Science Pack Production vs. Number of Laboratories

[hassad]

I haven't found any information on how much science should be produced as a function of the number of labs you have. What I have seen is the 5:6 red to green production ratio, and the general target of 1 pack per second, but I wanted to see if that was strictly necessary. Different research topics require different amounts of science packs, and don't take a uniform amount of time to process each step, so I did a bit of number crunching in Excel. The spreadsheet I made can be found here.

I made a number of assumptions to make this analysis. These assumptions were:

• Science pack assemblers are adequately supplied.
• Science pack assemblers are efficiently loaded and unloaded.
• Laboratories are efficiently loaded.
• The transportation of the packs to the labs is relatively short.
• Research is constantly being done without significant gaps between topics.
• Modules are not being used.

The intention of these assumptions was to reduce the effect of backlogging, which may be unavoidable for the blue and especially purple science packs. Modules were neglected to avoid further complicating the analysis.

The first sheet of the workbook contains a table of research topics and their costs, research times, and number of cycles needed. It also contains a table of total research packs needed per topic (as well as a total of each type of pack on the top row) as well as a table of packs per second per lab for each topic (with the maximums on the top row). The last table is total time in seconds to research a topic with either five or ten labs. The "Weighting" column was placed in anticipation of needing to tone down the effects of outliers, but was not used.

The sheet is hard coded in many places assuming that the player is using either five or ten laboratories at any one time. This was done because it seemed to be the general consensus that they are the "correct" numbers of labs to have.

The second sheet has the production analysis. While the maximum packs per second per lab on the previous sheet was interesting, it is complicated by the fact that there are some large outliers and also by the different production speeds of different assemblers. The bottom table compiles a calculated number of assemblers needed for each type of assembler, subdivided by the number of laboratories, and then subdivided again for each science pack type. The results were then sorted from high to low with the bias going Purple > Blue > Green > Red.

The first set of results tables at the top list the maximum number of science pack producing assemblers based on the pack being produced, assembler type, and the number of labs. The second set of results tables takes the average of the necessary factories and adds an adjustable number of standard deviations. I used two standard deviations in order to capture as much data as possible while excluding the total effects of the outliers. For the second set of tables I omitted almost all topics that certain packs were not necessary for, i.e. the average and standard deviation for blue science packs does not include the 0's from Oil Processing, Fluid Handling, etc.

There are some notable outliers in the consumption data that are noted here, but were not omitted from or altered in the analysis. These being:

• Steel Processing (red packs)
• Armor 1 (red packs)
• Logistics 3 (blue packs)
• Power Armor 2 (purple packs)

The first three are caused by the research topics' unusually low process times, and the last by the topic's unusually high purple pack demand. Steel Processing and Armor 1 are likely going to be researched manually before science pack automation is in full swing. I also concluded that possibly having a bottle-necked supply on four topics out of a total 177 is not anything to get too upset over. What's more, any amount of backlogging will mitigate that.

Wrong Data

In conclusion I found these recommended number of assemblers:

For 5 labs

Code: Select all

        Red  Green Blue  Purple
White    5    4     -     1
Blue     3    3     4     1
Yellow   2    2     3     1

For 10 labs

Code: Select all

        Red  Green Blue  Purple
White    9    7     -     2
Blue     6    5     8     1
Yellow   2    3     5     1

I pretty confident about the results for red and green packs. The results for blue and purple are almost certainly higher than they need to be, especially if you alternate topics that need blue and purple with topics that do not. That would allow production to catch up and refill any backlog space in your system. All that being said, I have not tested these numbers out yet in game.

I welcome any and all comment and criticism. I doubt very much that I didn't make any mistakes. I'm also aware that I've made a lot of assumptions (perhaps to the point of naivete) that may invalidate everything I've done.

EDIT: Put the tables into a spoiler, since they're wildly wrong.

[Xuhybrid]

Would be slightly better to display the data like this;

Code: Select all

        Lab   Red   Green Blue  Purple
White    5     5     4     -     1
Blue     5     3     3     4     1
Yellow   5     2     2     3     1
White    10    9     7     -     2
Blue     10    6     5     8     1
Yellow   10    2     3     5     1

So that the ratio is seen in line. Thanks for this btw.

[onebit]

One thing to consider is that if you don't start research immediately then there will be a backlog of science packs. I run 12 labs off 3 red/3 green/4blue.

[hassad]

After a bit of testing in game I've found my recommendations are wildly too high. Many thanks to onebit for shocking me into rechecking my analysis. The new spreadsheet can be found here, and is much more flexible. Also thanks to Xuhybrid for the better table format.

Corrected recommendations:

Code: Select all

+------+--------+-----+-------+------+--------+
| Labs |  Assm  | Red | Green | Blue | Purple |
+------+--------+-----+-------+------+--------+
|      | White  |   2 |   2   |   -  |    1   |
|   5  | Blue   |   2 |   2   |   3  |    1   |
|      | Yellow |   1 |   1   |   2  |    1   |
+------+--------+-----+-------+------+--------+
|      | White  |   4 |   4   |   -  |    1   |
|  10  | Blue   |   3 |   3   |   6  |    1   |
|      | Yellow |   2 |   2   |   4  |    1   |
+------+--------+-----+-------+------+--------+
|      | White  |   4 |   4   |   -  |    1   |
|  12  | Blue   |   3 |   3   |   7  |    1   |
|      | Yellow |   2 |   2   |   4  |    1   |
+------+--------+-----+-------+------+--------+

I still think that the blue production recommendation is too high, but the red and green production should be much closer to actual demand.

[waduk]

None of you guys using this ?
https://forums.factorio.com/forum/vie ... it=foreman
This works wonderfuly..

[hassad]

Wow, I had no idea that existed. Not sure how much I want to use it since I had just as much fun playing in Excel as I did in the game itself. Well, for anyone curious, the ideal ratio I've found is 1/30 packs per second per lab for all types of science packs.

[waduk]

Give it a try, i believe you will hooked when planning your next factory.
That wasn't just a tool. It was magic !

[hassad]

Indeed. After only a couple stabs at setups I'm already addicted.

[NAVAGOR]

I've experimented and found that the perfect ratio is 8 labs for every 5 red, basic beaker factories and every 6 green, basic factories.