Friday Facts #223 - Reflections on 2017

[BasketKees]

It is quite sad that the community seems to be so split about the fluid changes. But on the other hand at least we can have a good discussion

To be honest, I had not been aware that fluid handling by trains was so powerful. I simply played the game and took it for what it was. Now that this issue has come up an having thought about it, I can agree that it was very powerful. From that standpoint I can see why the changes were made.

First of, I can see why and don't object to changes, but I don't necessarily agree that it was necessary. Except for pipe mechanics, which I find to be unintuitive, I had no problem with fluid handling. I like a game to be balanced and would most often disagree with changes that provide for overpowered mechanics. Nuclear artillery would be absolutely bonkers. And I even think personal roboports are too overpowered; they are so powerful, they discourage constructing static bot networks. But I never felt this way about fluids. It had to be pointed out to me that it was so powerful and even now I'm not bothered by it. The complexity for fluids for me is in the design of refinery and production sites. Handling fluids in production is, intentionally so, a total pita. The real challenge of this game is logistics of course, not production. So you could argue that 'easy' fluid handling takes away from the true challenge of the game. I would counter argue that the logistics of the game are challenging enough; you can most literally spend thousands of hours to accomplish your own set goals without ever even looking at a mod. Adding extra logistic complexity for fluids does not really add to that in my opinion. To me that feels mostly like complexity for the sake of complexity. If you enjoy that, which is perfectly fine of course, you have the option of modding the game with Bob's/Angel's and other great options. Otherwise, why not leave fluid handling as the easier part of the game?

But okay, besides me arguing that the changes could be 100% reverted without hurting the game, the consensus seems to be that a nerf is in order. Like I said, I would not object to that per se since it was very powerful. The implemented nerf to the fluid wagon is unfortunate and mostly because of the lack of separation. That was a versatile feature that provided extra gameplay options. It is unfortunate that it has been removed. The lowered capacity doesn't hurt much I would say. Expanding the infrastructure to accommodate should not be much of a problem in most cases. But again, not more enjoyable either; just a bit more of the same. Losing the separation feature is a greater loss and I really dislike the removal being used as an argument that it 'buffs' barrels. That could be classified as New Speak. All in all I don't think it was necessary nor do I think that it improves the game, but I could live with the fluid wagon changes.

When it comes to barrels, I strongly dislike the changes. Again I don't really mind the lowered capacity. If most people think a nerf is in order, so be it. But the performance has been reduced by such a great amount that I argue that they have been made close to obsolete, except for those who want to make things more difficult for themselves just for the sake of it. In all but a few rare practical use cases, barrels are now worse than the other options. I would be very hard pressed to come up with a use case where the other options are not better or at the very least would suffice. You could now remove barrels from the game completely and it would hardly make a difference.
The argument has been put forth that the extra complexity required for barrels is worth it because they are more versatile. Well, they are more versatile partly because the other versatility option has been removed, so I don't find it a strong argument. And the versatility they had has been severely reduced because of the lower throughput they provide in transportation. Yes, you can mix fluids in the same inventory, but if you do you have hardly any capacity for each of the different fluids. In most cases you need many inventories for a single type in which case the possibility of mixing is irrelevant. As far as I can tell the only real use case for barrels now is a very low amount over a large, greater than possible by pipe, distance. And there are very few of those use cases in the game. And even those could be handled in other ways, so why even bother with barrels?
Another argument is that it does not make sense that barrels provide better throughput than other options. I agree. But quite frankly, I don't care. The realism argument is lost on me. This is a game and things have to make sense gameplay wise first, realistically second. <wait, let me pull this rocket silo out of my pocket before I continue> Gameplay wise barrels are more complex. The logistics behind them are decently complex. So much so that I hardly ever see anyone do them properly. Most hide or circumvent the complexity by filtering wagons and/or creating oversized buffers for barrels. I would argue that this complexity is the price you pay for their, now decreased, versatility. But with these changes, we have to pay double, because there is also a price in throughput. I say realism be damned; barrels are more expensive, more work and more complex and there has to be an advantage in throughput to make that worth the trouble.
Lastly, bots are extremely powerful with barrels. Yes, they are. Who cares? Bots are extremely powerful, period. They are the OP route in the game. They take more design and planning to do efficiently than a lot might argue, but when you do design well, they are as OP as you can get. I would say that is intentionally so. Bots make finer details simple, but they allow for an insane increase in upscaling. Bots are the reason you can go huge and I am often amazed by what people come up with to make all the logistics work at that scale. Bots make production trivial, but production is not the biggest challenge of the game to begin with. Well designed bot logistics on a large scale is very challenging indeed. And to make that possible they have to be OP as can be. Who cares they can carry a million fluid per second? I don't.

