| 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) |
Kazuo Ishiguro
House of Marbles
Zzz
   |
Posted - 2007.06.20 15:00:00 -
[301]
I've been getting some interesting numbers myself lately. I've done 16 invention jobs with drones so far, using the Test Reports decrytor + 4/3/4, and only 1 has failed. Either they have their own higher base chance of success, or I've been extremely lucky; the odds of that happening at the predicted chance are about 20000 to 1 against.
------
Spreadsheets: Top speed calculation - Halo Implant tanking |
Aykido
Gallente
Lobster of Babel
|
Posted - 2007.06.21 00:01:00 -
[302]
My formula is no more than an educated guess. It is probably not accurate.
Now we have have implants to increase the chances by 1% // 2-3% // 4-5% (judging from the various other science and industry implants).
I agree that the number of Datacores seems to affect the success rate, and the negative modifier for lack of base items was only to fit the skill modifier as the most important variable. This is not necessarily true.
But if I average the numbers we have documented my formule works pretty well, Caldari ships and Minmatar modules are way above the predicted success rate, but Gallente ships fall abyssmally short.
I am fairly confident that there have been at least two patches that did modify the success rate, i.e. lower it. But I am also certain that the odd success stories of abnormal success rates are just that; abnormal streaks of good fortune or bad luck.
Or perhaps an occassional bug 
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) |
Trak Cranker
Feral Tendency
Ratel Alliance
|
Posted - 2007.06.21 00:13:00 -
[303]
Would it not make som kind of sense, if the number and/or type of datacores was tied together with the level in the corresponding skill?
|
Aykido
Gallente
Lobster of Babel
|
Posted - 2007.06.21 00:19:00 -
[304]
Perhaps,
But other instances of such linking are only in the form of minimum requirements. So its hard to say without several 1000 attempt samples to compare.
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) |
Trak Cranker
Feral Tendency
Ratel Alliance
|
Posted - 2007.06.21 00:57:00 -
[305]
I still just hope that CCP are not mucking around with formula/chances a their whim.
In order for this to be a business, you have to be able to trust your experience and that there is a system to it. Otherwise its just a new lottery.
Not that this is not, to some extent, but in this system I hope you can trust that there is a system to the odds, and as such take your bet on a formula you believe in or simpel gut feeling, based on experience.
If there is a factor in there, that can change periodically overall or itemwise - it will become an industrial nightmare.
|
Fred Torrington
|
Posted - 2007.06.29 10:06:00 -
[306]
Edited by: Fred Torrington on 29/06/2007 10:07:07
Edited by: Fred Torrington on 29/06/2007 10:06:40
I'm in the middle of doing about 300-400 jobs on the test server to help you guys with numbers. Once I'm done, I'll post my results, along with whether i used a base item, what my skill levels were at, etc.
If ya want an update, look for me in game.
Character name is Denaris Torington.
|
Meau
|
Posted - 2007.06.29 13:56:00 -
[307]
Edited by: Meau on 29/06/2007 13:55:35
 Originally by: Qual
Originally by: Aykido
Invention Formulae
I think you are getting close.
But!
I am not ready to accept that the formula is aware of the job "type" (ship/mod/rig). When I see a formula only based the "reverse engeneering" chance, i'll be happy. (And you are nearly there. Actually it could be modified to be exactly like that with very little efford.)
While Aykido gives 20%,30%,40% base chances, arbitrarily, you can easily use the reverse engineering chance from the datadump, which is 20% for ships and 40% for nearly everything else(at least everything else i checked).
Furthermore i find it very unlikely that meta-level formula is different for different types.
It will only be one. There is no point in having a higher basechance and then also a higher negative modifier for rigs compared to ships.
And Base item = meta-level 0 = no item. Believe me ;P
These three points will simplify the whole thing into one universal simple formula only using variables which are found in game or the database.
Thus no arbitrary ones left.
@Aykido give the above a try :)
|
Miss Praise
|
Posted - 2007.06.29 14:29:00 -
[308]
Not wishing to change the subject here but.....
CCP should cough up with the formula or stop playing EVE..... DEVS involved with writing invention that play now have a distinct advantage in the game knowing exactly what the formula is and how to get the best out of it.
