Belts vs Bots - A response to FFF #224

[mrudat]

What would infinite research for belts look like? We have mods for infinite train and inseter research, which both amount to more items/second. There are compression mods, but they also impact bots and trains...

[Hannu]

TL;DR

One way to nerf bots would be by limiting how many can access a container at once.

I would prefer to give bots hitbox and collision avoiding. That would not affect to building (except huge things), mall feeding or player servicing operations, but make it impossible to build effective bot based transportation to main material flows. Also, the movement of bot flock would be visually awesome, if bots had to avoid tall objects, like power poles and high buildings.

However, it seems that huge but simple bot based megabases are the most popular youtube videos and devs want to keep it at most important developing direction. I do not believe that anything interesting which would nerf bots or consume significant amounts of CPU time can be implemented in game, even probably very small minority of players actually build such bases which gets benefit.

[TruePikachu]

The thing is that if you give the bots hitboxes and the need to avoid collisions, you're going to slow things down a lot. Collision detection and avoidance is relatively expensive, especially if you have tens of thousands of entities that all want to take the same path, and also need to run pathfinding to avoid crashing into each other. We have enough lag issues with the biters, and there's magnitudes fewer of them.

[Koub]

It would be nice something that hasn't been already suggested may times would come up. But I don't think it's possible, because people will not read the thousand posts about the bots vs. belts depate.
But there is a TL;DR :
There is no optimal balance point. It's just a matter of preference. And as such, any suggestion to "fix" the balance is bound to fail : there's nothing to fix.

[DustFireSky]

Each of us must decide how to play the game. If someone play the game only with bots. Why not ? The counterpart is the energy hungry bots to supply. Belts don't need power. It's personally not a way i love, cause in my opinion bots destroy the creativity to build production layout's. I prefer belts! Always! In some situations i build a little setup with request boxes etc. For example the satellite. Bots deliver the stuff. Easier to build as for satellites a belt paradise

[TruePikachu]

In my opinion, we should get belt item stacks, but with a limitation that stacks can only be put on a belt "at once"; additional items can't be directly added atop a stack, and the height limit corresponds to the inserter stack size limit. Some effects this could grant are:

• Belts can get a massive throughput boost, if designed correctly
• Belt-inserter throughput would increase as well (potentially to chest-inserter levels)
• Belts can get more interesting design patterns (e.g. if a machine can't produce 12 items in the time it takes a full inserter cycle, just repeating the machine won't allow you to entirely fill the belt: you'd need an assembly that uses power-hungry stack inserters to merge stacks together and create new stacks)
• Express underground belts can be used as a substitute in some forms of dense machine/beacon arrays for the logistic chests

Bots wouldn't be able to directly benefit from this buff, while it improves belt throughput per UPS impact by a factor of 8-12 (depending on how inserter logic changes). While the largest megabases would still go bot-only to avoid some of the unavoidable pitfalls belts have (e.g. underground distance limit of 8 prevents super-dense lattices of beacons), it would certainly level the playing field before then, making both bots and belts viable for smaller megabases (either going the expensive bot route which needs a lot of power, or the cheaper belt route which needs far less power but more thought in layout, while also being more consistent and reliable with deliveries).

[Italian Cuisine]

It would be nice something that hasn't been already suggested may times would come up.

A more oblique point was raised once before, but didn't gain traction. I whole-heartedly agree with it: The problem rests with BEACONS.

Rows of beacons are the only reason we can't use more parallel belt systems in later game factories. Once you get to that stage, the beauty of interesting and interacting 2D organization goes out the window, and you're constrained to these independent narrow rows with very little flexibility in order to keep up with production needs. That's one major factor that tends to ruin belts' usefulness within the assembler area itself.

I'd suggest removing the stacking effect of beacons, and maybe making the beacon effect wider. Make levels of research or higher tier/count modules for individual beacons that could raise assembler stats to the equivalent level that stacking currently gives, but without the ruinous constraints on factory design. They'd still have to be incorporated, and constructing at this level should still be equally expensive, but it wouldn't lock you into narrow bot-preferring rows to gain benefit.

[DustFireSky]

Rows of beacons are the only reason we can't use more parallel belt systems in later game factories. Once you get to that stage, the beauty of interesting and interacting 2D organization goes out the window, and you're constrained to these independent narrow rows with very little flexibility in order to keep up with production needs. That's one major factor that tends to ruin belts' usefulness within the assembler area itself.

For me it's not true. U can build also with beacons pure belts productions. U must build bigger! That's the only difference. Sure, between the beacon and the assembler machine must be maximum 2 tiles, but it's possible. U need more creativity to do this. One straight row of assemblers doesn't works anymore. U need some spacing between to make it.