Sorry for the long post. I love this game and want it to be better still. These changes do not achieve that goal in my opinion, so the very least I can do is spend some time to share my thoughts. On with the discussion!

[SamuelS]

The barrel changes were way to much. they completely broke mega-base or any base that used any significant quantity of fluids. On top of that, cargo wagons can only realistically carry half of their inventory in barrels because they need the other half to carry the empty barrels back.

[Tricorius]

You can't tell me that the 0.16.X release was one that was overwhelmingly positively received. Factorio is still an amazing game. It is well worth the $10 or $20 dollars people have spent on it. I am worried that if they keep making these type of changes the game will not be worth the $30 or $40 dollars they will want to charge when it gets officially released.

Another shining example of the selection bias occurring in the forums. Eight pages of comments are a *miniscule* percentage of the 1.2 million purchasers of the game. If I were a betting man, I’d place my wager on 99.5% of those people not basing whether to play Factorio based on the amount of the fluid that pipes or barrels carry. Or whether belts compress via side loading.

It is a *very* self-selecting set of people who actually care about any of this stuff. And I’m willing to bet quite a large portion of that group is on these forums.

I know the forums are a very small vocal minority. Does that mean we are any less important? When one of us has something to say, should it not at least be listened to? There is a group of forum members that are voicing their distaste of a decision made about the game they enjoy. I can only base my findings on what I see, and that is the forums.

As for the amount of people that have purchased the game and are playing it, my opinion is just as valid as theirs. If they want to voice their opinion then they have every opportunity to do so, just like I do.

I never said you shouldn’t discuss it. I never said you shouldn’t be heard. My philosophy is actually that “raving fans” should absolutely have input toward a product as they tend to evangelize and drive sales. However, the tough parts on managing a product is saying no.

I empathize with the Wube team. I feel they have either miscommunicated a couple of things, or their communications have been misconstrued. I think they are doing what they can to make a fun, balanced game. If I had to wager a guess, there was internal disagreement whether the connectable fluid wagons were worth the investment.

I never used them, so I don’t have an opinion one way or the other. However, it is pretty obvious to me the feature was near the chopping block for some time. I’m unsure if they have any heuristics or metrics in the game, but I for one would be fascinated by the statistics around that controversial little GUI toggle.

Finally I never said your opinion was more or less valuable than anyone elses’. Though based of the “raving fans” principle mentioned above it might be more valuable for monetary and “hype” reasons.

Again. It is easy to fall into the trap of creating community-driven development. This rarely ends in a gloriously wonderful product.

My comment was merely in regard to the increasingly-common notion that these forums are the end-all, be-all voice of every single person playing Factorio.

“You can't tell me that the 0.16.X release was one that was overwhelmingly positively received.”

Using these forums as a basis for the overwhelming majority of players is simply not accurate. Every single person on this forum could absolutely be fuming crazy that a certain feature was removed...and more than 1.1 million other Factorio players could be completely oblivious to the removal of the feature; happily playing and recommending the game to their peers.

The harsh reality: We are a microcosm of a microcosm. And the development team doesn’t owe us anything. And that doesn’t make them bad people or bad designers, developers, and project managers. In reality I think they do a fantastic job of building and refining Factorio. It is still one of the best damn games ever made. Am I happy with all the changes? No. I’m a person with opinions and preferences too. But one of the most fun aspects of Factorio, for my money, is working through the ebb and flow of a factory design within the constraints of the game design. And I think the devs have given us amazing tools, and a very open environment to do so, and to then discuss and share our creations.

[Zavian]

The barrel changes were way to much. they completely broke mega-base or any base that used any significant quantity of fluids. On top of that, cargo wagons can only realistically carry half of their inventory in barrels because they need the other half to carry the empty barrels back.

I fail to see how this prevents building mega-bases. (It might cause you to change the way you build them, and you will probably want to make changes to any mega-base uses lots of barrels, but it shouldn't prevent you building them. Also it is perfectly possible to carry load a full load of barrels each way in an unfiltered train. One way to do this is to have the train stop, unload all the barrels, then move forward a train length to another stop, and load a new set of barrels. (You can also do it in one stop, but this is simpler to explain).

