Friday Facts #376 - Research and Technology

[mmmPI]

Hmm, ok, I've just made a spreadsheet to come up with the numbers, and it seems right, and it's now giving 40x as expensive - 30x if you're clever and put L5 prod mods in a few assemblers...

Please share ! :d

If you use multiple layers of crafting with quality then it gets muuuuuch cheaper - every layer you use quality on roughly halves the cost - and that's actually independant of the quality of the modules (provided you're using the same quality of module everywhere), though the power of L5 prod modules evens things up, so if you're at L5 you have no reason not to just use prods everywhere, then qual on the last layer with a recycle loop (presumably drone driven, as honestly at that point you don't even need to recycle that much anyway - a full third of level 1 ingredients get upgraded anyway).

I'm not sure i'm sharing that conclusion, if you use quality only at the last layer, you remove yourself the chances of getting the "rare" +2 +3 +4 quality upgrade. You only expect the +1. That sound like not the most optimal way, maybe it's just because i haven't seen the numbers.

Also you first multiply the quantity of base input material with productivity, and then reduce it at the last step with the recycler, forcing a much larger footprint throughout the whole production chain, and many recycling recipe/loop. Whereas utilizing quality and recycler at the early stage of the production, will net only the "densest" material early, and later down the chain, only use productivity to multiply the quantity of this high quality material.

I know those two points are contradicting with each other (you can't use prod in many of the "last step" and you can't use "quality" either if it's maxed out already with the input).

But they are "reasons" to do differently than full prod and quality only at the last layer which may apply in one game or another depending the settings and objectives.

[Qon]

I'm not sure i'm sharing that conclusion, if you use quality only at the last layer, you remove yourself the chances of getting the "rare" +2 +3 +4 quality upgrade. You only expect the +1. That sound like not the most optimal way, maybe it's just because i haven't seen the numbers.

Why wouldn't you be able to get +2 on the last product when recycling?

[mmmPI]

I'm not sure i'm sharing that conclusion, if you use quality only at the last layer, you remove yourself the chances of getting the "rare" +2 +3 +4 quality upgrade. You only expect the +1. That sound like not the most optimal way, maybe it's just because i haven't seen the numbers.

Why wouldn't you be able to get +2 on the last product when recycling?

I meant to say something that i realize now is not what was discussed. You only remove yourself some of those chances when using quality module at the last layer WHEN USING some quality input. Same for recycling a quality4 something because you tried to get the +5 and didn't . at best you get quality5 material that's a +1 only and can't get better. whereas recycling a Q2 green circuit can yield Q5 inputs.

[BicycleEater]

Please share ! :d

It would be my pleasure!
It's quite scrappy (and in ods form, which I'm sure will cause trouble with excel - tho google docs copes I think).
The interesting variables are over on the upper left, with "Qual of modules" being the interesting one - set to 0 for L1, 0.3 for L2, 0.6 for L3, 1 for L4, 1.5 for L5.
You can also customise the number of modules each assembler type uses - simply setting "assembler prod modules" for each column will calculate everything else - including reducing the number of qual modules to fit.
The "layers on input" bit is determining the total cost per L5 module when using full beacons at different layer depths.
The "Cost" column on the middle left ish is the cost of an L5 module in terms of lower cost modules.

If anyone sees any problems with the spreadsheet, please point them out - its a bit of a jumble.
Note that the reason it's giving 30 cost per L5 is that I'm proposing using 4 L5 MK3 prod modules on the assemblers making the L5 results from the L5 ingredients (as quality does nothing there, so why not...)

EDIT: I just realised the old version was kinda limited. I've changed it to allow you to choose the number of prod/qual modules seperately for each quality.

[BicycleEater]

Also you first multiply the quantity of base input material with productivity, and then reduce it at the last step with the recycler, forcing a much larger footprint throughout the whole production chain, and many recycling recipe/loop. Whereas utilizing quality and recycler at the early stage of the production, will net only the "densest" material early, and later down the chain, only use productivity to multiply the quantity of this high quality material.

I know those two points are contradicting with each other (you can't use prod in many of the "last step" and you can't use "quality" either if it's maxed out already with the input).

But they are "reasons" to do differently than full prod and quality only at the last layer which may apply in one game or another depending the settings and objectives.

