These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

Science & Industry

 
  • Topic is locked indefinitely.
12Next page
 

Lets talk about Cost Index (warning: math inside)

First post
Author
Previous Topic Next Topic
Nina Smirnoff
Air Traffic Control
#1 - 2014-07-29 16:04:04 UTC
I like Crius. I like most changes. I love the idea of Cost Index since I read about it, but I have 2 disappointment with it. I'll talk about it later. Let's get down to business,

I'll talk about manufacturing only (no invention, research or copy). I'm watching cost index since Crius release and before that on SiSi. I took today Cost Index(CI) and I separated in 6 groups. They are grouped by Cost Index. Below I present the group median, average and standard deviation.

Group 1 - Jita (highest CI system)
9.06% - 9.06% - 0.00%
Group 2 - Top 2-10 highest CI systems
6.8% - 6.74% - 0.50%
Group 3 - Top 11-100 highest CI systems
4.73% - 4.6% - 0.59%
Group 4 - Top 101-250 highest CI systems
3.3% - 3.28% - 0.29%
Group 5 - Top 251-1500 highest CI systems
1.28% - 1.08% - 0.68%
Group 6 - Rest of the Universe
Almost no system activity. Most systems here have the default value.

Group 1 is Jita, it's so peculiar that it has a group alone. Is amazing how high Jita is comparing with the rest of the universe, even comparing to Group 2. Jita would be a lot higher but the system has a lot of stations and that gives it a huge discount. Anyway, if you are in Jita and you need help, here are 2 suggestion :
- 1 jump from Jita we have 3.93% (Maurasi - 0.9).
- 2 jumps from Jita we have 2.67% (Iyen-Oursta - 0.8)

Group 2 has very high values, I dont know why people are there. There must be a reason but I didn't try to understand it yet. Anyway, lets move on.

I want to compare Group 3 and Group 5. Comparing averages, you are saving 3.4%. That means, if you spend 1B/day, you'll save 34M/day if you are in a system of group 5 instead of group 3. So you move to a corner of universe to save 34M/day or 1B/month. Yes, you risk moving 1B/day to save 1B/month. Worth it the effort or risk? (That's my first disappointment)

Who we have in Group 3? New Caldari (1 jump from Jita), Perimeter (1 jump from Jita), Kakakela (2 jumps from Jita), Paara (2 jumps from Jita), Hek, Amarr, ...

Who we have in Group 5? That's my second disappointment. You see, I move to a system that was group 6. It has 0.09% tax (yeah, 100x lower than Jita!) and I start producing there. I'm alone (only my pos in the system and no stations with factory). I put 15 lines 23h/day and below I show you how it changed: 0.09%, 0.46%, 0.69%, 0,8%, 0,88%, 0,93%..... That's 15 lines of production...

So group 5 are systems where small groups are producing so you can't (usually) get any better then group 5. Group 6 is where there is no production or there is a really small production and/or huge bonus.

So here is my conclusion. Note that it's just a week since Crius release and numbers may change and I may be making a fool of myself, but right now, IMHO:

- cost Index spread should be bigger. I don't like the square root on CI equation. It mean that, for example, if your neighbor has a similar system but produces 10x more, he'll have a cost index just 3 times higher. 100x more? Just 10x higher index. If he has more stations the difference is even smaller.

- it's not worth the risk and effort to move too far away: you won't save much as index increase fast when you start producing.



Just my 0.2 isk.
Qoi
Exert Force
#2 - 2014-07-29 16:29:49 UTC
Math is my favourite thing Big smile

Hm, a 3 times higher system cost index can be very significant with some of the razor thin margins in eve industry. I'm a big believer in the square root. (Insert pun involving negative inputs, believing in imaginary things and riemannian manifolds here.)

http://eve-industry.org
Kireitsugu Secheh
Les chips electriques
#3 - 2014-07-29 16:45:51 UTC
Is this the same index than the one used in roid belts ?
I mean... Can I ruin an Indy income with 900 barges ?

Oh god James will have work to do soon.

How to clone swap in worm holes
Qoi
Exert Force
#4 - 2014-07-29 16:52:16 UTC
Kireitsugu Secheh wrote:
Is this the same index than the one used in roid belts ?
I mean... Can I ruin an Indy income with 900 barges ?

Oh god James will have work to do soon.

No

http://eve-industry.org
Carniflex
StarHunt
Mordus Angels
#5 - 2014-07-29 18:09:51 UTC
Well - I poked at it a little in Hek and then somewhere nearby with marginally lower index. To be frank in my opinion the cost difference was negligible enough to not justify picking up over 500 BPO's and running after some marginal savings.

Here, sanity... niiiice sanity, come to daddy... okay, that's a good sanity... THWONK! GOT the bastard.
Team Bidders
Doomheim
#6 - 2014-07-29 18:20:21 UTC
IMHO, if you are new to industry, choose wisely a low cost system.