[stretch611]

The barrel changes were way to much. they completely broke mega-base or any base that used any significant quantity of fluids. On top of that, cargo wagons can only realistically carry half of their inventory in barrels because they need the other half to carry the empty barrels back.

I fail to see how this prevents building mega-bases. (It might cause you to change the way you build them, and you will probably want to make changes to any mega-base uses lots of barrels, but it shouldn't prevent you building them. Also it is perfectly possible to carry load a full load of barrels each way in an unfiltered train. One way to do this is to have the train stop, unload all the barrels, then move forward a train length to another stop, and load a new set of barrels. (You can also do it in one stop, but this is simpler to explain).

I agree. Current blueprints may need to change, but megabases are not broken.

As for using full loads of barrels in a unfiltered train... To do it in a single station, use stack filter inserters when unloading. e.g. If you have barrels of crude unloading, have the stack filter inserters only unload crude barrels, and have stack inserters (or stack filter inserters if you prefer) load empty barrels.

[Engimage]

The barrel changes were way to much. they completely broke mega-base or any base that used any significant quantity of fluids. On top of that, cargo wagons can only realistically carry half of their inventory in barrels because they need the other half to carry the empty barrels back.

I fail to see how this breaks megabases. It is not removing any kind of game feature but only reduces its throughput and you just "need more" which is completely fine with megabases.
I have built megabases and I never used barrels at all. So your argument is simply not valid.
Yes it kinda decrease effectiveness of builds which did benefit from this exploity feature. But you can easily work around it.

[malventano]

The problem is not that barrels are OP, it's that the pipes are so weak that nobody wants to use them. Buff pipes, buff fluid wagons, and leave barrels as they are.

I haven't used barrels since 0.15 added tankers. I've never had problems with fluid throughput for pipes. You just need to plan your pipe network the same way you would a belt network, and use multiple pipes, if one pipe won't have enough throughput. With one pipe you can get over 1200 fluid/sec over short distances (less than 10 pipe segments), and over medium distances (between 10 and 200 pipe segments) you can get between 1000 and 1200, and if you need a long run of over 200 pipe segments (1100 tiles using underground pipes), you might get less than 1000 fluid/s (and you if you want 1000 fluid/s over 1000 tiles, then you should probably think about using a train anyway).

Also you are still ignoring the bit about bots being able to carry way too much fluid in barrels compared to plates.

If you need help splitting off of pumps into parallel pipe runs, I've created a mod that may help you out. It can support 200 segments (198 if you exclude the manifold segments at either end of the run) at the full 12,000/s pump speed across 10 parallel runs.

[OBXandos]

I guess I will go on record to specifically state what I am upset about. I have been avoiding that topic and watched many people infer what they think I am saying.

Belt compression should not have been changed and the fluid tanker should not have had its ability to be split apart removed.

I completely understand that barrels were way to overpowered when you compare straight throughput. I agree with them needing a nerf, I don't know if one this major was needed but time will tell on that. I don't really think the fluid tanker needed this much of a capacity nerf either but it is at least in line with what barrels can still accomplish and the weight of the car was reduced so in the end it is probably acceptable.

The excuse that splitting the fluid tanker apart was complicated, or underused, not noob friendly holds absolutely zero water with me. The tanker in its stock form works exactly like it does now, so there was no complication removed. Maybe it was underused, but why is that a bad thing? I completed my first rocket launch without using blueprints, the circuit network, or logistics/construction bots. I could argue that those systems too are underused and way more complicated. To make things more noob friendly you can add something on the tooltip or make a tutorial, that is an easy fix.

I have no idea where the belt compression changes came in. Now you have to setup timers on your inserters using the circuit network to have them properly lay down items on belts? That is the definition of over complicated and not noob friendly. I can understand it from a reality standpoint but that flies out the window when I can carry a train or nuclear reactor in my pocket.

The change from using alien artifacts(0.15), belt compression changes, and fluid tanker splitting are all changes I don't agree with and make me worried that the game is heading in a direction that I may not like. The artillery train was pretty questionable to me as well. It tries to fix a problem that shouldn't need to exist in the first place. The biters serve no purpose, and the artillery train helps you remove these biters. If I don't need or want biters in my game, why do I need or want an artillery train?

