Stating a preference is not attacking any one person. I think my choice of questioning Fulgora's benefits wasn't taken well. Hot's hot, so maybe my defense was not good.
You seem like you want people to share in your excitement. That's fine. It's better than being angry but you have to have an outlet for that too. It can come off as the same kind of energy. And someone who doesn't have a handle on their temper at a moment can come to meet you with as much force as they think they feel. It's not the right choice but it is a choice.
The response I'm getting from you could be seen in that way. Regardless of the context or order. And someone has a right to defend their opinion; to a point. I'm not going to defend what was discussed (or at least not at here). I feel like we weren't going in a good direction, I backed off.
Okay you're not mad. Quality presents you with a wide range of options. Looking at all those options is fundamentally exciting and dazzling.h.q. droid put me on a good track. I have a set of possible moves that include asteroid crushing, asteroid rerolling, em plants and a few things with boxes. I might start as wrong and contradictory but I'm getting options for a table of basic plays and moves.
Which is what you suggest to do.
I just doesn't make much sense to attack that unless you have a lot of energy and you need to vent some of it.
I'm happy to not argue with the numbers I'm getting, especially since I've got someone who has a math model that's producing consistent results.
That's not a process that is going to have the simulator come out looking like a totally consistent person. You're attacking me for doing what you proposed though.
I've played with the numbers. I've felt that contradictory energy from the numbers. Whatever else I can say I'm plenty confused.
We've got box crushing, some stuff on asteroids and em plants. I think quality presents problems that have solutions which are fundamentally contradictory. I think it's fundamentally exciting. Please don't mistake my reaction for shock or a lack of respect. You're good.
Biolabs are my standing case of "why taking an opinion driven approach" to quality doesn't work; and why you're always going to have that contradictory energy. But why what you say about having a table is the safest thing. I think I can also justify why that table needs a "rate" entry so someone can be confident in their choices.
I think we can have an interesting conversation about what we should go after next is in terms of module choice or material choice. I think we can have a conversation about why you need to approach this with an suitably sized set of options. In case you get into a conversation with someone who is saying "this is best" and insisting on one option. I might be a donkey but I can see eye to eye with you on that.
How to upcycle?
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
That's not about taking it well or not , i thought it made no sense to claim that the "mining bonus" is what should drive anyone to upcycle in Vulcanus or makes it "the best place" and to present your method as "the best" without any context and what you consider "the best". That's was borderline trolling imo as you did so quoting a post emphasizing the futility of searching for the best setup without defining such concept first.coffee-factorio wrote: Fri Jun 13, 2025 11:03 pm I think my choice of questioning Fulgora's benefits wasn't taken well. Hot's hot, so maybe my defense was not good.
I might be a donkey but I can see eye to eye with you on that.
I tried to understand why you'd say so and you added that Fulgora is surface limited which is unrelated and also no different than Vulcanus late game, so wrong, eventually you called me angry and backed your claim with what you called a monte carlo simulation on some other unrelated thing which wasn't one you admitted it ....
It would be nice if you could be a donkey without calling the others the same thank you.
I stand by my advices that the setup you should consider only after considering your goals and objectives and situation and that there is no generic answer as to which is the best that can be simplified and would always work, because players have different expectations for when something "works" or different playstyle, or what is "fast" or "big".
Since the first step of the reasonning seem to be problematic in your particular case, as mentionned there's no way for me to provide any help beyond. If you think i'm saying something wrong please feel free to point out the mistake in what appears to me comon sense. But when i inquiry "what is the best setup for you ?" if you answer "this one according to my monte carlo simulation" you have failed to understand the question unfortunately in my view. And if you insist your legendary setup is the "best" to make legendary material, i am going to point out that's it's not very serious to propose a setup full of legendary when the question is "how do i get them in the place ?" for many players.
Just in case in can help for more clarity, i was inspired this comparaison with the famous potato again. I feel like you're arguing about the method to measure something, like a potato, because you think the bigger number is the better potato, but your measurement method is wrong because it's only the circumference you're measuring, and you're claiming you've find the "tallest". Even if the measurement method was correct in the practice it would still be inappropriate to use to conclude about "the best potato" , and i feel like your answers are along the line of "ok let's forget about that and let's measure another one". So that's why i'm saying it appears to be a problem in the early step of reasonning that prevent me from providing help in that direction.
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
And regardless of the merits or why I'm sorry our conversation went the way it did.mmmPI wrote: Fri Jun 13, 2025 11:42 pmThat's not about taking it well or not , i thought it made no sense to claim that the "mining bonus" is what should drive anyone to upcycle in Vulcanus or makes it "the best place" and to present your method as "the best" without any context and what you consider "the best". That's was borderline trolling imo as you did so quoting a post emphasizing the futility of searching for the best setup without defining such concept first.coffee-factorio wrote: Fri Jun 13, 2025 11:03 pm I think my choice of questioning Fulgora's benefits wasn't taken well. Hot's hot, so maybe my defense was not good.
I might be a donkey but I can see eye to eye with you on that.
I tried to understand why you'd say so and you added that Fulgora is surface limited which is unrelated and also no different than Vulcanus late game, so wrong, eventually you called me angry and backed your claim with what you called a monte carlo simulation on some other unrelated thing which wasn't one you admitted it ....
It would be nice if you could be a donkey without calling the others the same thank you.
I stand by my advices that the setup you should consider only after considering your goals and objectives and situation and that there is no generic answer as to which is the best that can be simplified and would always work, because players have different expectations for when something "works" or different playstyle, or what is "fast" or "big".
Since the first step of the reasonning seem to be problematic in your particular case, as mentionned there's no way for me to provide any help beyond. If you think i'm saying something wrong please feel free to point out the mistake in what appears to me comon sense. But when i inquiry "what is the best setup for you ?" if you answer "this one according to my monte carlo simulation" you have failed to understand the question unfortunately in my view. And if you insist your legendary setup is the "best" to make legendary material, i am going to point out that's it's not very serious to propose a setup full of legendary when the question is "how do i get them in the place ?" for many players.
Just in case in can help for more clarity, i was inspired this comparaison with the famous potato again. I feel like you're arguing about the method to measure something, like a potato, because you think the bigger number is the better potato, but your measurement method is wrong because it's only the circumference you're measuring, and you're claiming you've find the "tallest". Even if the measurement method was correct in the practice it would still be inappropriate to use to conclude about "the best potato" , and i feel like your answers are along the line of "ok let's forget about that and let's measure another one". So that's why i'm saying it appears to be a problem in the early step of reasonning that prevent me from providing help in that direction.
The assumption that I was admitting anything by saying that I'm using the map editor is not accurate. https://en.wikipedia.org/wiki/Buffon%27s_needle_problem You don't need a computer to perform a Monte Carlo style simulation. A person in the 18th century dropped pins and figured out the constant PI to a reasonable amount of digits. I'm calling what I do in a map editor a Monte Carlo simulation because as far as I understand it I can throw a bunch of needles at a piece of paper and get a result of increasing accuracy.
My setups aren't the best. Where I'm from production means a something a bit different. I'm not putting up production setups. Right now what I'm doing with h.q. droid is figuring out how certain kinds of operations perform. My theory is not biggest or best. My theory is at a 15 or a 240/s rate, I can touch N*1000 samples and be within 10%, but 240 is going to be more convenient.