This is just one example of how DEVS have more benefits over gameplay.
|
Trak Cranker
Serenity Inc
|
Posted - 2007.06.29 22:54:00 -
[309]
Originally by: Meau
And Base item = meta-level 0 = no item. Believe me ;P
I actually do have a hard time believing you there.
Why should no item be as good as a meta 0 item? (except make a prettier formula)
|
Meau
|
Posted - 2007.06.29 23:57:00 -
[310]
Edited by: Meau on 30/06/2007 00:00:34
Originally by: Trak Cranker
Originally by: Meau
And Base item = meta-level 0 = no item. Believe me ;P
I actually do have a hard time believing you there.
Why should no item be as good as a meta 0 item? (except make a prettier formula)
To put it bluntly, i have done 300+ ship invention jobs without inserting an item, and about 250 with a ship inserted. If there is a difference in the success chance it should be extremely low*.
(while errors with statistics of 200-300 tries are still 1%-2%, my chance with base item is lower, thus giving a strict bound on how much higher it could be).
You will still quickly find that what most people do is:
Use a base item if the cost is negligible, you never know, use none if it would increase the cost by more than a few percent.
Simpleness of the formula is of course also a point  . Additionally all meta-levels are integers, so the natural choice would be -1. As the meta level can make quite a difference for the success chance the result of this would be at least 3%(f.e. with Aykidos formula some 7% or so) or something, which would be notice and measurable. I dont think they built in meta_level = 0.3 if no item is used or something like that.
If they did, its still not cost-efficient to use an item if it increases the chance by ~1% in the majority of cases, so as my final conclusion, could be that the effect is there, but then its negligible and we thus dont need to worry about it 
|
|
|
Paul Tarsus
|
Posted - 2007.07.07 22:58:00 -
[311]
There's a methodical way of doing this, but it would take a lot of data and a lot of isk. If any of you guys wanna try it on test server and get the data go for it...
You would have to focus on one blueprint and have a large amount of samples. You get the minimal skills necessary (level 1 for all required skills for the specific blueprint) and do a ton of invention attempts (without base item or decryptors just skills)... record success vs failure rate so you can get an average success %.
Then you train one of the skills to the next level and repeat the process of getting a lot of sample runs for a success %
repeat that until you get that skill to level 5
then you repeat the above for all the other skills all the while recording the success % at each level.
then through some simple algebra you can figure out how much of a difference training the skill made:
(success% level 1)*(variable)=(success % level 2)
solve for the variable at each increment (from level 1 to 2... 2 to 3.. and so on)
then you get an average value for the 4 variables on one skill and that should be how much it modifies the success rate.
Repeat this for all skills for the blueprint.
Then once you have the success multiplier(variable) for each skill you can do a lot of runs with a base item and use the same equation as for the skills to solve for the success multiplier for a base item.
This is a brute force method of doing this and it would only have information for the specific blueprint... to figure out the amount that skills increase success for a certain category you would have to once again repeat all of the steps above for a couple of blueprints.
Also the results won't be perfect because this is assuming that all of the skills are going to multiply the success rate rather than go through a complex equation... it basically sums up what might be a long and complicated equation
Also as you get closer to 5/5/5 skills the success rate doesn't increase as much so the multiplier will become less accurate unless you do a lot more runs. 
Like I said this is a brute force simple mathematical way of going about this and would work if CCP doesn't mess around with inventing while you go about this experiment... also it would take a lot of time and a lot of characters to do it efficiently and correctly.. as well as a lot of runs
Hope i didn't discourage you guys too much , but guessing the equation is hardly a way to go about it... maybe try and find someone who's a statistics major?
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.07 23:53:00 -
[312]
Originally by: Paul Tarsus
You would have to focus on one blueprint and have a large amount of samples. You get the minimal skills necessary (level 1 for all required skills for the specific blueprint) and do a ton of invention attempts (without base item or decryptors just skills)... record success vs failure rate so you can get an average success %.
Hey, have you been looking over my shoulder? :)
|
Paul Tarsus
|
Posted - 2007.07.08 03:30:00 -
[313]
well if you've been doing that I bid you good luck... your gonna either need a lot of time or a lot of chars and do it prolly on test server
I just figured since everyone was giving their '2 cents worth' I would just help everyone save their pennies and give something with a stronger mathematical basis
|
EDacIouSX
Yashio Factory
|
Posted - 2007.07.09 23:11:00 -
[314]
So it seems for ships inventions having a ship or not having one as an item makes no difference at all right?
|
Vladimir Tinakin
Caldari
Hadean Drive Yards
Archaean Cooperative
|
Posted - 2007.07.10 02:42:00 -
[315]
Originally by: EDacIouSX
So it seems for ships inventions having a ship or not having one as an item makes no difference at all right?