If you are an experienced industrialist, stick to where you are until the costs become unbearable.
Nina Smirnoff
Air Traffic Control
#7 - 2014-07-29 18:20:52 UTC
Qoi wrote:
Math is my favourite thing Big smile

Hm, a 3 times higher system cost index can be very significant with some of the razor thin margins in eve industry. I'm a big believer in the square root. (Insert pun involving negative inputs, believing in imaginary things and riemannian manifolds here.)


3 times higher system cost index means 35M on 1B. And we did not put any value on the extra effort or risk.
A 2%ME team would give you ~20M benefit.
Tau Cabalander
Retirement Retreat
Working Stiffs
#8 - 2014-07-29 18:58:14 UTC
Nina Smirnoff wrote:
Group 2 - Top 2-10 highest CI systems
6.8% - 6.74% - 0.50%

Group 2 has very high values, I dont know why people are there. There must be a reason but I didn't try to understand it yet. Anyway, lets move on.

Wild guess: within 4 jumps of Jita.

People that are already established, and are also playing the wait-and-see game.
Carniflex
StarHunt
Mordus Angels
#9 - 2014-07-29 20:00:41 UTC
Tau Cabalander wrote:
Nina Smirnoff wrote:
Group 2 - Top 2-10 highest CI systems
6.8% - 6.74% - 0.50%

Group 2 has very high values, I dont know why people are there. There must be a reason but I didn't try to understand it yet. Anyway, lets move on.

Wild guess: within 4 jumps of Jita.

People that are already established, and are also playing the wait-and-see game.


I would add into mix people with any locked BPO's. After unlocking few hundred of them few days back it is truly an experience I'm rather keen on avoiding in foreseeable future. It's even worse than setting up 50+ PI colonies in a row because every time you vote to unlock one of them the interface resets closing all open tabs. It literally took me several hours of frantic clicking to get it done.

Here, sanity... niiiice sanity, come to daddy... okay, that's a good sanity... THWONK! GOT the bastard.
Steve Ronuken
Fuzzwork Enterprises
Vote Steve Ronuken for CSM
#10 - 2014-07-29 20:28:55 UTC
https://www.fuzzwork.co.uk/blueprint/costindex.php may be of interest, to people looking for differences round a central point.

(I'm now storing a snapshot of costs once per day. But it'll be a while before that's useful, and I'll have to work out some way to display it)

Woo! CSM XI!

Fuzzwork Enterprises

Twitter: @fuzzysteve on Twitter
Ginger Barbarella
#11 - 2014-07-29 21:00:47 UTC
Steve Ronuken wrote:
https://www.fuzzwork.co.uk/blueprint/costindex.php may be of interest, to people looking for differences round a central point.

(I'm now storing a snapshot of costs once per day. But it'll be a while before that's useful, and I'll have to work out some way to display it)


Just a quick fyi, the NumberOfJumps calculation on that page is off.

"Blow it all on Quafe and strippers." --- Sorlac
Gully Alex Foyle
The Scope
Gallente Federation
#12 - 2014-07-29 21:02:42 UTC
If you manufacture at, say 15% margin (on sales value), a 3% cost difference (also calculated on approximate sales value) means +20% more profit (3/15).

So you'd make 200 mil/month more for every bil/month you're making in the busier system.

Figures are higher if your profit margins are lower.

Make space glamorous! Is EVE dying or not? Ask the EVE-O Death-o-meter!
Martin Corwin
Doomheim
#13 - 2014-07-29 21:26:56 UTC
Nina Smirnoff wrote:
Who we have in Group 5? That's my second disappointment. You see, I move to a system that was group 6. It has 0.09% tax (yeah, 100x lower than Jita!) and I start producing there. I'm alone (only my pos in the system and no stations with factory). I put 15 lines 23h/day and below I show you how it changed: 0.09%, 0.46%, 0.69%, 0,8%, 0,88%, 0,93%..... That's 15 lines of production...
[...]
- cost Index spread should be bigger. I don't like the square root on CI equation. It mean that, for example, if your neighbor has a similar system but produces 10x more, he'll have a cost index just 3 times higher. 100x more? Just 10x higher index. If he has more stations the difference is even smaller.

That's exactly the point why i don't like that square root in the equation. Just look at the graph. Dev blog said it was there to produce "nicer numbers" but that's just BS. It's just there to add another flat cost because spreading out your operation adds costs as well. Something along the line of x^n where n >1 would have been way better to spread out industry operations across new eden.
Nina Smirnoff
Air Traffic Control
#14 - 2014-07-29 21:37:28 UTC  |  Edited by: Nina Smirnoff
Tau Cabalander wrote:
Nina Smirnoff wrote:
Group 2 - Top 2-10 highest CI systems
6.8% - 6.74% - 0.50%

Wild guess: within 4 jumps of Jita..