Almost, but if you're using quality on a given layer, you can't use speed - since that reduces quality.
Hence prod will probably work out both cheaper (fewer modules as the assemblers with the modules will run faster), smaller, and higher throughput.
It's also usable with the rest of your base, so you can just tap mass produced products and use them.

[mmmPI]

Almost, but if you're using quality on a given layer, you can't use speed - since that reduces quality.
Hence prod will probably work out both cheaper (fewer modules as the assemblers with the modules will run faster), smaller, and higher throughput.
It's also usable with the rest of your base, so you can just tap mass produced products and use them.

Argh i had forgotten that, my argument was not correct, speed module in beacon + prod module would be more space efficient then just quality module in assembly. Maybe also more cost efficient on the long run , I suppose it depend on the exact cost of speed module & beacon as they make the initial investment higher. It'd be nice to have a spreadsheet to find out the exact cost to produce those high quality beacon and speed modules

If anyone sees any problems with the spreadsheet, please point them out - its a bit of a jumble.

I don't know if it's a problem but i found something puzzling :

1)change module quality to 1.5 ( cell T6)
2) change L1 assembly prod module from 0 to 4 alternatively(cell L8:O8)
3) Look at the cost (R17) not changing
4) change L1 assembly prod module from 0 or 4 to instead 2 or 3 alternatively(cell L8:P8)
5) Look ath the cost (R17) changing

0 or 4 module (L8 or M8 or N8 or O8 ) gives the same cost(R17) and 2(L8:O8) or 3(L8:O8) are giving a different cost(R17), that seem really weird no ? It would describe a situation in game where you either go full prod or full quality in one step is equally good but mixing them is bad because you have prod module at the end ? I would have expected that either quality or prod would be "best" and then going from 4 3 2 1 would be a unidirectionnal increase or decrease.

That oddity doesn't happen exactly the same if you don't use :

I'm proposing using 4 L5 MK3 prod modules on the assemblers making the L5 results from the L5 ingredients

If put less than 4 here. Then using 0 or 4 prod module (L8) doesn't yield the exact same result cost(R17), but i still find it weird that the worst (R17) cost is when using 2 prod module in (L8). It's not doing 01234 or 43210.

I choose some values and added a column "sum" to the layer on input to see the divisor of the right most column. To make it easier to see how much material is generated i thought and make a difference in the situations where the cost is the same.

QualityModules_Weird value 1.ods

(16.91 KiB) Downloaded 69 times

Now changing the previous figures shows a difference between them even if it's not the cost. I think it show the additionnal volume/quantity of material that one needs to shove into the recycling loop when using productivity early in the chain vs "normal". That would be 512 times if you have 9 layers of input and production everywhere since it's doubling everywhere. 2^9 = 512. It shows only 8,8170263313 if you only have 4 productivity module at the end and quality everywhere else. So the ratio between the 2 would be 64 times more material before the recycling loop ?

I wasn't able to find the 56 the devs announced from the spreadsheet. But i think that's because it doesn't model exactly the setup shown in the FFF where the input is green circuit of low quality, and the output is green circuit of high quality L5 but also L3 and L4 ( those are kept in a chest) AND low quality copper wire and iron plate ( those are seem discarded north) .

Whereas in the spreadsheet all material from the recycler are reused when not L5.( also in the FFF there is no L1 assembly ! ). Also in the FFF setup, the bottom recycler are having as input a mix of L1 and L2 quality whereas the top one only L1 not sure how that would play out if there are some different quality stacking allowed in the FFF's setup, or if not because of how inserter put item in recycler and their receipe.

There is a lot to take into account to try to find the 56 if that is the number from the setup or hinting at a rate calculator in game x)

It's also usable with the rest of your base, so you can just tap mass produced products and use them.

Yes that's a big positive for ease-of-use.

However, if your factory is multiplanetary, things may change. Extreme worst case, you can only ship 1 stack per travel, and the spaceship used is not re-usable and cost around 200 circuit L2 to be made. Doesn't feel good to use that spaceship to carry less than L2 circuits, as the cost of delivery would already be equal to payload. So 50% loss ? So much math x)

[BicycleEater]

...