That does appear to be the case. Or at least, it's not statistically significant enough to notice without someone rolling out 1000 invention attempts with the same setup.
I typically just use the "max run, crap chance" decryptor and toss in a half dozen+ jobs when I want to get a T2 ship BPO semi-reliably. Just got 20 runs of hulk goodness out of 7 attempts, so I'm happy. 
-----------------------------------------------
Adm Vladimir Tinakin
CFO
Hadean Drive Yards |
Elisa Day
Koshaku
Interstellar Starbase Syndicate
|
Posted - 2007.07.10 17:19:00 -
[316]
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): 0.4x
Item meta level (0/1/2/3/4): none used
Attempts: 912
Successes: 171
|
Vlammetje
|
Posted - 2007.07.11 07:08:00 -
[317]
Originally by: Miss Praise
This is just one example of how DEVS have more benefits over gameplay.
Go away Goon.
or
It's their game, if you don't like it: Go play another game.
|
Dr Fighter
|
Posted - 2007.07.17 15:56:00 -
[318]
im almost finished with 100 attempts at a module with encryption methods 5 and the two sci skills at 4, using meta 1 item for all 100 attemps and no decryptors. I will post a link to a screeny of the table ive made as each batch has gone in to bake over the last month with live success rate statistics.
so far its looking like 50%, which tbh with the skills and the meta lvl 1 item, seems quite a low success rate  Also interestingly since the very first batch the running success rate has droped EVERY batch, which might point to a ingame control over the number of items being invented/created.
see for your self, just 2 more attempts cooking atm and 4 more in the next day.
interesting, also the profit is looking like 300m for a total investment of over 2 billion isk, which to me is a really crappy turnover compared to others (tho could just be the module im doing).
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.17 17:06:00 -
[319]
Edited by: Trak Cranker on 17/07/2007 17:13:39
100 attempts is nothing in terms of having a solid basis for conclusions.
But I will give you this: 50% is pitiful, in terms of what you would expect. And I have seen numbers to insinuate (it is that at best) that skills might be less important here than most would like.
But it is pure conjecture at this point.
|
William Alex
Viscosity
|
Posted - 2007.07.17 17:46:00 -
[320]
Originally by: Trak Cranker
Something to the tune of:
chance of reinvention from BP *
1+0.1*EncSkill *
(1-0.05*(5-DC1skill))^#DC1 *
(1-0.05*(5-DC2skill))^#DC2 *
Decryptor *
1+0,5*metalevel*
implant
I really like this formulea so far i've got almost exactly these results with all my attempts.
Please give us a 1 depth skill queue CCP.
|
|
|
Dr Fighter
|
Posted - 2007.07.19 15:26:00 -
[321]
Below are my invention results in a table recored exactly what every batch resulted over the past month and a half.
All the invention attempts where done with lvl 5 encryption methods lvl 4 for both sci skills and a meta 1 item, no decryptors where used.
My invention batches
because i had a few datacores left and no more meta items, i did 3 attempts wihtout an item, the results were two out of th 3 worked lol
|
Trak Cranker
Serenity Inc
|
Posted - 2007.07.19 15:39:00 -
[322]
Can you flesh it out a bit with what item(s) you invented?
|
Galtan Deus
|
Posted - 2007.07.21 20:40:00 -
[323]
Originally by: Miss Praise
Not wishing to change the subject here but.....
CCP should cough up with the formula or stop playing EVE..... DEVS involved with writing invention that play now have a distinct advantage in the game knowing exactly what the formula is and how to get the best out of it.
This is just one example of how DEVS have more benefits over gameplay.
