Belts vs Bots

[jmsbrone]

I feel like I should state my personal opinion about bots vs belts before suggesting changes.
When I started playing Factorio it was fun figuring out different things, setups, belt spaghetti, etc. Since I was playing by myself and figuring out all the things by myself I didn't know of bot-based factory, to be honest. I did get to bots and used them for simple logistics around the initial base and all the way through several campaigns I used belts for everything. When I decided to go on bigger scale and make it into megafactory all of my stations used belts. Iron station was (I think,don't really remember that well) 8 blue belts with speed modules and it was fun taking module effects into account to know how many furnaces to put to eat and produce a full blue belt. But then it started to be not enough and I rebuild the stations to have 16 lines, 8 on each side for input/output 2-8-2 trains.
But I think about that time I started to run into issues with preffering belts. I haven't watched any YouTube videos at that point, so I still didn't know about bot-based megafactory, but I've started to experience issues I didn't really like. I think UPS started to drop, buffer setup before and after each furnace column weren't working as expected, some chests were preffered over others, belts weren't saturated, some end furnaces started to accumulate iron since inserters couldn't place it on the belt even with underground belts. And the entire station with 16 lines was just huge in size, but my issue was mostly WHY ISNT IT PRODUCING 38.4k/min with full input? That's what 16 blue belts should do. When I looked at production it was like 38.0, or even 37 sometimes dropping to 36. It never had a stable 38.4. And that's an OCD-trigger. I know now that were probably some workarounds and solutions for saturation, but I didn't know back then and none of them are intuitive when you need something for the first time on your own.
I got annoyed a bit and I think that's when it went south. I started to watch YouTube videos and saw bot-based builds. They were elegant, small, simple and producing consistenly more per required space for them. After I've built my first ones I realized belts suck for big production. They need too much maitenance (insertertes, underground, buffers, belt balancers (which aren't intuitive as well), train input balancers) and any belt-based build requires a lot of space for big factory. Even with space not being a problem - it still is when you have to go warfare for hours upon hours to clear land. And for belt-based stations I'd need 3-4 times the amount of space. Don't get me wrong, I love a challenge. But I do not like doing same thing over and over again. It's not a challenge, it's just tidious and boring. Clearing a lot more space for belt-based builds that don't even work properly - it's not really something I want to do.
After 0.16 came out I, as usual, started a new campaign and I immediately noticed belts sucked more because inserter-to-underground stopped compressing the belt. Even with belt optimizations I knew immediatelly I'm not even gonna give it a shot since it triggers OCD practically 30 minutes in a new game. I'm now at the point of having all bot-based and the only thing I use belts for are miners because bots flying over entire patch is just a waste of energy and doesn't look nice. Belts to the stations, then to providers and then requesters to trains. And even that isn't great because belts aren't saturated. Even if I were to find a way to have exact X number of miners per belt to saturate it - next Miner Productivity research will break the ratio.

So, here's some of my points:
- Required space.
Bots: Less space. Any assembler/furnace/etc. just needs 2x2 tiles in front of it for 2 inserters and 2 chests. Clearing space on one's own is tidious, I'd prefer to spend less time doing it. I don't even want to think how much space I would need for my current megabase to have it move to belt-based. I already had my share of having to clear biters on High since I wanted them to be more of a problem to play around with different defenses setups. Plus, each station needs 1 input train-to-providers and 1 output with product.
Belts: For 1 input recipe (or 2 if demand is low enough to put 2 items on 1 belt) required space is already either inserter+long-handed and 2 belts on one side, or 2 stack inserters+2 belts on 2 sides for high-throughput recipes. It's okay for up to 4 ingredient recipes, but for 5 it starts to get more complex when you want to have maximum throughput. Plus, each station needs at least one input station per each of recipe igredients.
- UPS. Goes without saying.
- Template building.
Bots: Each bot-based build can be made the same if all requesters are wired to a constant combinator. Then it's just a matter of changing combinator settings for each new station. And that's a good thing for me since it's progress. I didn't get to the end game to figure out different designs for making stuff, I had enough of it in the beginning/middle stages of each campaign.
Belts: Each station will have to be built from scratch since different ingredients are needed in different ratios; different number of ingredients; different stations for each input ingredient since there isn't a way to re-route items from belt to belt unless doing some circuit magic. Scaling production also means figuring out yet new design.
- Time efficiency.
Bots: Less time to make setups. Plop down a template, change settings, setup trains and good to go. Exactly what progress feels like when I don't have to do much.
Belts: Much more time. Prop down a template for required number of ingredients, change up how many belts go from input depending on what items are needed more, figure out if there should be different ratio of input trains. I had 6 input stations with copper, 4 with iron for my old green circuit belt-based build.
- Buffering.
Bots: Buffer can be set in requester chests. Also, each inserter can be conditioned for logistic network to stop working when amount is reached. Each bot-based build will have exactly as much as I want and it will stop itself from making more. It can have just a couple over the limit. And it will have some items in requester chests. If I take all logistics out of the area - machines will burn through what they have in their buffers + what's left in requesters and in most cases they will have a couple of spare items since requester chests respect ratios. That's efficient if I want to rebuild something.
Belts: each piece of belt is a buffer for 8 items. Recipe ratios are sometimes very different. One blue belt can be eaten fast while even yellow delivers a lot more than needed. If I want to rebuild - I'll have to deal with a lot of spare resources in my inventory and run up and down to move them somewhere. Belt-based setups do not respect buffering ratios and are really tidious to rebuild. Yes, point can be made to build them perfectly once and for all, but that is rarely the case. Train traffic management, OCD, proximity can require any station to be moved anywhere.
- Production scaling.
Bots: Any station can be scaled by adding more of the same until hitting a point of bots travelling too far and being inefficient, in which case a new station will be build. But until that point each station can be scaled independently of trains. It'll be 1 input train-to-providers and 1 output. If input is low then more trains can be set for delivering items.
Belts: There's a number of inserters to saturate a belt, so number of belts coming from input station is limited to number of wagons for input trains. If maximum input from 1 station is not enough - new station will have to be build and whatever belt balancer I had there will have to be x2-to-x2 which can and most likely will cause problems. Scaling is really problematic.

