Crops Broken?

[DYefimov]

I'm posting to social like once in a year, but this thread is definitely worth it!

Stumbled across in a search of why the heck A+S is not showing crop quality albeit wiki says opposite.
The reading was entertaining with ever growing curiosity...
...nearly screaming at the end: why everyone is hand-guessing the science never mentioning things like law of large numbers, statistical distribution or RNG implementation?

And here it goes, at the end - it was a catharsis.

Can I compliment to further reading some further watching, as it may be more entertaining to someone.
It's about 1 hour long some what like documentary with a couple interactive pages to see all of the stuff for yourself.

Actually, always suggesting this course to ppl far from the field - who are not yet fascinated with the simplistic powerful magic that our life is driven by.
Of cause - the core of it being Bernoulli trial and Binomial distribution (which is covered by Origin of Markov chains video in this set).
You'll see for yourself that magic exists and understand how they search for extraterrestrial activity at SETI.

Enough of lyrics.

Regarding crops.
As a 20+years professional C++/math/R&D, my best (low-informed) guess will be:
with our crop fields we are struggling against the law of large numbers and pseudorandom generator here.

In computer software the module that generates what is called RNG usually utilizes Pseudorandom number generator capabilities.
Which is not even near to any kind of fair statistical distribution.

There are some other types of generators, e.g. hardware random number generator (true random number generator, TRNG) or may be some software gaussian e.t.c., which can give you such distribution.
But I doubt that's the case here.

To keep it simple and brief - any % displayed in an item tooltip can not be relied on.

But is this the case?

Let's look at our field: each crop harvested is a trial against pseudorandom generator.
If we have a small field - amount of trials is small (population), and it can roll 9 out of 10 -quality with 99% +quality probability displayed.
But the bigger our field is the more trials we have, kicking in the law of large numbers, which will alleviate the pseudorandom generator quirks over time.

I will not state this 100% true, and will not dig any deeper.

But this is an interesting paradox to think of:
two not interconnected systems in any kind of straightforward way I'm trying to describe here, are affecting each other, showing us the true nature of crop harvesting problem!

Smile often!

[sMartins]

Interesting...but now farming is being restored to World 9 system, otherwise we could make some test in the game....it could been nice....
P.S. Well we can try with dices....they are still in the game....some volunteers to throw the dice 1,000 times and make a graph?

[DYefimov]

Wait what?

Anyway...

I don't feel strong enough to put full explanation in plain English here, as it will take quite some time and effort.
Here is the brief one:

I spent half a day modeling and experimenting with all this stuff...

3.png

...and it brings us to several interesting conclusions, being:

1) law of large numbers proves to be working just fine with pseudorandom number generator
2) the deviation is quite big on a small sample sizes (small crop fields)
3) ~100+ fields alleviate the deviation, and we can expect approximately the same amount of randomness as we see in real life

About new system:

Removed generational quality randomization per plant, and thus seed sorting, from farming, and returned to the previous W9 quality randomization mechanic per locality and point in time. Removed inspect feedback on crops. Remade crop quality fields to crop speed fields, so that crops now grow with somewhat varying speeds in varying locales, rather than have their qualities affected by the crop fields.

If I got this right, it does not matter either we plant 1 field or 10 of same quality, cause "Removed generational quality randomization per plan".
Meaning, we can run an experiment every crop life cycle (i.e. being several days), hence to reach 100+ count experiment zone ("real life randomness" zone) we need several hundred days?

Now imagine there is some reset counter - we plant 1 field of same crop+quality every, le'ts say, half an hour, and we get different quality outcome on harverst.
As it were in some previous implementation (we A+S 1 crop, wait 5 hours, rinse n repeat)
Thus we can run our experiment more often, bringing "real life randomness" zone closer in time to us.

Hey! Was it tedious to sort seeds? How about replanting 1 crop of every type every half an hour?

Actually, I was thinking of building a full probability model here, that...
...given the amount of fields, starting seeds amount/quality, "quality reset" timer, amount of time spent tending to crops, and some other parameters...
...can build a graph of: how probable the quality of a seed in your seed bag now (i.e. calculate probability mass function).