If I understand the rate of machine speed. Then I can total the number of normal inputs into a machine. If I set up a production line on a surface, disable enemies (biters love things that run for hours), I use iron at 240. I get out 27774 after 10 hours at 64 speed. 3600 seconds * 10 = 36000. And 240*36000 is... a number I'm copy and pasting from a calculator.
Droids' program is accurate, as far as I can tell. But I've got no idea what he's using. Wherever he's getting his numbers from they match mine. Including the one where I make a foundational mistake.
If you propose that this is not a method, I can give a blueprint that anyone can copy and paste into a map editor and run using gears that will produce the same outputs given the same inputs.
Droid can explain what they're doing.
There where a number of limitations to what I was doing that I discussed. Those are the classic limitations of Monte Carlo style simulations. If I didn't honestly discuss those limitations. Then I probably couldn't have a discussion with droid. Well, they'd get to say my math was faulty a few more times if they thought it was worth the effort.
Now if you don't like that approach, I can drop that for a minute to show why you need a list of methods and not to engage the problem with "and this method is best" approach so you know more or less I'm not limiting a persons options. I can also show you one of the ways that quality creates a situation which causes a person to be inherently contradictory. Biolabs is a solidly defined problem where we can see that you are very clearly right.
I think it's best if I say why you are right for a moment?
Re: How to upcycle?
I'm sorry you think it's a conversation. I think it's best if you adress personnal message in private.coffee-factorio wrote: Sat Jun 14, 2025 2:33 am And regardless of the merits or why I'm sorry our conversation went the way it did.
I think it's best if I say why you are right for a moment?
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
That's fair. I did send you one. Someone else probably might talk about Monte Carlo and it's strengths and weaknesses when you're given a problem you're using it to look at quality. It isn't being inconsistent to admit there are tradeoffs that make one better or another worse. I plan on continuing my conversation with droid at some point.
Re: How to upcycle?
I believe if you want to have a conversation with a certain individual so you can learn more on the math concept you used the name inapropriatly you don't have to quote me and say things that are wrong and grotesque on " upcycling" in the gameplay help section, that's what private messages are for.
You're being defensive when your wrong claim are corrected, so you don't look like you're willing to receive help. And you're not providing any, as such you could also open a thread on general discussion instead of polluting a thread with incorrect statements that could be confusing for players just trying to upcycle their things
You're being defensive when your wrong claim are corrected, so you don't look like you're willing to receive help. And you're not providing any, as such you could also open a thread on general discussion instead of polluting a thread with incorrect statements that could be confusing for players just trying to upcycle their things
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
Had a conversation with someone out by the flag pole. I'll respond to their criticisms at my leisure when they are valid.
This went a bit long, so the tldr summary is that quality produces deep results that are both technically intense and philosophically deep. At the end of the day it does influence how to do the thing at a beginner-novice level.
Now we have this particular issue and it's a simple puzzle for the beginner. However, it's absolutely good fun. I will explain how a Monte Carlo method/Buffon's Needle style simulation can be put together for the intermediate and maybe some of the difficulties you face later.
However I do believe quality is played in different ways by different people.
Now, I am going to play philosopher with quality, they upcycle different. This is how they do it:
With a collection of skills they demonstrate a point.
I have a build in mind to demonstrate how quality presents people with deeper problems than can be solved by generic solutions. This is actually a bit contradictory to what experience is teaching me and theory suggests:there's common cases and solutions within reach of a beginner. But there's not a single "How to/best" approach. I have an infinite input, and infinite sink outputs. This all goes to a chemical plant station. This build tests a feed mechanism a beginner can use on a space platform. And the production graphs produced from testing will show that it has contradictory qualities from that light.
The feed mechanism is an unguarded sushi feed. So a river of parts hits the belt and that river is drawn from. For the purposes of this test, four assemblers and parts at best quality where used to eliminate any question that the result would be affected by throughput rate. An issue with that logic was noted that caused the production of a fifth build, which is operating as I type this. A bulk inserter swings for each item it can see, so it might miss an item on a belt. Therefore a build with four inserters per station was produced. This build consumes all parts made by one station.
To be used for production purposes, the infinite source needs to be replaced with source of parts. The infinite sink may be replaced with a recycle loop. But in doing this we eliminate two sources of irrelevant questions.
Four surfaces where set up and used to eliminate any question of whether or not the result could be reproduced. Due to operator error, three surfaces had biters on them which will show up as a sharp dip in the production graph. That's me removing them and resetting the experiment.
Chemical plants was selected as the assembly recipe for a number of reasons: it has an even parts distribution (this is important to the philosopher). It builds fast, so even at a large time scale it produces a result in a timely fashion. One assembler produces chemical plants. Four recyclers are used, to prove the point and suggest further exploration. For expediency, max tier quality modules are used.
The blueprint is provided for the amusement of Seablock players and independent verification.
The 2nd variant is here.
Four production graphs produced the following result: at random times over a 50 hours long, every single one of them halted; but they would not halt in so short a time as to discourage a beginner. Graphics are at the end of the post behind a spoiler tag since they are screen shots, and rather large, as well as a shot of a chemical plant from the shortest run with its production number.
The quantity of normal chemical plants on those graphs shows why this is of no hazard to a beginner:
Quantities at the point of the jam are: 44k, 70k, 81K, 101k. The worst recorded rate droid and I produced was 1 in 2400. The actual rate is going to be something like 65 in 10000 (.65%/100 if I've done my math right). If all you're doing in the time it takes to run the build is chemical plants you've still got low hundreds of them.
When I looked at the timescale of the graph from the point where I removed the biters on the shortest one, time seemed to be at a day of playing time.
It might seem odd that 0.5 ips legendary buildings produce that result. The read on the building is ~38K items produced. So 10 hours should be the run time. A single legendary bulk inserter was used. Bulk inserters handle only one kind of items at a time. This means that it must grab a stack of an individual item, of the total required.
However a legendary inserter moves 2160 degrees per second. This divided by 360 is 6. Assuming it rotates half a circle forward and then back, there might be animation that changes this rate somewhat and there is. However, you make only a half a chemical plant per second output on this build. So the inserter gets 10 shots to fill four slots with a hand 12 deep.
As I was doing a technical review of this before posting, I thought that a part might slip past a single inserter on a sushi belt and I have a hard standard for myself. I might argue that caused the jam if that I made a beginner mistake. So the aux build has four legendary arms and they absorb every part at quality a building can receive at a given time. The second building in the auxiliary build should rarely build an entity. In 70K parts only 1 chemical plant was built by it.
This is because as the thing slowly jams, a recycler will fill the second side of the sushi belt up. One belt will be stuck, but there's a 1 in... very large and very random chance that the 2nd building magically be filled and produce a second. This produces a practically interesting result. It's the stuff of madness but 14 in 70K uncommon chemical plants where made on the 2nd building. Five rare and 2 epic where also produced. And on this build, only the remanufacture step is examined. The first buildings all show it.
Alright, this might be something that can happen if this result applies generally to entities that take four inventory slots:
Beacons will be fine. Biochambers will be fine. Okay, bulk inserters use the same build they are stopped dead in their tracks. And two hours into making a circuit or something to fix it you look up and chemical plants are stopped too. But beacons are still running.
I'm always a beginner at something new. This belt issue will be corrected and someday I'll be able to say I'm okay at this game.