On the other hand, u don't need beacons. U can reach the same output with just more assemblers contained. Into the lategame i use beacons too, but with belts!

Bot's are nice to supply yourself or build big areas, but i personally don't like pure bots deliver production sides. It's a little bit like cheating. Just my humble opinion. And a pure bot base looks very very boring. The beauty contains belts. Simple, but sure

The moderator has told they don't fix something. I think, it's right. We don't must fix something. It's by yourself to fix the issue through creativity

[Italian Cuisine]

I don't use bots personally, and do constrain my building to getting beltworks to feed beacon rows. But there's much less scaling and flexibility in the core design itself there, and it's impossible to use belted fully-beaconed assemblers in anywhere near the same design efficiency as non-beacon belt builds. Bot builds & designs are comparable before and after beacons.

It would be interesting to see how large a non-beaconed setup is that does 1kspm or more. I would guess that it would start to impact performance on many machines. Beacons are supposed to allow you to scale beyond those limitations in late game, but do not allow the interesting & efficient designs of how the game was up to then. Again, they end up promoting the use of bots, not strictly enforcing them.

[TruePikachu]

Beacons are going to be the cause of at least some late-game issues, so probably the first goal would be to make it so belts and bots are on even footing with regard to machine access past beacons.

If nerfs are permitted, I'd possibly say to limit the number of modules that can affect a machine to 16; this is six vanilla beacons paired with a tier 3 assembling machine, which builds in a straightforward manner. If a machine can't have modules inserted but can get beacon effects, that would be eight vanilla beacons, which is still relatively straightforward to build for with belts. Of course, this won't make everyone happy, as it limits the potential designs for beaconed setups.

Maybe a buff for things could be to allow transferring items through beacons, to some degree. I was at first thinking about beacons allowing belts to go through their center, but I then realised that an alternative could be to give beacons a small bit of chest inventory room. For vanilla, at least, this would be a buff for beacons overall (as designs could potentially use slightly less space if items are being transferred via the beacons), but in doing so it would make the space saving feature that bots get in such dense designs more-or-less a moot point; if the input and output is external to the assembly itself, the space that belts take up doesn't even matter.

[DaveMcW]

I agree with DustFireSky, machine access is not a problem with beacons.

However beacons do introduce another challenge for belts to overcome, while bots can effectively ignore them.

[TruePikachu]

I mean, strictly speaking you can get equivilant machine access with beacons on a single machine; the issues with access arise when you change from a linear construction to a lattice in order to best take advantage of the beacon bonus area, as belts aren't quite designed for that use case. You end up (in dense configurations) with a machine surrounded by beacons which themselves are surrounded by machines and beacons -- and no way to route the underground belt without losing some beacon efficiency.

When you also consider that beaconed layouts are also frequently faster than nonbeaconed layouts, you can start running into throughput issues, which require either greater belt throughput or more belts; the latter of which can be impossible to implement without losing beacon efficiency. Or, you can just use a swarm of bots and avoid the logistical problem entirely.

---

Typing the above helped me realise that the bots actually are a useful thing to have around, since it can help to reduce the amount of belt design needed, especially in these sorts of situations where you have very little space to work with inside an assembly. But, the bots can be used for everything, completely avoiding the belt puzzle. While belts are free energy-wise, and bots consume a lot of energy, these factors don't matter for the solar-powered megabase. While belts have better throughput compared to small numbers of bots, since the only space that bots take up in crowded factory areas is the space for their chests and roboports, inner-factory throughput issues can be remedied just by spamming more bots. And since the bots can be used for multiple different kinds of items, those bots you spammed for the "main bus" can also help out with side jobs, e.g. unloading trains or refilling turrets -- a constructed belt section rarely can see this kind of multipurposing.

Perhaps bots need to be made more expensive in running costs; I had initially typed out an idea for them to consume lubricant, but since nothing else has a similar form of maintenance, it would stand out a bit. Maybe it doesn't matter, though, since bots otherwise can serve as a universal solution.

[Hannu]

The thing is that if you give the bots hitboxes and the need to avoid collisions, you're going to slow things down a lot. Collision detection and avoidance is relatively expensive, especially if you have tens of thousands of entities that all want to take the same path, and also need to run pathfinding to avoid crashing into each other. We have enough lag issues with the biters, and there's magnitudes fewer of them.

Yes, you are right. I would prefer few hundreds of effective bots which seems to be real machines instead of tens of thousands immaterial ghostbots. In my opinion bots should be only for construction works and small scale irregular transports, not for main material flows.

