I feel like the general policy of not allowing both imported blueprint and set seed is precisely to prevent the use of blueprint fitted specifically to a map, and I kind of support the idea.

So I'd expect using BP generated by a tool during the run to be considered tool assisted and would probably vote so if it's ever voted.

That being said, I also love tool assisted runs and more generally tools for factorio, and I can see how this stance of mine means one less incentive to develop tools. I guess one solution would be for Wube to put recursive blueprint into the base game :D

It's possibly reaching the same conclusion as you, indeed. I'm not saying incremental building isn't powerful, it certainly is. A 0.69 factor in time is huge and that's only comparing incremental build to doubling up every now and then, there are less incremental things to compare to.

What I'm saying is that building incrementally has diminishing return and can be totally accounted for in the computing of the theoretical fastest possible growth.

"You could even improve that to 7x with faster printing". And if you just ignore the printing cost and latency, you still get an upper bound of the maximum growth rate (bigger than the actual growth rate with printing, however fast it is). Printing speed does matter, but again, it can 'easily' be taken into account.

"a bot that travels to the edge of a roboport range from the middle, can place 1 entity and then return. The same bot, when placing entities near its logistic chest, can place up to 25x in the same amount of time." That sounds like a genius idea. I haven't speedrun myself enough to tell if it can make a big difference in any%, but it definitely could in 100% at the very least. Definitely worth adding to the opening post once tested and documented. You still have to move the machines from where they are produced to where they are placed though, in a way better than by bots. were you thinking of moving items by hands into cleverly placed chests?

I can agree to some extent with the everyone copy each other issue, but top runners tend to try to differentiate themselves, with some success (their build is suited to their strength and they make breakthrough on occasion). There are like 3 or 4 different seeds in the top 5 of any%, Anti used to use gun turret in default setting whereas Nefrums used to use Laser turrets, Nefrums uses giga buses of red belts for 100% whereas Warger uses trains (the longer the run the starker the differences), etc. Spending large amount of time off the leader board and coming up with crazy ideas and breaking the meta is exactly what Rain usually does (ok, his incentive is not driven by speedrun.com, but it doesn't really matter IMO).

I'm not sure how segmented runs would help the matter. segments are arbitrary, and hinting new runners at a single road to follow does the opposite of fostering variety (but if newcomer do needd that sort of guidance they can already choose one of the top runner to copy cat). Besides, factorio runs are not really segmentable, precisely because they are incremental: when you do red science, you're already building up resource to help you build green science. Your constantly chasing and preparing for several goals at a time.

Increments can only do so much. You can prove mathematically that if a build can double every period of time T, without incremental build, the growth with infinitesimally incremental building is in exp (t/T), in other words, your factory grows in size by 2.71 every period T, or equivalently doubles every 0.69T.

The 5-6 minutes of Nef to double are consistent with what I remember computing for electric drilling, steam electricity stone smelters and assembler1, accounting for incremental growth. If I do compute it again, I'll try to make it clean and clear and will notify you as I bet you'd be interested.

Bots do not accelerate this exponential growth, quite the oppposite. This doubling every 5-6 minutes assumes building time is free. Beyond a certain growth rate, you can no longer assume building time is free and this is where bots come in, with there own cost. In short, Bots don't accelerate exponential growth, they allows the growth to be exponential for far more longer than could otherwise possible, albeit a slower exponential growth than when building time was negligible.

Blaze, you cannot just say left and right "speedrunners do things wrong" and then shout "ad hominem" every time someone disagrees with you or suggest you do a run.

You have interesting ideas, but don't expect people to listen to you if you present them as more than that, because for now, they aren't more than interesting ideas.

Math is cool, but it's not a science, Physics is. Mathematical models are incredibly powerful to describe and predict the world around us, but they only become useful to physics after being confirmed by experiment.

That's why peope get annoyed (at least I get annoyed) when I read anyone say "speedrunners are doing it wrong", when they only have a bunch of conjectures to present. However valuable they may be, however correct they may be, however mathematically sound they may be, unverrified theory are just conjectures, and shouldn't be used to say "speedrunners are doing it wrong". Thankfully, you do use more sensible wordings usually, but I've read you once saying something close to "speedrunners are doing it wrong", or "admins are more interested in views than in being good", and it does get people off the wrong way, with no redemption possible if you've got nothing substantial to back it up (and you don't).

In any cases, people pointing out you don't have a single speedrun to your name are not attacking you personally for not being a speedrunner, they are playing some of your bold claims down, because a claim unsupported by experience is weaker than a claim which is.

For example, in: https://forums.factorio.com/viewtopic.php?f=5&t=92439 You use Dave's name and math from: https://forums.factorio.com/viewtopic.php?p=521707#p521707 to support the idea that sub 1hour run is possible. It is bogus beacause, Dave's math is wrong and way off-base, and anyway applies to a scenario very remote to actual gameplay, let alone to speedrun.

Having a prototype run on the other hand, would speak more than a thousand formula (even when the thousand formula where necessary to produce the run).

Again I love math and I enjoy following your endeavors, as you can see by my answer to your topic: https://forums.factorio.com/viewtopic.php?p=529680#p529680 But: "If you do respond tho, please stick to the math. APM, resource costs, etc. ad hominems reasoning like the above will just be ignored with an 'lol'." simply does not make any sense, there is more to speedrun than just math, intuition often beats non rigorous math, and rigorous math requires amounts of work that approach actually doing a run.

Hey, I just made a run with the following rules :

• time ends when 50 construction bots, 10 logistic chests (any) and 5 roboports are in my inventory.
• seeded, any map gen setting allowed (including peaceful)
• no opinion on blueprint (I don't import any but it could be nicer to be allowed to if a new sub category is created)
• video

Once uploaded, it'll be available here (probably in one day with my potato connexion) :

(about 2h15 1h57 in game time because I'm terrible and forgot to disable autosave and played single player, and my computer is terrible too).

What'dya think of it?

IMO it cuts any% nicely in half, requiring not so much preparation and lot of micro (as opposed to the later half of an any% speedrun) and could be a nice option for new bees (which I myself am).