A hypothetical operator with a similar mindset will use the second variation on all four builds.
75 hours in everything jammed on beacons. Bulk inserters is fine. Chemical plants and biochambers are fine.
At 125 hours you fancy you have an instinct for how fast something should go though and biochambers "feels" off. Like it hasn't made enough stuff. You've had that wrong feeling for awhile. And you've read about peasant uprisings and your turrets are working. After all, you have this build operating with moats the pentapods can't reach over. You get alerts. You have a damage output that indicates they've shot several pentapods.
This is the sort of thing that make a person sound a bit contradictory, and take up mathematical models in their hobby time. You see a lot of odd things by hour 750 if you go that far.
For a single player that's pretty fun. This must be great to discuss on multiplayer servers. That might sound like sarcasm. Actually, if you have the right group of excited people with you, no one will blame each other because... I mean it's one belt.
If you put quality on a space platform you'll run into a variation of this problem quickly. Space limited (fun challenge), resource rich except for petrol-chem but at rate-limited infinite, cost to ship parts is zero. Needs chemical plants, only way to make artificial coal. mmmPi is right about being able to deliver an entire volume of coal from orbit. It's also a produces a volume of rocket fire. There's a few other "unique to chemical plants" recipes which mean you can't simply use a substitute.
Space limited applies to the physical sense. But it also applies to to the tools being used. This is a fully stacked belt, jammed by being filled on both sides. That means it has about a third of an inventory slot if everything was full... which it won't be. Because the recycler doesn't have a memory.
In terms of solving this problem, it might be a manual reset (you tear apart the belt and place the inventory on it somewhere else, then put down) and an alarm. You can at least get the buildings if this is a "sometimes over hours" event. If you hook this up on a space platform that's instant, on the ground that's a race between you and the robots. Which works for a beginner setup, if you've put hours into this game you know you want to automate.
But now instead of playing space engineer you get to play space manager and ask some interesting questions.
Would this always happen to the space engineer regardless of the storage system they used? Because if this wasn't happening randomly, it could be something trained around. NineNine's build at the start of this thread uses chests. But I see something there where they might have run into something like this and solved it.
Would this phenomenon be unique to chemical plants? In short can we safely swap them out for a different entity, as coffee is doing?
Why would biochambers probably take a little longer to jam if an otherwise technically experienced player had a perfectly functioning build without being aware of this problem?
If it does generalize it probably generalizes to parts with similar inventories... would it generalize to all parts inventories and how could you demonstrate that by playing philosopher at 64x speed?
These questions might seem irrelevant. Well, someone might get discouraged from seeing a stopped belt if all it is, is getting enough resources. Without a rate table they could verify, they might not have a quantitative metric of how much work they where doing. If they do not realize it is random, they might leave this system entirely instead of looking for a simple engineering solution for its problems. Someone might think they need to have an understanding of matrix multiplication they can't get, because of work-life balance. And that isn't true.
At 50 hours to fail, this also suggests that if you want to use a brute force simulation to check a "quality calculators" numbers or someone telling you something in a chat or on a forum... it doesn't necessarily have to be a perfect build. But at that scale it also suggests that you should tread with a bit of care, because that's 44 minutes gone for a number. A deep problem can help someone grow and do something they're afraid of normally.
This went a bit long, so the tldr summary is that quality produces deep results that are both technically intense and philosophically deep. At the end of the day it does influence how to do the thing at a beginner-novice level.
Now we have this particular issue and it's a simple puzzle for the beginner. However, it's absolutely good fun. I will explain how a Monte Carlo method/Buffon's Needle style simulation can be put together for the intermediate and maybe some of the difficulties you face later.
However I do believe quality is played in different ways by different people.
Now, I am going to play philosopher with quality, they upcycle different. This is how they do it:
With a collection of skills they demonstrate a point.
I have a build in mind to demonstrate how quality presents people with deeper problems than can be solved by generic solutions. This is actually a bit contradictory to what experience is teaching me and theory suggests:there's common cases and solutions within reach of a beginner. But there's not a single "How to/best" approach. I have an infinite input, and infinite sink outputs. This all goes to a chemical plant station. This build tests a feed mechanism a beginner can use on a space platform. And the production graphs produced from testing will show that it has contradictory qualities from that light.
The feed mechanism is an unguarded sushi feed. So a river of parts hits the belt and that river is drawn from. For the purposes of this test, four assemblers and parts at best quality where used to eliminate any question that the result would be affected by throughput rate. An issue with that logic was noted that caused the production of a fifth build, which is operating as I type this. A bulk inserter swings for each item it can see, so it might miss an item on a belt. Therefore a build with four inserters per station was produced. This build consumes all parts made by one station.
To be used for production purposes, the infinite source needs to be replaced with source of parts. The infinite sink may be replaced with a recycle loop. But in doing this we eliminate two sources of irrelevant questions.
Four surfaces where set up and used to eliminate any question of whether or not the result could be reproduced. Due to operator error, three surfaces had biters on them which will show up as a sharp dip in the production graph. That's me removing them and resetting the experiment.
Chemical plants was selected as the assembly recipe for a number of reasons: it has an even parts distribution (this is important to the philosopher). It builds fast, so even at a large time scale it produces a result in a timely fashion. One assembler produces chemical plants. Four recyclers are used, to prove the point and suggest further exploration. For expediency, max tier quality modules are used.
The blueprint is provided for the amusement of Seablock players and independent verification.
The 2nd variant is here.
Four production graphs produced the following result: at random times over a 50 hours long, every single one of them halted; but they would not halt in so short a time as to discourage a beginner. Graphics are at the end of the post behind a spoiler tag since they are screen shots, and rather large, as well as a shot of a chemical plant from the shortest run with its production number.
The quantity of normal chemical plants on those graphs shows why this is of no hazard to a beginner:
Quantities at the point of the jam are: 44k, 70k, 81K, 101k. The worst recorded rate droid and I produced was 1 in 2400. The actual rate is going to be something like 65 in 10000 (.65%/100 if I've done my math right). If all you're doing in the time it takes to run the build is chemical plants you've still got low hundreds of them.
When I looked at the timescale of the graph from the point where I removed the biters on the shortest one, time seemed to be at a day of playing time.
It might seem odd that 0.5 ips legendary buildings produce that result. The read on the building is ~38K items produced. So 10 hours should be the run time. A single legendary bulk inserter was used. Bulk inserters handle only one kind of items at a time. This means that it must grab a stack of an individual item, of the total required.
However a legendary inserter moves 2160 degrees per second. This divided by 360 is 6. Assuming it rotates half a circle forward and then back, there might be animation that changes this rate somewhat and there is. However, you make only a half a chemical plant per second output on this build. So the inserter gets 10 shots to fill four slots with a hand 12 deep.
As I was doing a technical review of this before posting, I thought that a part might slip past a single inserter on a sushi belt and I have a hard standard for myself. I might argue that caused the jam if that I made a beginner mistake. So the aux build has four legendary arms and they absorb every part at quality a building can receive at a given time. The second building in the auxiliary build should rarely build an entity. In 70K parts only 1 chemical plant was built by it.
This is because as the thing slowly jams, a recycler will fill the second side of the sushi belt up. One belt will be stuck, but there's a 1 in... very large and very random chance that the 2nd building magically be filled and produce a second. This produces a practically interesting result. It's the stuff of madness but 14 in 70K uncommon chemical plants where made on the 2nd building. Five rare and 2 epic where also produced. And on this build, only the remanufacture step is examined. The first buildings all show it.