Argh i had forgotten that, my argument was not correct, speed module in beacon + prod module would be more space efficient then just quality module in assembly. Maybe also more cost efficient on the long run , I suppose it depend on the exact cost of speed module & beacon as they make the initial investment higher. It'd be nice to have a spreadsheet to find out the exact cost to produce those high quality beacon and speed modules

Yeah, it's absolutely worth using speed modules, as each speed module pair on a beacon increases the amount of value you get from each productivity module. This is especially true if you're using L5 prod mods, as you can use L1 speed mods, and still quadruple (or more) the output from the L5 prod mods.

0 or 4 module (L8 or M8 or N8 or O8 ) gives the same cost(R17) and 2(L8:O8) or 3(L8:O8) are giving a different cost(R17), that seem really weird no ? It would describe a situation in game where you either go full prod or full quality in one step is equally good but mixing them is bad because you have prod module at the end ? I would have expected that either quality or prod would be "best" and then going from 4 3 2 1 would be a unidirectionnal increase or decrease.

Yeah, it is kinda odd, so maybe there's something wrong, but it is quite a complex system, so maybe it does do that?


Anyway, glad you found it interesting.

[factoriouzr]

I don't like the changes to make more recipes unlock by mining or crafting a certain number of them because it will be a pain for blueprints. It's already bad having to manage updating blueprints in mid to end game that can be built from the beginning due to tiers of assemblers, belts etc. Now with the quality system this will be exponentially worse. 5 tiers for every item and building will make updating and using bluerpints in new games a pain unless the blueprint system is updated to handle this elegantly and easily. I don't want to have 5 blueprints for the same setup. Plus another 3 for the tiers of assemblers, plus belts. In even the simplest cases, I don't want to have 5 blueprints for every blueprint that I make, one for each tier. This just increases complexity and annoyance. Now when you update a blueprint say mid to late game of a tier 4 assembler crafting something, you would have to manually build 4 more blueprints, make the same change to all of them. then re-blueprint every one of them again. If this isn't handled properly to the end user in blueprints, you will upset a lot of players and introduce complexity just for the sake of complexity.

Glad you finally decided to make the research queue always enabled. I asked for a feature to at least save your default preferences for research queue years ago but it was never implemented. We had to remember to always turn this on for every new game as there was no way to turn it on after. It never made sense to me that the developers wanted it disabled as you could just ignore it anyway and just research one item at a time. It just added complexity for the sake of complexity and no benefit.

[Axs1]

I have a question, how have they carried out the calculations to obtain that the cost of creating a legendary item is 56 times more, 116 according to a correction they made or

Hmm, ok, I've just made a spreadsheet to come up with the numbers, and it seems right, and it's now giving 40x as expensive - 30x if you're clever and put L5 prod mods in a few assemblers...

I have tried to do the calculations and it doesn't get me anywhere close to those numbers.

These are my values obtained (Quality modules - Legendary):

Q 2 ------- Q 3 -------- Q 4 --------- Q 5
3.48 ------ 11.00 ------ 34.42 ------ 106.51
3.48 ------ 10.37 ------ 30.59 ------ 89.09 -------------------- (I think these are the closest)

[BicycleEater]

I have a question, how have they carried out the calculations to obtain that the cost of creating a legendary item is 56 times more, 116 according to a correction they made or

Hmm, ok, I've just made a spreadsheet to come up with the numbers, and it seems right, and it's now giving 40x as expensive - 30x if you're clever and put L5 prod mods in a few assemblers...

I have tried to do the calculations and it doesn't get me anywhere close to those numbers.

These are my values obtained (Quality modules - Legendary):

Q 2 ------- Q 3 -------- Q 4 --------- Q 5
3.48 ------ 11.00 ------ 34.42 ------ 106.51
3.48 ------ 10.37 ------ 30.59 ------ 89.09 -------------------- (I think these are the closest)

I don't know how you got those numbers, but feel free to look at the spreadsheet.

[Axs1]

I have a question, how have they carried out the calculations to obtain that the cost of creating a legendary item is 56 times more, 116 according to a correction they made or

Hmm, ok, I've just made a spreadsheet to come up with the numbers, and it seems right, and it's now giving 40x as expensive - 30x if you're clever and put L5 prod mods in a few assemblers...

I have tried to do the calculations and it doesn't get me anywhere close to those numbers.