If people can honestly tell me that these type of changes are good for the game, then so be it. I don't think they are.
OBX

[Tricorius]

The barrel changes were way to much. they completely broke mega-base or any base that used any significant quantity of fluids. On top of that, cargo wagons can only realistically carry half of their inventory in barrels because they need the other half to carry the empty barrels back.

I can assure you that barrels are fine. They might *not* be good in existing bases built based on previous throughput characteristics. That sucks, but things break from time to time (even with updates to fully-released games).

I’m building out my current 0.16 base into a megabase and still using barrels (botted-in, no less) and I’m still currently bottlenecks by iron and copper. I just built out the first 2-4-0 belted oil barrel station. It can handle a surprising amount of throughput. Yes, you have to build out. But that is Factorio.

[meganothing]

If one pipe doesn't have enough throughput, why not use two pipes in parallel?

Because over any significant distance (>10 pipe tiles), it takes *12* pipes in parallel just to carry the flow rating of a *single* pump, and even then you need a modded manifold just to be able to distribute that single pump exit across to 12 pipes (using standard pipes to split off of the pump cuts its output to a fraction of its rating).

Why do you compare the pipes to the throughput of a pump? Pumps might intentionally be set to pressurize/fill nearly anything so that you just have to balance length of pipe but never pump pressure. Two pipes in parallel, no matter the distance, doubles the throughput.

[malventano]

If one pipe doesn't have enough throughput, why not use two pipes in parallel?

Because over any significant distance (>10 pipe tiles), it takes *12* pipes in parallel just to carry the flow rating of a *single* pump, and even then you need a modded manifold just to be able to distribute that single pump exit across to 12 pipes (using standard pipes to split off of the pump cuts its output to a fraction of its rating).

Why do you compare the pipes to the throughput of a pump? Pumps might intentionally be set to pressurize/fill nearly anything so that you just have to balance length of pipe but never pump pressure. Two pipes in parallel, no matter the distance, doubles the throughput.

Due to the way that flow exponentially falls off over the first few pipes after the pump, you can't split off to multiple pipe runs and get full flow. Using your example, you can't just keep doubling up on pipe runs and reach the 12,000/s rate of the pump because there is only a single outlet from the pump discharge, so using standard pipes you would drop too much pressure just trying to split out to the runs needed to achieve higher flow rates from singular sources.

Why care about such high flow rates? Because the folks who have such a high fluid demand are the same folks who have been coming up with 'creative' ways to move fluids at a higher rate than possible by pipes.

Why compare pipes to the throughput of the pump? Because the pump is pumping fluid through the pipes. Not everything in the game can be direct tank to tank.

[Zavian]

Due to the way that flow exponentially falls off over the first few pipes after the pump, you can't split off to multiple pipe runs and get full flow. Using your example, you can't just keep doubling up on pipe runs and reach the 12,000/s rate of the pump because there is only a single outlet from the pump discharge, so using standard pipes you would drop too much pressure just trying to split out to the runs needed to achieve higher flow rates from singular sources.

Why care about such high flow rates? Because the folks who have such a high fluid demand are the same folks who have been coming up with 'creative' ways to move fluids at a higher rate than possible by pipes.

Why compare pipes to the throughput of the pump? Because the pump is pumping fluid through the pipes. Not everything in the game can be direct tank to tank.

Doesn't that depend on where and how you do the split? eg use 2 pumps to pump out of a tank, that way each pipe should be able to get full flow. And in almost all cases what you care about is the fluid flow rate. Comparing pipe flow rate to a pump is to a pump is silly, because pumps are cheap, and you can easily have one (or more) pumps per pipe.

[Unclebod]

I just find this "No megabase possible" logic a bit off. Fluids (except from crude oil) has only been possible to ship in barrels from 0.15.0 (the same version the tanker wagon bunced down on the rail)
If I remember correct, the pump (which had a different footprint up to 0.15.0) didn't have enough throughput to fill a pipe (you had to use 3 or more pumps in parallell for each pipe).
So, no Megabases were possible befor 0.15.0. Forget it...

BTW, 0.15.0 broke almost all factories (unless you were all solar and had allready explored all science...)

So, this is an alpha release, things will change. And, besides, isn't the game about adjusting to the situation?

/UncleBod
To old to remember

PS
Was to lazy to find info about the filer in cargo cars, but AFAIR, that was also 0.15.0...