Alright, this might be something that can happen if this result applies generally to entities that take four inventory slots:
Beacons will be fine. Biochambers will be fine. Okay, bulk inserters use the same build they are stopped dead in their tracks. And two hours into making a circuit or something to fix it you look up and chemical plants are stopped too. But beacons are still running.
I'm always a beginner at something new. This belt issue will be corrected and someday I'll be able to say I'm okay at this game.
A hypothetical operator with a similar mindset will use the second variation on all four builds.
75 hours in everything jammed on beacons. Bulk inserters is fine. Chemical plants and biochambers are fine.
At 125 hours you fancy you have an instinct for how fast something should go though and biochambers "feels" off. Like it hasn't made enough stuff. You've had that wrong feeling for awhile. And you've read about peasant uprisings and your turrets are working. After all, you have this build operating with moats the pentapods can't reach over. You get alerts. You have a damage output that indicates they've shot several pentapods.
This is the sort of thing that make a person sound a bit contradictory, and take up mathematical models in their hobby time. You see a lot of odd things by hour 750 if you go that far.
For a single player that's pretty fun. This must be great to discuss on multiplayer servers. That might sound like sarcasm. Actually, if you have the right group of excited people with you, no one will blame each other because... I mean it's one belt.
If you put quality on a space platform you'll run into a variation of this problem quickly. Space limited (fun challenge), resource rich except for petrol-chem but at rate-limited infinite, cost to ship parts is zero. Needs chemical plants, only way to make artificial coal. mmmPi is right about being able to deliver an entire volume of coal from orbit. It's also a produces a volume of rocket fire. There's a few other "unique to chemical plants" recipes which mean you can't simply use a substitute.
Space limited applies to the physical sense. But it also applies to to the tools being used. This is a fully stacked belt, jammed by being filled on both sides. That means it has about a third of an inventory slot if everything was full... which it won't be. Because the recycler doesn't have a memory.
In terms of solving this problem, it might be a manual reset (you tear apart the belt and place the inventory on it somewhere else, then put down) and an alarm. You can at least get the buildings if this is a "sometimes over hours" event. If you hook this up on a space platform that's instant, on the ground that's a race between you and the robots. Which works for a beginner setup, if you've put hours into this game you know you want to automate.
But now instead of playing space engineer you get to play space manager and ask some interesting questions.
Would this always happen to the space engineer regardless of the storage system they used? Because if this wasn't happening randomly, it could be something trained around. NineNine's build at the start of this thread uses chests. But I see something there where they might have run into something like this and solved it.
Would this phenomenon be unique to chemical plants? In short can we safely swap them out for a different entity, as coffee is doing?
Why would biochambers probably take a little longer to jam if an otherwise technically experienced player had a perfectly functioning build without being aware of this problem?
If it does generalize it probably generalizes to parts with similar inventories... would it generalize to all parts inventories and how could you demonstrate that by playing philosopher at 64x speed?
These questions might seem irrelevant. Well, someone might get discouraged from seeing a stopped belt if all it is, is getting enough resources. Without a rate table they could verify, they might not have a quantitative metric of how much work they where doing. If they do not realize it is random, they might leave this system entirely instead of looking for a simple engineering solution for its problems. Someone might think they need to have an understanding of matrix multiplication they can't get, because of work-life balance. And that isn't true.
At 50 hours to fail, this also suggests that if you want to use a brute force simulation to check a "quality calculators" numbers or someone telling you something in a chat or on a forum... it doesn't necessarily have to be a perfect build. But at that scale it also suggests that you should tread with a bit of care, because that's 44 minutes gone for a number. A deep problem can help someone grow and do something they're afraid of normally.
- Screenshot 2025-06-15 223807.jpg (412.74 KiB) Viewed 642 times
- Screenshot 2025-06-15 223523.jpg (492.87 KiB) Viewed 642 times
- Screenshot 2025-06-15 223322.jpg (505.62 KiB) Viewed 642 times
- Screenshot 2025-06-15 223049.jpg (435.5 KiB) Viewed 642 times
- Screenshot 2025-06-15 222137.jpg (303.96 KiB) Viewed 642 times
- aux-01-11_48zulu.jpg (477.18 KiB) Viewed 642 times
- aux-02-12-19zulu.jpg (401.32 KiB) Viewed 642 times
- aux-03-12-24zulu.jpg (435.41 KiB) Viewed 642 times
Re: How to upcycle?
That's not true you, you ignore most and just repeat the same 'wrong' thing. It is quite fairly obvious to me that you have made a setup that's functionnaly the same as the second post on this thread. But you made mistake when building it and added a lot of misconception, and improper math terminology that should be ignored in the first place. Which is again very close to trolling imo considering how was introduced the setup from which you made a copy :coffee-factorio wrote: Mon Jun 16, 2025 9:04 am Had a conversation with someone out by the flag pole. I'll respond to their criticisms at my leisure when they are valid.
reference
There is indeed no math needed to realize this method "can" work, albeit when it's built properly, which doesn't seem to be the case in your blueprint where the belt is clogged, and the input are not making sense, unlike in the setup linked.I think this is a step back in the discussion because it was discussed quite largely by now that this constitute only a single method to do. ( when properly implemented) It sounds like you are on purpose repeating the same wrong reasonning when presenting a working setup in the worst way possible when describing it as "generic" and avoiding to show any sign of the slightest understanding of any of the critics that have been raised. It sounds to me like you're doing it on purpose to make the same grotesque claim everytime so i feel like i have to repeat and over myself something that anyone reading the discussion would have realized by now, but it may need to be precised again as visibly you are not having it understood and a partial reading of such ridiculous assertions of yours, if left unchallenged because of the tedium to interact with roleplaying donkey, may be akin too receiving wrong or misleading informations when presented the way you did.
What you presented is a poorly built "dedicated loop". Presented out of any logical scenario to me, why would you have 3 level of different quality of chemical plant as input ? how does that makes sense in any real game to build a setup designed to handle this ? Why mix the intermediate and risk clogging ? It's non-sense compared to the first "dedicated loop" that was shown imo, it's a misleading example you provided. for more than the reason of the clogging.mmmPI wrote: Sun May 11, 2025 12:54 am
- 1)At least the way i understand it, if you want certain specific items like substation or beacons, it's easier to make dedicated loop for those items.
- 2)If you want a full spaceship or a full armor, you want a bit of many different material, but not in mass production, you don't really need recycler you can have a few % of your machines dedicated to this, or a specific small quality subfactory switching receipe that backs up when nothing's needed.
- 3)If you want most everything in high quality, you will probably use different methods that are only available/making sense with advanced infinite research in productivities area and try to have the raw ressources of high quality for fully legendary production lane.
Conclusion :
How to upcycle depend a lot on what you want to have and when, the mass production method are only available late game, they involve producing raw ressources at high quality. Some players consider the game as "finished" by then , others the game as "only starting". Thus it's hard to advise a particular setup without more information on "what do you need ?" and "when ?". Hopefully i provided explanations on the second part for why i think it's important to answer those question before thinking of the actual design, beause there are options to choose from that would work to different degree. I think having an idea of the different ways that exist allow to plan better the strategy over the course of a full game, to only focus on quality when necessary without over investement or postponing it forever, but it's a delicate balance to anticipate our own needs during a game and choose the design that match best.