There are no really lag issues in factorio at all. The game can handle incredible big system with cheap desktop computer. Of course extreme players would always find and break limits, but increasing maximum possible system would not bring any new interesting elements in the game. Just more brainless copying.

[Hannu]

It would be interesting to see how large a non-beaconed setup is that does 1kspm or more. I would guess that it would start to impact performance on many machines.

I am currently building a beaconless belt base intended to 500 spm. If I remember I can check some numbers at evening when I play. It is buildable, but I have been frustrated tens of hours and thought that 1/4 had been more entertaining level. But of course it is nothing compared to largest megabases in videos.

[audigex]

To me this whole question is even simpler:

Belts are the fun early-mid game. Bots are the late game convenience.

Factorio is a sandbox - if I don't want to use bots I don't have to, but like so many other lategame tech they exist to make life easier late game when the player is focusing on the "big picture" of how to expand.

For the same reason I like mining productivity (less time spent finding new ore patches), artillery/nukes (less time spend slowly killing biter nests), and nuclear power (less time spent masses of solar fields), I like bots

If bots are removed from the game, literally the first thing I will do is add a mod that puts them back in... but then I'll lose any future achievements, which will be a shame.

If the devs want to make a "harder" mode, then sure, make a harder mode - disallow certain techs, reduce the effectiveness of others etc - but I think it would be a mistake to remove bots from the base game

[Hannu]

For the same reason I like mining productivity (less time spent finding new ore patches), artillery/nukes (less time spend slowly killing biter nests), and nuclear power (less time spent masses of solar fields), I like bots

In my opinion bots are too easy. Heart of Factorio is connecting material flows correctly and effectively with belts (maybe electronic engineering background have something to do with my Factorio addiction). Bots make everything trivial copying of simple basic blocks.

Do you think bots are necessary? How immaterial magic bots are better than for example teleporting chests would be? Current animation is just annoying. It is very traditional game programmers lazy cheat to let game objects go through each other. It was annoying already in 80's, even it was more or less only option when you had 1 MHz 8 bit processor.

If bots are removed from the game, literally the first thing I will do is add a mod that puts them back in... but then I'll lose any future achievements, which will be a shame.

They will not remove bots or nerf them. Majority of players are used to easy bot solutions and it would be economic suicide. This is just harmless small talk and not severe development suggestions.

[planetmaker]

Construction bots are definitely more than a convenience: they keep the game playable beyond midgame, removing the grind, helping players focus on what's important.

Logistics bots are something in between, but mostly a convenience:
Logistics bots which supply the player with whatever s/he orders for his inventory or disposing of trash: again it keeps grinding away and helps players focusing on the important aspects.
Logistic bots which are used in a way that you simply setup factories, miners, stations, etc, each with its own requester and provider chests... that makes things too boring for my taste, reducing difficulty to basically zero. I'm always saddened when I see these "constructions" appear on multiplayer maps (any good coop MP servers where such excessive use of bots is banned?).
Yet... the time you can actually do that from a ressource POV is usually around the time when you can launch your first rocket... so it's a convenience to offer different play style for the late / endless game, expanding the player base.

[Mmmmmmorten]

Speedrun: I’ve never seen a fast speedrun of Factorio that includes bots and I don’t think I ever will. Bots are not a dominant strategy when trying to launch a rocket as quickly as possible as the cost to acquire bots does not pay itself off.

Nice post and comments... guess this will alway be a topic where we just have to agree that we all will have different opinions...
Easy solution - if you do not like bots - Don't use em!

Seen a lot of speedrun's as they first of all are awsome and second give some very good ideas for what to do and not to do in my own builds.
Most speedruns in 0.17 are building with bots (very slow bots). See link to youtube video of Nefrums "Factorio 0.17 Any% speedrun WR 2:46:40" here.

it's almost 3 hours, but amazing to launch a rocket in that time... I usually have red & green science at this time

/old dude

[cappie]

What about longer charging times for bots, so that they're only effective when used in bursts, like when resupplying a train or a player, or helping in the production of some finite-storage product like repairpacks or complex ammo types?

I really think giving bots realistics recharging times and wear over time, especially when a bot's usage approaches >80% of the time... and maybe a system for them to have changeable batteries as a tier 2 system, would be nice. It would add some more realism and would cause bots not to be used ALL the time.

I'd love to see overused bots fail, crash and drop their load of items on the spot where they crashed; causing mayhem in a large factory, thus teaching players to use them only in burst-situations.