Well, I guess so.
Itamo 7.86%
Oimmo 7.09%
Uemisaisen 7.01%
Oiniken 6.84%
Saisio 6.74%
Ansila 6.55%
Ohmahailen 6.52%
Piekura 6.48%
Muvolailen 6.12%

Steve Ronuken wrote:

(I'm now storing a snapshot of costs once per day. But it'll be a while before that's useful, and I'll have to work out some way to display it)

Oh, I would love to see that. Any format works. Raw data from crest (json), text, xml, data base dump, whatever you have. =)


Gully Alex Foyle wrote:

So you'd make 200 mil/month more for every bil/month you're making in the busier system.

You have to consider that you have a higher risk moving stuff around the universe, right? And an extra effort. That has a cost too, right? Can be a direct cost if you use PushX or Frog or any other service to move your stuff around or your time and ships, but there is a number.
I don't deny that it's good, but I think it should be much better....
Steve Ronuken
Fuzzwork Enterprises
Vote Steve Ronuken for CSM
#15 - 2014-07-29 21:41:55 UTC
Ginger Barbarella wrote:
Steve Ronuken wrote:
https://www.fuzzwork.co.uk/blueprint/costindex.php may be of interest, to people looking for differences round a central point.

(I'm now storing a snapshot of costs once per day. But it'll be a while before that's useful, and I'll have to work out some way to display it)


Just a quick fyi, the NumberOfJumps calculation on that page is off.



Really? Can you give an example? Might be able to run the reason for it to ground. (It's not going to be using the same method to determine routes as Eve, but the distances should be valid.)

Woo! CSM XI!

Fuzzwork Enterprises

Twitter: @fuzzysteve on Twitter
Nina Smirnoff
Air Traffic Control
#16 - 2014-07-29 21:48:09 UTC
Steve Ronuken wrote:


Really? Can you give an example? Might be able to run the reason for it to ground. (It's not going to be using the same method to determine routes as Eve, but the distances should be valid.)


Try Jita. It doesnt show any system with 2 jumps. Funtanainen shows as 3 jumps for example (instead of 2).
Ginger Barbarella
#17 - 2014-07-30 02:19:51 UTC
Steve Ronuken wrote:
Ginger Barbarella wrote:
Steve Ronuken wrote:
https://www.fuzzwork.co.uk/blueprint/costindex.php may be of interest, to people looking for differences round a central point.

(I'm now storing a snapshot of costs once per day. But it'll be a while before that's useful, and I'll have to work out some way to display it)


Just a quick fyi, the NumberOfJumps calculation on that page is off.



Really? Can you give an example? Might be able to run the reason for it to ground. (It's not going to be using the same method to determine routes as Eve, but the distances should be valid.)


Bika to Nifshed shows 3, only 2 jumps away.

"Blow it all on Quafe and strippers." --- Sorlac
Steve Ronuken
Fuzzwork Enterprises
Vote Steve Ronuken for CSM
#18 - 2014-07-30 09:12:54 UTC
bah. Going to have to run the route generation again as something's screwed up with it.

Which is going to take a while, to generate all 27 million routes.

Woo! CSM XI!

Fuzzwork Enterprises

Twitter: @fuzzysteve on Twitter
Velicitia
XS Tech
#19 - 2014-07-30 10:06:23 UTC  |  Edited by: Velicitia
Steve Ronuken wrote:
bah. Going to have to run the route generation again as something's screwed up with it.

Which is going to take a while, to generate all 27 million routes.


Sounds like it's counting with base-1, instead of 0

(i.e. "1,2,3,4..." instead of "0,1,2,3...")

One of the bitter points of a good bittervet is the realisation that all those SP don't really do much, and that the newbie is having much more fun with what little he has. - Tippia
Banko Mato
Republic University
Minmatar Republic
#20 - 2014-07-30 12:02:54 UTC
Nina Smirnoff wrote:

...
Who we have in Group 5? That's my second disappointment. You see, I move to a system that was group 6. It has 0.09% tax (yeah, 100x lower than Jita!) and I start producing there. I'm alone (only my pos in the system and no stations with factory). I put 15 lines 23h/day and below I show you how it changed: 0.09%, 0.46%, 0.69%, 0,8%, 0,88%, 0,93%..... That's 15 lines of production...

So group 5 are systems where small groups are producing so you can't (usually) get any better then group 5. Group 6 is where there is no production or there is a really small production and/or huge bonus.
...


I experienced something similar. Moved a part of my production to a system that had no manufacturing going on and started with 12 alts there (no 24/7 line coverage, only slightly above 60% currently, so its more like having 7 full time alts working there). According to the "Jobs started during the last 24h" system info almost 100% of the manufacturing jobs in that system have been installed by me.

The cost index instantly jumped to 3.xx% after the first 2 batches of jobs. I therefore have the strong feeling that something is very wrong with the underlying algorithm when the numbers are really meant to reflect the last 28 days' job activities. In other words, I think the model reacts way too fast to changes in a system's load factor. This might be caused by not properly seeding the algorithm before the crius patch, or it is in fact calculating in a faulty way.
12Next page