And in case you're wondering, if you throw 3 coins in the air and it gives you 3 tails and 0 head, you can claim your monte carlo simulation tells you the coins as 100% chance of tails and 0 of heads, you'd be wrong the same way you are when explaning you did a monte carlo simulation by looking at game for a few minutes, because in both case amongst other things you show no mathematical correlation between the amount of sample and the meaningfullness of the result. That's then not a monte carlo simulation. It's just "looking at it a lot and taking the average while eyeballing how much is 'a lot'". It's better to call it this way i think because it's not using complicated term inappropriatly, which confuses those who never heard of the thing and provoke
or facepalm otherwise.Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
Now, some statements have been made.
The relevant ones are that a beginner is insulated from error due to the time scales involved. And a more curious mind will observe some interesting phenomenon which makes the problem seem a bit deeper. And at 50 hours in, well...
This is where a discussion of engineering tolerance comes in. A carpenter, a game developer and a mathematician will all use a constant number. The carpenter and the game dev. both need to make decisions on the basis of a table. A carpenter is working in an environment where 2-4 digits give a close enough value.
A game dev. usually needs a program to run in time. But they might be simulating physics, so they need a bit more digits. A mathematician wins their game if they come up with another method with a novel property or demonstrate the capability of the state of the art. They can throw a super computer and a lot of human time at just getting the constant.
So from the point of view of engineering tolerance. 50 hours and several hundred machines built at legendary quality justified me looking over at this feed mechanism on an automall, tearing the belt up and going. "Oooo. Tasty. Spicey. ". By the time it jammed it'd redone my engines, beacons and crushers.
It's tricky. The map editor gives me the capability to act without showing some care and tolerance by simply copy and pasting another's work in every detail.
Without isolating the build I cannot isolate an interesting fact about the build being shown. I could also call into question NineNine's workmanship. I'm a shrewd man. Using infinite sinks removes the option of saying that this issue, if it even could be replicated in their system, was due to operator error on the input or output side of build.
From the point of view of being technically flawless it halts so it is not a beginners build. From the point of view of completing operations a beginner should use it for it is an excellent build and a fantastic learning experience and they should continue. That's fundamentally inconsistent. From the point of view of being relevant, you'll be wanting those chemical plants for coal synthesis.
So I'm happy as a philosopher.
The question of whether we should put NineNine's build to a long trial, over a 250+ hour timescale, does verge on absurd. I would only do it if we agreed on some ground rules: I have to demonstrate it produces an output that solves a beginners issue. I need to use infinite parts sources and voids so there's no question about inflow or outflow feed. The other value of this is that we aren't in a childish contest of "who has best" because we can look each other in the eye and say "when we get this rate, we can multiply it by a constant, and be found to be equal".
When it's jamming randomly, c*random=random. It'll be as kind to anyone.
The build I produced would have to be technically correct from the point of view that it feeds parts only. As such, any story includes a blueprint and multiple trials (though they can be done at the same time by copy and pasting the build onto multiple surfaces). They may decline at any time that they choose to play this game including choosing not to play it at all. I've had a discussion in pm's, if they are curious they're allowed to change venue.
On that note:
I'm confident my result will provoke curiosity and show the issue has more depth than a hundred hours of play and throwing material at the problem at a production rate. But I've already shown how a philosopher might play. It might do well to change the venue to another thread and not continue this conversation from this point here. I'd like some feedback on that from others. PM's are appropriate if you feel an opinion would be a bit too public.
The relevant ones are that a beginner is insulated from error due to the time scales involved. And a more curious mind will observe some interesting phenomenon which makes the problem seem a bit deeper. And at 50 hours in, well...
This is where a discussion of engineering tolerance comes in. A carpenter, a game developer and a mathematician will all use a constant number. The carpenter and the game dev. both need to make decisions on the basis of a table. A carpenter is working in an environment where 2-4 digits give a close enough value.
A game dev. usually needs a program to run in time. But they might be simulating physics, so they need a bit more digits. A mathematician wins their game if they come up with another method with a novel property or demonstrate the capability of the state of the art. They can throw a super computer and a lot of human time at just getting the constant.
So from the point of view of engineering tolerance. 50 hours and several hundred machines built at legendary quality justified me looking over at this feed mechanism on an automall, tearing the belt up and going. "Oooo. Tasty. Spicey.
". By the time it jammed it'd redone my engines, beacons and crushers.It's tricky. The map editor gives me the capability to act without showing some care and tolerance by simply copy and pasting another's work in every detail.
Without isolating the build I cannot isolate an interesting fact about the build being shown. I could also call into question NineNine's workmanship. I'm a shrewd man. Using infinite sinks removes the option of saying that this issue, if it even could be replicated in their system, was due to operator error on the input or output side of build.
From the point of view of being technically flawless it halts so it is not a beginners build. From the point of view of completing operations a beginner should use it for it is an excellent build and a fantastic learning experience and they should continue. That's fundamentally inconsistent. From the point of view of being relevant, you'll be wanting those chemical plants for coal synthesis.
So I'm happy as a philosopher.
The question of whether we should put NineNine's build to a long trial, over a 250+ hour timescale, does verge on absurd. I would only do it if we agreed on some ground rules: I have to demonstrate it produces an output that solves a beginners issue. I need to use infinite parts sources and voids so there's no question about inflow or outflow feed. The other value of this is that we aren't in a childish contest of "who has best" because we can look each other in the eye and say "when we get this rate, we can multiply it by a constant, and be found to be equal".
When it's jamming randomly, c*random=random. It'll be as kind to anyone.
The build I produced would have to be technically correct from the point of view that it feeds parts only. As such, any story includes a blueprint and multiple trials (though they can be done at the same time by copy and pasting the build onto multiple surfaces). They may decline at any time that they choose to play this game including choosing not to play it at all. I've had a discussion in pm's, if they are curious they're allowed to change venue.
On that note:
I'm confident my result will provoke curiosity and show the issue has more depth than a hundred hours of play and throwing material at the problem at a production rate. But I've already shown how a philosopher might play. It might do well to change the venue to another thread and not continue this conversation from this point here. I'd like some feedback on that from others. PM's are appropriate if you feel an opinion would be a bit too public.
Re: How to upcycle?
This was long and I'm confused as to what prompted it....
But I wanted to remark on the cause of the jams, since it's not clear to me if you are aware. When you recycle a chem plant, for each ingredient you get a 75% chance of one item back and a 25% chance of two items back. (IIRC everything out of one recycle craft is the same quality? I can't remember for sure)
So you usually won't have the same amount of each type of item over time -- at any point in time you are likely to have a surplus of some item. On average, over N crafts, the size of that surplus will be proportional to sqrt(N). And when the surplus exceeds your buffer capacity, it jams.
Adding a buffer chest to each building will vastly increase the amount of time your design can run. But if you actually want something that can, in principle, run forever, you either need to have an overflow valve to drain excessive surplus, or supplemental material feeds to cover deficits.
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
The jam occurs because the belt recirculates entities. The recycler can be replaced with an equivalent machine that reads from a tape an input symbol and produces output symbols randomly onto that tape. Position on a belt very naturally corresponds to position on a tape, with repeated symbols corresponding to 1-n stacks of parts or runs of parts. If there where some argument in being vague a symbol on a tape indicating stack/not stack would clear it up immediately.Hurkyl wrote: Tue Jun 17, 2025 8:23 amThis was long and I'm confused as to what prompted it....
