 32,5% success rate (or 33,85% with a built ship).
______________________________________________ Husaria recruits
based on legendary XVII century Polish winged cavalry |
Kirja
Rampage Eternal Ka-Tet
|
Posted - 2007.06.11 12:05:00 -
[285]
Edited by: Kirja on 11/06/2007 12:11:59 Edited by: Kirja on 11/06/2007 12:09:27 I'll add my data
Skill level (Encryption/DC1/DC2): 3/4/4 Interface (Ship/Tuner/Data): Data Number of datacores from one group: 2 Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): none Item meta level (0/1/2/3/4): 2 Attempts: 30 Successes: 15 (50%)
Skill level (Encryption/DC1/DC2): 4/4/4 Interface (Ship/Tuner/Data): Data Number of datacores from one group: 2 Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): none Item meta level (0/1/2/3/4): 2 Attempts: 25 Successes: 11 (44%)
Skill level (Encryption/DC1/DC2): 4/4/4 Interface (Ship/Tuner/Data): Data Number of datacores from one group: 2 Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): none Item meta level (0/1/2/3/4): 0 do i understand correctly that t1 is 0 meta lvl? Attempts: 32 Successes: 17 (53%)
|
Sinova Mortario
|
Posted - 2007.06.11 13:04:00 -
[286]
I've not seen this discussed anywhere after a brief bit of searching, so forgive if this has already been gone over, but my results are showing that Level 5 in Minmatar Encryption methods has actually decreased my success rate.
Here are my numbers:
Skills at level 4, using no decryptor, and best named item:
146 attempts: 108 Successful, 38 Failed = 73.97%
Skills at 4 for secondary and 5 for Encryption Methods, no decryptor and best named item:
154 attempts: 112 Successful, 42 Failed = 72.73%
Overall drop of 1.24% in success.
Has anyone else seen this type of change? Is this a bug perhaps? |
Trak Cranker
Feral Tendency Ratel Alliance
|
Posted - 2007.06.11 13:38:00 -
[287]
I think you need more deviant numbers and/or more data sets to conclude anything is wrong with those numbers.
|
Sinova Mortario
|
Posted - 2007.06.11 14:29:00 -
[288]
Originally by: Trak Cranker I think you need more deviant numbers and/or more data sets to conclude anything is wrong with those numbers.
Really? You don't think that 150 attempts is enough to get a good idea of what the success rate would be.
All of those were using the exact same thing except for the first half (about 150) was at Encryption level 4 and the second half (about 150) were at Encryption level 5.
Not being sarcastic, just trying to understand. |
Titus Quintus
|
Posted - 2007.06.11 14:53:00 -
[289]
Originally by: Sinova Mortario
Originally by: Trak Cranker I think you need more deviant numbers and/or more data sets to conclude anything is wrong with those numbers.
Really? You don't think that 150 attempts is enough to get a good idea of what the success rate would be.
All of those were using the exact same thing except for the first half (about 150) was at Encryption level 4 and the second half (about 150) were at Encryption level 5.
Not being sarcastic, just trying to understand.
There is nothing wrong with your numbers unfortunately. It's just the game of statistics. Try doing a hypothesis test whether your data corresponds to the same binomial distribution and see the error margins of such a small sample.
|
Sinova Mortario
|
Posted - 2007.06.11 15:53:00 -
[290]
Thanks for the replies, but 300 samples is too small a number to get an average?
I'll not get into deep statistical analysis on this, as it really doesn't need it as I don't have 10's of thousands of attempts recorded. However, that said, it's hard to believe that 300 samples is not enough to give an average of success, especially when there are only 2 results being looked at, success or fail. Heads or tails. You flip a coin, yes, you can get 150 heads in a row. Odds are that you will not get 150 in a row, but that you will get around 50% of them heads.
All I'm wondering about is that training the encryption methods skill to level 5 has not had a positive impact on my results, and in fact has had a slightly negative result, and I curious if anyone else has had this same situation arise?
|
|
Mckayla X
|
Posted - 2007.06.11 16:29:00 -
[291]
Originally by: Sinova Mortario Thanks for the replies, but 300 samples is too small a number to get an average?
I'll not get into deep statistical analysis on this, as it really doesn't need it as I don't have 10's of thousands of attempts recorded. However, that said, it's hard to believe that 300 samples is not enough to give an average of success, especially when there are only 2 results being looked at, success or fail. Heads or tails. You flip a coin, yes, you can get 150 heads in a row. Odds are that you will not get 150 in a row, but that you will get around 50% of them heads.
All I'm wondering about is that training the encryption methods skill to level 5 has not had a positive impact on my results, and in fact has had a slightly negative result, and I curious if anyone else has had this same situation arise?
Statistically you get about a 3% error at 1000 sample size which is generally regarded as close enough, so 300 is more like a 10% error, which would easily cover the difference you are seeing.
|
Sinova Mortario
|
Posted - 2007.06.11 17:27:00 -
[292]
Originally by: Mckayla X
Statistically you get about a 3% error at 1000 sample size which is generally regarded as close enough, so 300 is more like a 10% error, which would easily cover the difference you are seeing.