[roman566]

Of course megafactories were possible. But the amount of jumps you had to go to get everything working with pipes only was massive. Then we get barrels and megafactories are suddenly much more fun and easy as you no longer have to use pipes for everything. Now we got barrels nerfed because people liked them a bit too much. You wonder why people complain? How would all people using pipes only feel if output was cut to 20% of normal and the answer for their problem was 'just build more pipes and add pumps'?

[meganothing]

Due to the way that flow exponentially falls off over the first few pipes after the pump, you can't split off to multiple pipe runs and get full flow. Using your example, you can't just keep doubling up on pipe runs and reach the 12,000/s rate of the pump because there is only a single outlet from the pump discharge, so using standard pipes you would drop too much pressure just trying to split out to the runs needed to achieve higher flow rates from singular sources.

Zavian already provided a good answer to this. The question is: What are you trying to optimize here? Surely not 100% pump utilization itself, but pipe throughput over a non-trivial distance, right?

Now, if energy and resources don't matter you just could fill the whole distance with pumps (no pipes at all) and get about 8000 units/s out of that single pipe track. That is what seems to follow from viewtopic.php?f=5&t=46030.

Any design that doesn't use a ridiculous amount of pumps might use a pump every ~150 pipes (or underground pipes) and still achieve over 1000 units/s over one pipe track .(https://wiki.factorio.com/Fluid_system) (T=tank, P=pump, o=pipe).

TPooo....oooPooo...T

Now if you need to double the throughput, you simply do the following:

oTPooo....oooPooo..T
o
oTPooo....oooPooo..T

Lets check if that is really reasonable: To achieve the same throughput with more pumps on a single track instead you would already need a pump every 2 pipe segments, i.e. PooPooPooP, clearly ridiculuous over long distances.

Either tell me what is wrong with my calculation or what is wrong with my priorities/goals.

Why care about such high flow rates? Because the folks who have such a high fluid demand are the same folks who have been coming up with 'creative' ways to move fluids at a higher rate than possible by pipes.

Everyone cares about how much he can transport over a distance. But no one should care about " reach the 12,000/s rate of the pump" as a goal by itself (and that is what your answers up to now boil down to). Usually in any chain the weakest link defines the limit not the strongest link. If "creative" ways achieve more, good for them. But your words still seem to be just another redraft of "It used to be like this and I don't want change".

The only relevant question we can ask is: Does the highest throughput you can achieve with any method (pipes, train,belt) prevent something desirable anywhere else in the game to work (and be fun)? It is possible for us players to define/declare the "prevent something" part (through play-testing) but only the factorio developers can really define the "desirable" part.

[malventano]

Due to the way that flow exponentially falls off over the first few pipes after the pump, you can't split off to multiple pipe runs and get full flow. Using your example, you can't just keep doubling up on pipe runs and reach the 12,000/s rate of the pump because there is only a single outlet from the pump discharge, so using standard pipes you would drop too much pressure just trying to split out to the runs needed to achieve higher flow rates from singular sources.

Why care about such high flow rates? Because the folks who have such a high fluid demand are the same folks who have been coming up with 'creative' ways to move fluids at a higher rate than possible by pipes.

Why compare pipes to the throughput of the pump? Because the pump is pumping fluid through the pipes. Not everything in the game can be direct tank to tank.

Doesn't that depend on where and how you do the split? eg use 2 pumps to pump out of a tank, that way each pipe should be able to get full flow. And in almost all cases what you care about is the fluid flow rate. Comparing pipe flow rate to a pump is to a pump is silly, because pumps are cheap, and you can easily have one (or more) pumps per pipe.

Let’s say you want to achieve 10,000/s flow. You’d have to have a source, let’s say a tank. Tanks only have 4 connections, and one is likely the input, so that’s only 3 outputs. You’d have to pump to 3 other tanks and split off of them with additional pumps. That’s an awful lot of convoluted effort at both ends just to ramp up flow rates to higher levels. As others have said in this thread, ‘just add another parallel pipe’ is not effective when you can’t split off of the output of the primary fluid mover in the game (a pump) without a convoluted mass of pumps and tanks.

[malventano]

Due to the way that flow exponentially falls off over the first few pipes after the pump, you can't split off to multiple pipe runs and get full flow. Using your example, you can't just keep doubling up on pipe runs and reach the 12,000/s rate of the pump because there is only a single outlet from the pump discharge, so using standard pipes you would drop too much pressure just trying to split out to the runs needed to achieve higher flow rates from singular sources.