But I wanted to remark on the cause of the jams, since it's not clear to me if you are aware. When you recycle a chem plant, for each ingredient you get a 75% chance of one item back and a 25% chance of two items back. (IIRC everything out of one recycle craft is the same quality? I can't remember for sure)
So you usually won't have the same amount of each type of item over time -- at any point in time you are likely to have a surplus of some item. On average, over N crafts, the size of that surplus will be proportional to sqrt(N). And when the surplus exceeds your buffer capacity, it jams.
Adding a buffer chest to each building will vastly increase the amount of time your design can run. But if you actually want something that can, in principle, run forever, you either need to have an overflow valve to drain excessive surplus, or supplemental material feeds to cover deficits.
Recirculating symbols can be done by rotating a single tape while editting it with a machine that is equivalent to a bank of assemblers. It could also be done by issuing an instruction that causes a tape to move left and right, and the turn on the belt would be equivalent to seeing that instruction once every n steps on one tape.
Tasty yummy very spicy
The machine demonstrated halts. But the same machine given a limited instruction set that is always accepted by a bank of assemblers would be like changing to a machine that only feeds in input from one direction and either accepts it or rejects it. You could run it for countably infinite time if it always accepted inputs. In lay terms "awhile". Those instruction sets wouldn't capture complicated behavior but using them for a Monte Carlo method attack on a problem upcycling presents would be productive.
In plain english, gears or wires recipes but not chemical plants in those sims lets you build models easier, but you'll be able to answer less questions. The fact that it does halt means you need some trial and error or a really good math model of when it can halt. Finding a math equation can take a lot of trial and error but if you have it than you can skip trial and error completely. The method is not flawless and has well known and acknowledged tradeoffs.
That might go some way in explaining why I "took jabs" at asteroid reprocessing with minute long models instead of solving it with one brilliant stroke. I needed to see if my approach had the effect of a converging accuracy or I needed to resort to math that showed it would. Not everyone has math. The discrepancy droid pointed out in machine counts indicated I could be in error on account of operator skill. But with a technical adjustment, I was confident in putting more energy into my method and running a ten minute sim. Our numbers match and I can be confident in both our methods.
There is a quality calculator mod up, https://mods.factorio.com/mod/quality-calc. If it has some limits, playing around could reveal them. I mean, 1 in 2400 on asteroids is a hard number to accept, particularly if you arrived at the same logic in another case like sulfur for blue science upcycles
Be careful giving an exact percent because there's multiple ways of saying "2 items/all possible inputs" here. Cargo bays would be different than chemical plants. I'm not saying you're wrong, but I couldn't hazard you to always be right; 2 items can mean 2 aces in poker or all hands or 2 pairs in poker. I'll accept the O(n)=sqrt(N) logic on faith.
You should also think carefully on what happens if you give a machine too much of a buffer of any size. There's an awful lot of gears on Fulgora. And an awful lot of robot parts, but there's a strange discrepancy there. Robots can't have hard discussions and that isn't always to their benefit.
I might ask for proof and it might lead to some discussion
. That's the primary reason for questioning if a change in venue might be better rather than other factors that may be at play.Re: How to upcycle?
No, you just use used complicated word inapropriatly. You described removing a belt to unclog your system. That's quite the lame way to fix your original system imo. That wouldn't make the system better at all, unlike if you were to try and use the advices and example that were given on this thread.coffee-factorio wrote: Tue Jun 17, 2025 6:57 pm Those instruction sets wouldn't capture complicated behavior but using them for a Monte Carlo method attack on a problem upcycling presents would be productive.
That's the true epistemology of math there x) !
And i've seen people doing things like this : https://dfamonteiro.com/posts/factorio- ... cler-loop/
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
It's a good article. In the final graph of the article, there's a graph of each edit: quality module effect level. I originally just used recycler level because that's how I think of it "grade of recycler power".
For four quality modules (the furthest point on the graph from 0), the percentage given for a recycler only upcycle is between 1 in 4800 and 1 in 2400 - 0.02% and 0.04%. That's about what I'm getting from simulation, I'll take accept the author's logic. A "recycler only upcycle loop" is the equivalent of only upcycling asteroids, only upcycling sulfur and directly upcycling plastic. There exact number appears to be under 1 in 2400 so I'll need to run my sim for longer if I need better than that. I'm fine with what I have though; it's not for me where I can avoid it.
For four quality modules (the furthest point on the graph from 0), the percentage given for a recycler only upcycle is between 1 in 4800 and 1 in 2400 - 0.02% and 0.04%. That's about what I'm getting from simulation, I'll take accept the author's logic. A "recycler only upcycle loop" is the equivalent of only upcycling asteroids, only upcycling sulfur and directly upcycling plastic. There exact number appears to be under 1 in 2400 so I'll need to run my sim for longer if I need better than that. I'm fine with what I have though; it's not for me where I can avoid it.
Re: How to upcycle?
coffee-factorio wrote: Tue Jun 17, 2025 7:57 pm It's a good article.
I'll take accept the author's logic.
You should see the math on the wiki too :
https://wiki.factorio.com/Quality
They probably could do with a review from a trusted monte carlo simulatorist
Edit: Note that if anyone doesn't understand those math on the wiki, they can ask question, but it's not necessary to upcycle, the advices on the early post of this thread are more than enough. They vulgarize most of it.
It's not recommended to pretend you are doing complicated math using wrong terminology when correct reasonning and solutions are already widely available publicly. It highlight not only your inability to do the proper math. But worse it prevent you from learning them, show you didn't even try to document and propagates non-sensical long text devoid of any useful content.
Last edited by mmmPI on Tue Jun 17, 2025 11:17 pm, edited 1 time in total.
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
This is a proof of concept run based on my "failed build" that has been running since last night, with some occasional pauses.
I have another build where the sushi belt is cut and goes to belt which lets me see if an item gets past legendary bulk inserters. It hasn't in ten hours simulated time. The blueprint will be provided in a bit.
The beginner may find some value in the blueprint provided for making simple upcycles, although since it uses an infinite source and sink they still need to design the rest of the build and scale the inputs appropriately. This in effect converts what might be called a Turing Machine into a simpler automata by "fixing the direction of the tape feed so it only moves in one direction, and either accepts or rejects its input".
This however, does not reproduce the conditions given on the wiki where there are 2 quality modules and 2 productivity modules in each step of the recycling process. In short max modules 4, 2 productivity modules and 2 tier 3 quality modules. Predicted value is 80 crafts to one part or 1.25%. I'm having an interesting time of it. I'm giving the sims a hundred hours time at 64x before I report. Someone with a program might want to check that, I read Daniel's blog some more and didn't see him run that. Daniel took the wise step of sanity checking his program against the wiki, though I'm not using python I seek to emulate that logic
- proof-of-concept-build.jpg (404.54 KiB) Viewed 364 times
The feed system is untouched. The only difference is that the recipe has been changed from chemical plants to gears, which is what I suggested in the post you quoted:mmmPI wrote: Tue Jun 17, 2025 7:26 pmNo, you just use used complicated word inapropriatly. You described removing a belt to unclog your system. That's quite the lame way to fix your original system imo. That wouldn't make the system better at all, unlike if you were to try and use the advices and example that were given on this thread.coffee-factorio wrote: Tue Jun 17, 2025 6:57 pm Those instruction sets wouldn't capture complicated behavior but using them for a Monte Carlo method attack on a problem upcycling presents would be productive.
coffee-factorio wrote: Tue Jun 17, 2025 6:57 pm
In plain english, gears or wires recipes but not chemical plants in those sims lets you build models easier, but you'll be able to answer less questions.