There're probably some more things I can say, but this post is getting too big. My points against belts imply what can be done to improve them, but I have some more ideas. So, here's my list:
1. Configurable belt speed. Instead of having 3 tiers it'll be easier to just setup ratios with belt configuration. If I need 1:10 ratio I'll just make blue belt carry 4/s instead of 40.
2. Intuitive or built-in belt balancers. If you have to go to copy something from somewhere to play the game better - it's not a good approach. Belt balancing is a big thing. It's not hard to setup 2^n balancers, but it takes a lot of time. Any other balancer will take much more time to figure out on one's own.
3. Production line. Instead of having 1 item being made per cycle it can be done in stages. In any factory there's a conveyor belt with items that get added-to/changed/pressed/etc. It's 1 conveyor belt that has same item being manipulated with by different machines. This way each stage can have 1-2 items to each machine for that stage and any recipe can be scaled indefinitely. 1 machine will take item from that belt and items from other belt, craft the item to next stage and put it back down. This makes belts a requirement for some recipes. But this will require to make new belts that move in bursts instead of continuous and sync up belts with machines.
4. Item weight. If bots are good for throughput then they can be less good for things that don't need high throughput. If bots have weight restriction instead of stack size they won't be able to carry heavier items like rocket silos, engines, etc. They shouldn't do it as it is if we start to apply logic, but everybody's used to it, so fine.
5. More in-between recipes. If something needs 5 ingredients it can be split into needing same amount of items but shared between 3 stages recipes: 2 items combine in 1, 3 other combine in 1, end recipe is just 2 items. Having 5 items come to a machine is okay when you have a couple of them. Not dozens.

I can go on, but it's already too many letters. Summing it up, bots feel like end-game, belts give a feeling of being in the same place as 100 hours ago. The only difference is that they are blue and faster. It's more of a loyalty test. Whether you want to stick with something older and inferior vs AI-bots simplifying your life.

[Lez B Ann]

TL:DR

[Tarka]

Interesting ideas here - a few points:

If item weight is used to limit robot capacity, balancing item weight so that players do not have to memorise a different stack size for each item based on its carrying weight would be very difficult. Perhaps for simplicity's sake have all raw materials weigh one amount, all plates weigh another amount, all intermediate products again have the same weight, etc. Trains seem to be already affected by cargo weight reducing their acceleration, so this addition would make sense.

Built-in balancers - I presume you mean a device which has a belt input on one side and an output on the other side which simply takes the input to produce a balanced output? Good idea, but likely to cause many complaints because 'its too easy'. Also, I do not think it likely that a simple balancing machine would be able to replace player made balancers when trying to balance the throughput of many belts side-by-side, instead only serving to balance the two sides of a single belt.

Perhaps also a belt compressor machine with much the same external design, that has an internal storage buffer into which it takes items as they come and releases them at exact intervals that space them out an exact number of item lengths apart, leaving gaps just the right size to insert more items further down the line. Again, I would expect many players to make the 'too easy' complaint.

Changing manufacturing to have more intermediate steps and multiple possible combinations of input products is an optimiser's dream indeed, but I do not think that it could be done without increasing the complexity of manufacturing to a point that most players would find unacceptable.

Syncing belts with machines, making belts able to move in bursts, or throttling the speed of belts. seem quite unecessary. If there is nowhere for items on a belt to go, they just sit there while the belt slides past underneath them. Items move at their own pace, the belt speed is only a maximum speed. If belt is faster than it needs to be, the item flow just starts and stops instead of being smooth.

Perhaps if the sight of a belt moving under stationery items is unpleasant, then the belt movement animation could be made to stop if that belt segment is covered in items going nowhere. However some real conveyor belts do move non-stop and just skid under the items on them if the items are not able to move. Such as those used to move parcels in mail sorting centres. Depends on whether or not the items the belt is designed to move would be damaged by having the belt move under them while they are still. So we can simply assume that Factorio belts are made in such a way that moving under stationery items causes no damage.