Zavian already provided a good answer to this. The question is: What are you trying to optimize here? Surely not 100% pump utilization itself, but pipe throughput over a non-trivial distance, right?

Now, if energy and resources don't matter you just could fill the whole distance with pumps (no pipes at all) and get about 8000 units/s out of that single pipe track. That is what seems to follow from viewtopic.php?f=5&t=46030.

Any design that doesn't use a ridiculous amount of pumps might use a pump every ~150 pipes (or underground pipes) and still achieve over 1000 units/s over one pipe track .(https://wiki.factorio.com/Fluid_system) (T=tank, P=pump, o=pipe).

TPooo....oooPooo...T

Now if you need to double the throughput, you simply do the following:

oTPooo....oooPooo..T
o
oTPooo....oooPooo..T

Lets check if that is really reasonable: To achieve the same throughput with more pumps on a single track instead you would already need a pump every 2 pipe segments, i.e. PooPooPooP, clearly ridiculuous over long distances.

Either tell me what is wrong with my calculation or what is wrong with my priorities/goals.

Why care about such high flow rates? Because the folks who have such a high fluid demand are the same folks who have been coming up with 'creative' ways to move fluids at a higher rate than possible by pipes.

Everyone cares about how much he can transport over a distance. But no one should care about " reach the 12,000/s rate of the pump" as a goal by itself (and that is what your answers up to now boil down to). Usually in any chain the weakest link defines the limit not the strongest link. If "creative" ways achieve more, good for them. But your words still seem to be just another redraft of "It used to be like this and I don't want change".

The only relevant question we can ask is: Does the highest throughput you can achieve with any method (pipes, train,belt) prevent something desirable anywhere else in the game to work (and be fun)? It is possible for us players to define/declare the "prevent something" part (through play-testing) but only the factorio developers can really define the "desirable" part.

The natural way to calculate maximums in this game is to figure the most you can get out of any one miner, assembler, refinery, etc. Getting the most out of a pump only makes sense, but if you prefer, choose an arbitrary figure, say 8,000/s. That is not achievable by ‘just splitting in two’, as the split itself cuts flow down to half of a pump output.

[dasiro]

my feeling is that the flow rate of a pipe decreases too fast. Simply look at real-world pipelines stretching hundreds of miles with only a few pumping stations.
the capacity should be limited by the pumping power and the pumps should create a boost above the normal output rate, now it feels like the opposite

[SuicideJunkie]

People just LOVE OP things. Give em barrels with 10k contenance, and stacks of 50, 90% will bless you, and then yell at you that pipes are SO underpowered and need a buff, and all other items need a buff too, because you can't convey 1k plates as a single item.
I can remember of only one thing that was consistently asked for a debuff in this forum, and it was mostly because of the brainless ease of use of it, more than sheer OPness I'm sure the regular contributors will recognize it at once.
People viscerally hate nerfs, and one needs a serious sense of honesty to say "Yes, that nerf hurts my ability to do OP things, but seen as a whole, the game is better balanced now".
I'm not saying I'm describing 100% of the people here, but I think I'm well into the vast majority.

People will do those things everywhere.
Years ago on the SE4 IRC channel, we even ended up with a set of bot triggers to make fun of people that would multiply the size of both vehicles and parts by 100x and thus change nothing at all except for causing integer overflows.

[Zavian]

The natural way to calculate maximums in this game is to figure the most you can get out of any one miner, assembler, refinery, etc. Getting the most out of a pump only makes sense, but if you prefer, choose an arbitrary figure, say 8,000/s. That is not achievable by ‘just splitting in two’, as the split itself cuts flow down to half of a pump output.

I guess I consider pumps and pipes in the same way I consider inserters and belts, for most builds I don't care about maximising an inserters throughput, I just use inserters as appropriate to get the job done. At a loading/unloading station I do want to minimise the amount of time the train spends loading unloading, so I do use stack inserters directly into boxes, and pumps directly attached to tanks. But when making red science, I don't add modules and beacons so that I am maximising the inserters swing rate of 2.3 swings per second. Nor do I say I'm feeding this 0.5 yellow belts of copper and 1 yellow belt of iron, so I need to keep adding assemblers/speed modules until they produce at 6.66 flasks per second. No I choose a target production rate of 0.75/sec or 1.5 or 2.5 or whatever, then I build the right number of assemblers, with beacons and modules as appropriate, and add input and output belts as appropriate.