These are my values obtained (Quality modules - Legendary):

Q 2 ------- Q 3 -------- Q 4 --------- Q 5
3.48 ------ 11.00 ------ 34.42 ------ 106.51
3.48 ------ 10.37 ------ 30.59 ------ 89.09 -------------------- (I think these are the closest)

I don't know how you got those numbers, but feel free to look at the spreadsheet.

Here:

Cost.ods

(15.23 KiB) Downloaded 76 times

[BicycleEater]

Here:
Cost.ods

I don't really understand your spreadsheet.
What does the "Total" row represent?
- From the formula, it's the number of modules of a given level you'd get if you inputted one of each level of module?
What does the "% Level Up in 1 level" represent?
And what do the "Consecutive" rows do? I've had a look, but given my non-understanding of the above rows, I can't really reason through it.

One big reason we're getting different results is that we've made very different assumptions about how the bonus scales with quality. I've assumed that at max quality you get x level 2, x^2 level 3, x^3 level 4, x^4 level 5... which might be kinda dumb tbh, I'm not sure. I just used my spreadsheet to recalculate dividing quality by 10 each time, and I can get as low as 67 L1 ingredients per L5 module, by using a fair amount of productivity (2 in each assembler, and 4 in the L5 assembler).
Using just qual modules, I get 130 L1 ingredients per L5, or 97 if I use prod in the L5 assembler.

[Axs1]

Here:
Cost.ods

I don't really understand your spreadsheet.
What does the "Total" row represent?
- From the formula, it's the number of modules of a given level you'd get if you inputted one of each level of module?
What does the "% Level Up in 1 level" represent?
And what do the "Consecutive" rows do? I've had a look, but given my non-understanding of the above rows, I can't really reason through it.

One big reason we're getting different results is that we've made very different assumptions about how the bonus scales with quality. I've assumed that at max quality you get x level 2, x^2 level 3, x^3 level 4, x^4 level 5... which might be kinda dumb tbh, I'm not sure. I just used my spreadsheet to recalculate dividing quality by 10 each time, and I can get as low as 67 L1 ingredients per L5 module, by using a fair amount of productivity (2 in each assembler, and 4 in the L5 assembler).
Using just qual modules, I get 130 L1 ingredients per L5, or 97 if I use prod in the L5 assembler.

What does the "Total" row represent?
Total represents the sum of said row or column.

What does the "% Level Up in 1 level" represent?
"%" represents the probability of going up 1 level

And what do the "Consecutive" rows do?
Sorry, I forgot to change it, I were previously performing other calculations.

One big reason we're getting different results is that we've made very different assumptions about how the bonus scales with quality. I've assumed that at max quality you get x level 2, x^2 level 3, x^3 level 4, x^4 level 5... which might be kinda dumb tbh, I'm not sure. I just used my spreadsheet to recalculate dividing quality by 10 each time, and I can get as low as 67 L1 ingredients per L5 module, by using a fair amount of productivity (2 in each assembler, and 4 in the L5 assembler).

I understand, but I'm seeing that you are not considering the probability of going up two levels, for example, normal quality has a 10% chance of going up to level 2, 0.1% at level 3, 0.01% at level 4 and 0.001% at level 5
I had not considered that it could be that way

I have made the changes in my spreadsheet and I also based them on your probabilities. I just have one question, I still don't understand the recycling part you have in your spreadsheet.

Cost.ods

(5.53 KiB) Downloaded 67 times

[BicycleEater]

What does the "Total" row represent?
Total represents the sum of said row or column.

I know, the thing I'm curious about is why the sum over the column is meaningful - as it sums over values which don't correlate.

What does the "% Level Up in 1 level" represent?
"%" represents the probability of going up 1 level

I don't see how that can be higher than the "Level Up Ingredient"?

I have made the changes in my spreadsheet and I also based them on your probabilities. I just have one question, I still don't understand the recycling part you have in your spreadsheet.

The recycling section counts the output values of the recycler. If the recyclers and assemblers have the same modules, this is the same as the assembler version (but divided by 4), but I've also allowed for putting productivity modules into the assemblers, which gives better returns under some conditions. This means their output can be wildly different.