That go for ships too?
|
Elisa Day
Koshaku
Interstellar Starbase Syndicate
|
Posted - 2007.07.22 17:55:00 -
[324]
Skill level (Encryption/DC1/DC2): 5/4/4
Interface (Ship/Tuner/Data): Data
Number of datacores from one group: 2
Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): 0.4x
Item meta level (0/1/2/3/4): none used
Attempts: 480
Successes: 100
|
Sallie Wonder
|
Posted - 2007.08.07 00:00:00 -
[325]
bump for more results!
|
Elisa Day
Koshaku
Interstellar Starbase Syndicate
|
Posted - 2007.08.18 16:17:00 -
[326]
Edited by: Elisa Day on 18/08/2007 16:17:20
Skill level (Encryption/DC1/DC2): 5/5/4
Interface (Ship/Tuner/Data): Data
Number of datacores from one group: 2
Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): 0.4x
Item meta level (0/1/2/3/4): none used
Attempts: 703
Successes: 131
This shows a disturbing trend to me, compare previous results:
4/4/4: 18.75% on 912 tries
5/4/4: 20.83% on 480 tries
5/5/4: 18.63% on 703 tries
I'd like to believe that these samples are stasticially significant enough to tell us that science lvl 5 skills for invention might be broken. My characters all have 5/5/5 now though so we'll see if the "bad luck" holds or if it was just a big fluke. Comments from any statistics gurus?
|
Lady Natacha
Minmatar
Water and Power
|
Posted - 2007.08.18 17:05:00 -
[327]
Originally by: Elisa Day
Edited by: Elisa Day on 18/08/2007 16:17:20
Skill level (Encryption/DC1/DC2): 5/5/4
Interface (Ship/Tuner/Data): Data
Number of datacores from one group: 2
Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): 0.4x
Item meta level (0/1/2/3/4): none used
Attempts: 703
Successes: 131
This shows a disturbing trend to me, compare previous results:
4/4/4: 18.75% on 912 tries
5/4/4: 20.83% on 480 tries
5/5/4: 18.63% on 703 tries
I'd like to believe that these samples are stasticially significant enough to tell us that science lvl 5 skills for invention might be broken. My characters all have 5/5/5 now though so we'll see if the "bad luck" holds or if it was just a big fluke. Comments from any statistics gurus?
Can you say "Tactical Shield Manipulation skill bug"? At least there it could be demonstrated to be broken because the bonuses for that skill are documented.
We have noticed a substantial decrease in success rates after achieving level 5 in science skills required for our regular invention jobs.
250M implant may disappear: Linkage
|
Liisa
Infinitus Odium
The Church.
|
Posted - 2007.08.18 20:53:00 -
[328]
Originally by: Lady Natacha
Originally by: Elisa Day
Edited by: Elisa Day on 18/08/2007 16:17:20
Skill level (Encryption/DC1/DC2): 5/5/4
Interface (Ship/Tuner/Data): Data
Number of datacores from one group: 2
Decryptor (0.4x/0.8x/1.0x/1.1x/1.3x): 0.4x
Item meta level (0/1/2/3/4): none used
Attempts: 703
Successes: 131
This shows a disturbing trend to me, compare previous results:
4/4/4: 18.75% on 912 tries
5/4/4: 20.83% on 480 tries
5/5/4: 18.63% on 703 tries
I'd like to believe that these samples are stasticially significant enough to tell us that science lvl 5 skills for invention might be broken. My characters all have 5/5/5 now though so we'll see if the "bad luck" holds or if it was just a big fluke. Comments from any statistics gurus?
Can you say "Tactical Shield Manipulation skill bug"? At least there it could be demonstrated to be broken because the bonuses for that skill are documented.
We have noticed a substantial decrease in success rates after achieving level 5 in science skills required for our regular invention jobs.
So the encryption skill works, the other skills are bugged at level 5? It looks like this to me, but just to be sure that I understand correctly what is being presumed.
----------------------------------
Currently pursuing a "deviant playstyle." |
Elroth
Clerks Industries
|
Posted - 2007.08.20 09:29:00 -
[329]
hmm so i am thinking stopping my skill at the 3 hours left its on right now :P
|
Chruker
|
Posted - 2007.09.09 11:38:00 -
[330]
Here is how the average success rate evolved for a couple of my invention scenarios.
http://junk.chruker.dk/files/temp/average_success_changing_with_more_attempts.gif
Notice how the blue scenario starts changing from an average of 39% at 850 attempts to an average of 40.6% at a 1000 attempts.
----- http://games.chruker.dk/eve_online -----
Top wishes:
- No daily downtime
- Faster training on sisi
- Speedup IGB table rendering |
|
|
|
|
| |
|
| 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 |