Not quite sure: a) if we have "quality reset" timer b) how to incorporate seed amount into the model
So... it's time consuming... should stop myself of being distracted again...

One interesting thing to mention here is:
probability mass function the model is based on is analogous to what we can see in real life:
- you have a bag of same seeds
- you can not inspect every one of thousands in detail
- you can expect with some probability what's the quality of seeds inside based on previous harvests
- when you plant - you roll against this function for what quality of seed you pull from the bag
- when you harvest - you update the prob mass func with quality of seed you push into the bag
- prob mass func always sums to 1, thus scalable to any amount of seeds it is describing

I'm not aware of any way to evaluate this expression non-iteratively

That's being introducing the probability distribution on the whole field + seed bag?
Anyway, sum and product with range constraints are often used in here, and they are quite iterative

Fuh.... tired of this nonsence...
I wonder - will someone read this brainfudge till the end? )))

Science for the win!

[sMartins]

1) law of large numbers proves to be working just fine with pseudorandom number generator

Cool....btw if that function is designed to replicate random reality . I don't know how the algorithm works but probably it could fail only with huge numbers for a very small deviation, that is the intrinsic error of the calculator.... this is my bet.

Not quite sure: a) if we have "quality reset" timer b) how to incorporate seed amount into the model
So... it's time consuming... should stop myself of being distracted again...

I don't know how it works exactly but I feel there are simply different timers, where you get the quality...and probably they are not divisible 24h, like localized resources are...so they will keep rotating during the days....I expect something like that from Loftar
Someone who was really dedicated during W9 could explain to us. But maybe it's better if we don't know how it works exactly, just plant and hope for the best.

[DYefimov]

Cool....btw if that function is designed to replicate random reality . I don't know how the algorithm works but probably it could fail only with huge numbers for a very small deviation, that is the intrinsic error of the calculator.... this is my bet.

Your explanation is quite close.
One may say, it is based on laws of space-time itself.
The logic says it should not work this way, but it is... and this is truly amazing about our universe

I don't know how it works exactly but I feel there are simply different timers, where you get the quality...and probably they are not divisible 24h, like localized resources are...so they will keep rotating during the days....I expect something like that from Loftar
Someone who was really dedicated during W9 could explain to us. But maybe it's better if we don't know how it works exactly, just plant and hope for the best.

We can run an experiment - plant e.g. 48 seeds of same quality with a half an hour interval, and observe a quality change.
Based on the data collected put an interactive webpage providing probability model calculation.
Sell access for salt

No, just thinking... winers where wining about stacking? making the situation even worse, as in current implementation the time-frame for awaiting consistent result was increased by the orders of magnitude.

Mine crop field now looks like this:

4.png

[sMartins]

Yeah but being a combination on when you plant and when you harvest , good luck

Sell access for salt

eeheh

[MagicManICT]

Are you still talking about the old legacy crop system or the one that was reimplemented two weeks ago now from w8/9?

If you guys have a math lesson going on, carry on, I just don't want anyone to be confused about what's going on.

[pietin2]

I think the farming system stinks. Maybe they should emplement some type ferterlizer like they did in Salem that would make crops grow faster and give better q.

[neofenris]

To think that crops Q is given to all the crops planted is terrible, Went from 36Q to 32Q for like 100 carrots, if each carrot had a different Q I would have some 41 some 31, and that would be perfect, replant the 41, eat the 31...With the remove of inspect on seeds, how am I supposed to raise my seeds quality ? I don't get it at all

Edit the only way I can see is to plant the same crop at different time interval and pray for the best ?

Plant 10 carrots, wait 5 hours plant an other 10, repeat and hope that some of those are going to get a nice modifier and trash the other ?
Looks way too slow/chaotic/random/time consuming for me but maybe I don't get it right

[Beezer12Washingbeard]

I know this is kinda separate from the grousing about crop Q implementation, but I gotta say I'm enjoying the speed nodes. Kinda has some weird results though, like my flax is WAY faster than my hemp, at almost a 2:1 harvest ratio it seems. Gonna be a while before I can regularly smoke dank buds with my toon