I have another build where the sushi belt is cut and goes to belt which lets me see if an item gets past legendary bulk inserters. It hasn't in ten hours simulated time. The blueprint will be provided in a bit.
The beginner may find some value in the blueprint provided for making simple upcycles, although since it uses an infinite source and sink they still need to design the rest of the build and scale the inputs appropriately. This in effect converts what might be called a Turing Machine into a simpler automata by "fixing the direction of the tape feed so it only moves in one direction, and either accepts or rejects its input".
This however, does not reproduce the conditions given on the wiki where there are 2 quality modules and 2 productivity modules in each step of the recycling process. In short max modules 4, 2 productivity modules and 2 tier 3 quality modules. Predicted value is 80 crafts to one part or 1.25%. I'm having an interesting time of it. I'm giving the sims a hundred hours time at 64x before I report. Someone with a program might want to check that, I read Daniel's blog some more and didn't see him run that. Daniel took the wise step of sanity checking his program against the wiki, though I'm not using python I seek to emulate that logic
Re: How to upcycle?
Sorry i added the edit saying that it was not recommened to make up some non-sensical text devoid of useful content too late maybe.
It's not the place to show your creations, this is gameplay help
You're not providing any help, and not accepting any, why would you quote me on this and ignore my personnal message ?
You're trying to obtained some kind of reaction by trolling poorly ?
It's not the place to show your creations, this is gameplay help
You're not providing any help, and not accepting any, why would you quote me on this and ignore my personnal message ?
You're trying to obtained some kind of reaction by trolling poorly ?
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
It gave me some fight on the numbers side.
So for the Monte Carlo simulation: I've got a machine that acts like a PDA or and FSA, basically it's simple and it don't jam. So it's as much time as I want to put in it. Parts are defined as time*parts a machine could produce. But this appears to be a mistake in light of what a matrix method produces.
1) I set up how long this will run and what I am going to test on it.
2) I run it.
3) Do math on it.
4) Interpret that math
On step three it took me some time because I was getting number that indicated I was 1.5x short of Konage and Daniels computed values. After I did a lot of work (including looking at Konage's beautiful table https://docs.google.com/spreadsheets/d/ ... id=0#gid=0) what I came around to was that there's a 1.5x productivity bonus that's on the machine that more or less isn't factored into the total tally of gears. What I came up with is that a legendary machine outputs 2 gears per second * 1.875. That means it is doing 3.75 gears per second of work. Then I multiplied this by seconds to get total work, for a hundred hours.
Once I did that, I looked onto three different works surfaces. On each surface this is gears produced, and my model is gears produced/ gears made = rate of gears. After a lot of looking at my own work and some effort to make sure all parts where getting used, I got that this was ~17000 gears / 11350000 parts.
The slowdown on my end was that I originally used the output for the assembly machine number on parts which factors in productivity, which is going to divide the output by 1.5. That 1.5x discrepancy was tying me in knots for hours, because after some work I got some python running which was showing that it was more likely that I had made this sort of mistake and needed to go over the source of my numbers one at a time.
With that error accounted for I can verify the wiki's number of 1 in 80 and Konage's work as well. The addition of productivity threw me through a loop for a moment, but I'm now confident in my method and its result. Considering all things, I agree with him on three digits and I'm impressed by his skill.
For the entertainment of Seablock players and the curious. The blueprint for the test build pictured is provided.
The numbers part, grabbing the 17000 from the production graph was the easy part. Fixing down the input so I could see exactly how the graph was being calculated was the hardest part. The best clue might have been that the machine takes 7.5 iron plates as input. So that's going to be the hot spot in rate calculations that if you do this, you want to watch very, very carefully.
So for the Monte Carlo simulation: I've got a machine that acts like a PDA or and FSA, basically it's simple and it don't jam. So it's as much time as I want to put in it. Parts are defined as time*parts a machine could produce. But this appears to be a mistake in light of what a matrix method produces.
1) I set up how long this will run and what I am going to test on it.
2) I run it.
3) Do math on it.
4) Interpret that math
On step three it took me some time because I was getting number that indicated I was 1.5x short of Konage and Daniels computed values. After I did a lot of work (including looking at Konage's beautiful table https://docs.google.com/spreadsheets/d/ ... id=0#gid=0) what I came around to was that there's a 1.5x productivity bonus that's on the machine that more or less isn't factored into the total tally of gears. What I came up with is that a legendary machine outputs 2 gears per second * 1.875. That means it is doing 3.75 gears per second of work. Then I multiplied this by seconds to get total work, for a hundred hours.
Once I did that, I looked onto three different works surfaces. On each surface this is gears produced, and my model is gears produced/ gears made = rate of gears. After a lot of looking at my own work and some effort to make sure all parts where getting used, I got that this was ~17000 gears / 11350000 parts.
The slowdown on my end was that I originally used the output for the assembly machine number on parts which factors in productivity, which is going to divide the output by 1.5. That 1.5x discrepancy was tying me in knots for hours, because after some work I got some python running which was showing that it was more likely that I had made this sort of mistake and needed to go over the source of my numbers one at a time.
With that error accounted for I can verify the wiki's number of 1 in 80 and Konage's work as well. The addition of productivity threw me through a loop for a moment, but I'm now confident in my method and its result. Considering all things, I agree with him on three digits and I'm impressed by his skill.
- Screenshot 2025-06-18 003048.jpg (434.41 KiB) Viewed 313 times
The numbers part, grabbing the 17000 from the production graph was the easy part. Fixing down the input so I could see exactly how the graph was being calculated was the hardest part. The best clue might have been that the machine takes 7.5 iron plates as input. So that's going to be the hot spot in rate calculations that if you do this, you want to watch very, very carefully.
Re: How to upcycle?
The quick answer is that when you repeat independent random trials with identical distributions, (i.e. each recycler output), the mean and the variance of the overall result is computed by simply multiplying the mean and variance of a single trial by the number of trials.coffee-factorio wrote: Tue Jun 17, 2025 6:57 pmI'll accept the O(n)=sqrt(N) logic on faith.
...
I might ask for proof and it might lead to some discussion
Thus, the standard deviation is proportional to the square root of the total number of trials.
You could now do some calculus with the binomial distribution to argue things, but an easier route is to invoke the central limit theorem: with a lot of trials, the actual distribution of the end result is going to be extremely close to a normal distribution, which is just scaled up from a standard normal by multiplying by the standard deviation (and shifted by the mean).
So if we're asking the average value of difference between the most and least of four independent samples (i.e. each item samples independently from the overall distribution of N trials), the answer will be scaled up by a factor of sqrt(N) from the corresponding question about the standard normal distribution (and with an additional error due to the fact we're using the CLT approximation to the true result, but the error is increasingly negligible as you do lots of trials).
Re: How to upcycle?
It's a setup limited in use because it works only for things that have a single ingredient, as it would jam otherwise.