[Axs1]

I know, the thing I'm curious about is why the sum over the column is meaningful - as it sums over values which don't correlate.

The sum represents the probability that the item will level up regardless of how many levels it raises, it is to know what percentage does not level up.

I don't see how that can be higher than the "Level Up Ingredient"?

That is only the part of the recycler, to know what percentage increases in level, it is related to the probabilities of what was mentioned above and the % Level Up in 1 level represent the % that specifically goes up 1 level. the values are low because I am representing the ingredients as a part of the product that is recycled

The recycling section counts the output values of the recycler. If the recyclers and assemblers have the same modules, this is the same as the assembler version (but divided by 4), but I've also allowed for putting productivity modules into the assemblers, which gives better returns under some conditions. This means their output can be wildly different.

I also did that part, just applying productivity in the recycler and it gives better results, only that to determine from which quality of productivity and quality give better results, it depends on the way in which the costs were made.

Share my spreadsheet on the cost of getting premium quality.

You can find the results for:

• Assembler with Quality and Recycler with Quality

• Assembler with Quality and Recycler with Quality(Old Version)

• Assembler with Quality and Recycler with Productivity

• Assembler with Quality and Recycler without modules

Spreadsheet

Quality - Cost.ods

(8.84 KiB) Downloaded 66 times

[BicycleEater]

I know, the thing I'm curious about is why the sum over the column is meaningful - as it sums over values which don't correlate.

The sum represents the probability that the item will level up regardless of how many levels it raises, it is to know what percentage does not level up.

Sorry about the short reply, its really late here, but the sum over the horizontal is the probability it levels up at all. What is the sum over the verticle - that is C19-F19?

[Axs1]

I know, the thing I'm curious about is why the sum over the column is meaningful - as it sums over values which don't correlate.

The sum represents the probability that the item will level up regardless of how many levels it raises, it is to know what percentage does not level up.

Sorry about the short reply, its really late here, but the sum over the horizontal is the probability it levels up at all. What is the sum over the verticle - that is C19-F19?

These are the probabilities that each level has, but although I think that putting them together as one is not correct.

I'm still working on it.

I have a new idea on how to calculate the cost and it seems to me that it is more accurate or it could also be the correct way to calculate the cost.

I'll share it when I finish it.

Edit:

This is the new spreadsheet, here it contains the new method and also the old ones. The equations used in the new method are included.

Quality - Cost.ods

(29.85 KiB) Downloaded 65 times

Note: if you have any questions let me know

[Axs1]

I know, the thing I'm curious about is why the sum over the column is meaningful - as it sums over values which don't correlate.

The sum represents the probability that the item will level up regardless of how many levels it raises, it is to know what percentage does not level up.

Sorry about the short reply, its really late here, but the sum over the horizontal is the probability it levels up at all. What is the sum over the verticle - that is C19-F19?

These are the probabilities that each level has, but although I think that putting them together as one is not correct.

I'm still working on it.

I have a new idea on how to calculate the cost and it seems to me that it is more accurate or it could also be the correct way to calculate the cost.

I'll share it when I finish it.

Edit:

This is the new spreadsheet, here it contains the new method and also the old ones. The equations used in the new method are included.

snip

Note: if you have any questions let me know

I have made several corrections to the spreadsheet, I have also changed the probabilities according to the data provided by the wiki: https://wiki.factorio.com/Quality

Here:

Quality - Cost.ods

(27.21 KiB) Downloaded 70 times

[Dimswit]

The triggers should be like picking expensive mode from normal. I dislike the triggers, some of us don't like it. Quality sounds good + modules though.

[Axs1]

snip

Sorry for posting the spreadsheet so many times, I have made changes that maybe someone will be interested in.

Now in the spreadsheet you can enter the number of quality or productivity modules for each assembler or recycler (recyclers do not support productivity, but I included it anyway)

Spreadsheet

Quality - Cost.ods

(9.38 KiB) Downloaded 71 times

Edit:
If you place 12 productivity modules (this is done to simulate productivity research and 12, because the productivity limit is 300%) and 4 quality modules in the assembler, you obtain that the cost, only for the items that can be apply productivity, from normal quality to legendary quality is 4.12

In conclusion, for the late game it is better to always make legendary items. I wouldn't say all of them, but the science packs for sure.