Thanks.
|
Alvara
Kuiper Belt Industries Empire Research
|
Posted - 2007.06.11 19:36:00 -
[293]
Wish I had been documenting since the beginning. Since I've done an easy 500 invention jobs with all decryptors and skill levels.
Good job for those that are keeping documentation and sharing.
One tid bit I can add, is that ccp does change % values regularly to control the amt that is invented of an item.
Tired of Waiting? Use Empire Research |
Trak Cranker
Feral Tendency Ratel Alliance
|
Posted - 2007.06.11 21:21:00 -
[294]
Originally by: Alvara One tid bit I can add, is that ccp does change % values regularly to control the amt that is invented of an item.
Can you confirm this with more hard evidence? If true, and I sould be very surprised to learn that it is, its one of the dumber things I have heard of.
They have no incentive to control the amount of BPCs invented. It is a totally self regulating market.
|
Whip Slagcheek
|
Posted - 2007.06.12 02:16:00 -
[295]
In response to Alvara's comment about them regularly changing the chances:
Are you really saying that CCP is watching invention / market closely enough to alter invention rates? This is the company that went for over a year before they looked at the t2 situation and realized there were now 3x as many people and even less BPOs due to some being taken out of circulation.
I have a hard time believing CCP is regulating the invention market when their past response to the T2 market was a dev blog basically saying, "oh yeah. Guess we should've been looking at the T2 BPO production / demand those couple of years. Guess we should seed a bit of everything now." (Even ammo which was debatable if we needed more of that.)
|
Kirja
Rampage Eternal Ka-Tet
|
Posted - 2007.06.12 10:34:00 -
[296]
Edited by: Kirja on 12/06/2007 10:39:14
Originally by: Sinova Mortario Thanks for the replies, but 300 samples is too small a number to get an average?
300 sample is not too small to get an average. But average of random set of data is random value itself. Therefore it has a deviation. The bigger data set we have the lower deviation we will get. The difference you recieved can be easily explained by this deviation therefore we can not conclude that lvl 5 encryption actualy reduced your chances.
What i find more interesting in this case (since its rather large set of data) is that you received a success rate of ~73% with lvl 4 skills which is ~13% more than Aykido's equation suggests with maxed out skills.
|
Mighty Baz
HUSARIA Curatores Veritatis Alliance
|
Posted - 2007.06.12 12:56:00 -
[297]
Edited by: Mighty Baz on 12/06/2007 13:00:18 Edited by: Mighty Baz on 12/06/2007 12:56:40
My two cents, It seems that decryptors and named items are the most important to increase odds receiving t2 bpcs.
Secondary the above listed skills are not corresponding to generated runs properly. Probably CCP fixed avarage of need skill's levels. They have thought that huge numbers of players as a inventors are achieved 3-4 lvl in game (like Gauss figure), because of long time training of lvl5 (25-30 days, and CCP would like to take advantage of invention as a first. Plz remember that invention could replace bpo t2 in some definied time (Kali 2 ?). Who knows maybe t2 bpo will be reversing to limited bpc.
Thirdly, I dont understand why invention is such a expensive. Avarage of success rate is moreless 50%, and cost of decryptors, datacores, named items are too high at least. We expect much more components in hidden encounters.
And last one, datacores are limited in game, because of limited research points...
anything else? yea it's true invention is limited like DEV's ideas..
cheers
______________________________________________ Husaria recruits
based on legendary XVII century Polish winged cavalry |
Thira Rans
|
Posted - 2007.06.19 09:13:00 -
[298]
any news on the invention formula?
P.S: can't we make this a sticky?
|
Meha Mott
Minmatar Carebear Research and Produktion Agency
|
Posted - 2007.06.19 18:56:00 -
[299]
First of all, sorry for my bad english.
I made, up to now, more than 1000 inventions, but unfortunately i am not a fan of writing down a lot of numbers.
What i can say, is, that my % chance is somewhat between 50 and 80 % depending on the meta item and the decryptors i use, or not
I only invent modules, no ships, no rigs.
But there is 1 module that is totaly breaking all rules.
My sucsess rate is near to 100 % with large shield extender.
I made round about 150 inventions and only 2 without sucsess.
My skills are 4/4/4, i use the best meta item and the 1.3 decryptor.
My bee some other people have same experiences and are willing to share it here.
|
Trak Cranker
Feral Tendency Ratel Alliance
|
Posted - 2007.06.19 20:16:00 -
[300]
Given the latest formula that has been posted here, I think it is possible to get to or above 100% on some combinations. Correct me if I am wrong?
|
|
|
|
|
Pages: 1 2 3 4 5 6 7 8 9 [10] 11 12 13 14 .. 14 :: one page |
First page | Previous page | Next page | Last page |