It's made full of legendary module, so it's a late game setup, yet it's using assembly plants to make gears. It's a non-sense from a gameplay perspective not to use foundries there, assembly plant with legendary productivity module for gear upcycling is another reason this setup is cursed.
Now for math demonstration which seemed the point, it's also a bad example, because the reasonning is backward, if you built a setup first, and then you look at the average result for arbitrary amount of time without any clue as to when is enough to stop, or how much time correspond to how much significance in the result, or wether the result will be stable , it will lead to ridiculously many tests for every module permutation that can exist or any alteration to the original design that needs a lot.
The purpose of the math on the wiki or linked articles is to do predictive math, you calculate how much you need to have to meet your objectives, you do the math first, to know how much to build.
If you don't want to use math, and you just want to look at a picture of the average over sometime, then anyone can just play the game and look at their production stats in their actual game with the module they can afford at that time, not pretending they're doing anything more than that, it's fine too x), you can compare 2 platform's chain of production, and use asteroid as source of infinite input and space as void.
I can understand the thrill of testing setup in editor, if that's what you enjoy doing, fine, but for anyone that could be tempted to do those test because they care about the results only, then it's quite the waste of time, just ask for explanations on the math if you don't get them, which parts or what you're trying to math out.
"this is switzerland"
"no"
"this is switzerland"
"no"
"this is switzerland"
"no"
"this is switzerland"
"YES"
"i'm progressing in geography"
"yes"
"because my method works so well"
"no"
"you bet i can find belgium ?"
"it's there on the map where it's written"
"this is belgium "
" no this is switzerland .... and your method appeared even worse after that last one try"
That's not always the case, obviously you can also increase linearly the errors when you do more trials as was demonstrated on this thread.Hurkyl wrote: Wed Jun 18, 2025 9:07 am but the error is increasingly negligible as you do lots of trials
The quick answer doesn't explain how more trials make the error increasingly negligible in this context unfortunatly. Do you just look at it long enough to have a good feel ? or do you use some kind of math method with complicated name to know how many is "lots of trial" ? Something that would maybe clear the misunderstanding between throwing a coin 3 time in the air and claiming it's not a fair coin due to the simulation result, and a proper simulation that would actually meaningfully represent the increasing certainty that the coin is actually fair or not ?
Because from what i understood you have explained the reason's why "poorly built dedicated loop" are bound to fail with more than 1 ingredient with increasing certainty over time. As it grows as sqrt of number of trials ,which tend to infinity, it also tend to infinity, and you can't have infinite buffer in real game so such setups are going to fail eventually with more than 1 ingredients.
But it seems to me that the latest part is explaining that in order to know when exactly the failure will happen, or how likely the failure is to occur after a certain amount of time, it's possible to run many trials, and in doing so knowing with increasing precision the probablities ? Sounds like a task for a monte carlo simulator enthusiast x).
Not sure it's overall useful for upcycling in practice, as you can just add extra ingredients from elsewhere or void some extra in case one of the multiple ingredients from a "properly built dedicated loop" is reaching too high of a quantity, to avoid any problem in the first place, that was mentionned in this thread, but in the case where a player has followed none of the advices or made a mistake in the building of their dedicated loop, inspired by the recent pictures posted, it could be useful to understand when it fails, not just why
I'm considering using those failed blueprints that will clog randomly to make music, it could be something like "the sound of failure", and instruments would start randomly playing music when their dedicated loop fail btw !
Check out my latest mod ! It's noisy !
-
coffee-factorio
- Fast Inserter
- Posts: 125
- Joined: Thu Oct 17, 2024 10:56 pm
- Contact:
Re: How to upcycle?
You're going to post an image of a gigantic plastic refinery that uses perhaps the least efficient upcycle practical? That can only run indefinitely because it uses a single parts stream at the point of recycle/upcycle? And talk about how it's a late game build and a mark of how much a player has? And then make this argument?mmmPI wrote: Wed Jun 18, 2025 12:49 pmIt's a setup limited in use because it works only for things that have a single ingredient, as it would jam otherwise.
It's made full of legendary module, so it's a late game setup, yet it's using assembly plants to make gears. It's a non-sense from a gameplay perspective not to use foundries there, assembly plant with legendary productivity module for gear upcycling is another reason this setup is cursed.
Your build is a species of this class of machine PI. I'm not criticizing your art here; a drill that can do a high ips and is making nothing is as useful as a rotting orchard. The ultimate efficiency is immaterial at that point.
But you can see how I might look at what you yourself are doing and find your arguments lacking.
We've already taken criticizing things for producing limited sets of items off the table because NineNine's build runs enough to solve beginners problems. A beginner with the modules available to them could happily switch out modules. I'd be at the same point I was when people where suggesting using t2 quality modules at legendary, and I'd say the same thing: "That's a start but you probably want to do something else."
Also, using the wiki's numbers, an em plant doing copper wires is going to be very efficient. Using Konage and Daniels more full table, it's still going to be very efficient with a different module choice. You can hook up your choice on Gleba with copper bacteria whether or not you decide to also include calcite in that mix. Calcite is a more efficient process and I'd be forced to choose between wires in assemblers and gears in assemblers to match, instead of a universally available if inefficient benchmark. That matches the wiki's numbers and independent publishing.
I've been getting numbers that match the wiki's and people on this thread using other methods for some time. That simply isn't something you get to decide individually.
I hope you don't mind that I put your full words up; I didn't follow all of itHurkyl wrote: Wed Jun 18, 2025 9:07 am
The quick answer is that when you repeat independent random trials with identical distributions, (i.e. each recycler output), the mean and the variance of the overall result is computed by simply multiplying the mean and variance of a single trial by the number of trials.
Thus, the standard deviation is proportional to the square root of the total number of trials.
You could now do some calculus with the binomial distribution to argue things, but an easier route is to invoke the central limit theorem: with a lot of trials, the actual distribution of the end result is going to be extremely close to a normal distribution, which is just scaled up from a standard normal by multiplying by the standard deviation (and shifted by the mean).
So if we're asking the average value of difference between the most and least of four independent samples (i.e. each item samples independently from the overall distribution of N trials), the answer will be scaled up by a factor of sqrt(N) from the corresponding question about the standard normal distribution (and with an additional error due to the fact we're using the CLT approximation to the true result, but the error is increasingly negligible as you do lots of trials).
. The central limit theorem I looked up on wikipedia, and it seems like something profoundly beautiful. I'm in a difficult spot because I'm looking for convergence to some tolerance only so I only have to use part of it's beauty. It justifies giving a method more time iff it appears to be heading towards what is on the wiki and independent publication. Where I'm from "iff" means something shrewd: if and only if. I found this in the references and I think someone else did very well with it. https://www.mctague.org/carl/blog/2021/ ... t-theorem/. It recalls the song "Happiness is Warm Gun" by John Lennon, which was composed by cutting up and rearranging sequences of words. He was inspired by interacting with William Burroughs, who used the same technique to write "Naked Lunch". Although due to the topic of his writing his material is infamously difficult, even ugly; hard men do argue his work is beautiful as frequently as they find it ugly. Neither man is a student of hard mathematics though both lived hard lives
I'm glad my faith was met with a full proof. Sometime later someone might see a difference in our post counts, but you're an inserter to a fast inserter now and you've schooled me well.