Hi,

I did some test as yield from this was unnaturally low. And I think default settings are just not right. I converted 1 million of ores with 2 different tests.

1. quality condenser test

Outcome is as expected with 1 legendary module, 4 steps is: 1000000 * 0.0625^4 is about 15, and indeed data was converging to 15 legendary ores per 1 million normal. So that's 0.000015

1. 4*legendary module with single recycles test: chest -> recycler -> chest.

and the yield was as follow in order of quality: 188k 56k 5.5k 535 61
which matches the math on https://wiki.factorio.com/Quality

If we upcycle further so that we are left with only legendary, the final yield is about 350 ledengary ores, or 0.00035 which is very close to 0.000375 that factorio wiki specifies.

So not only you get so 4 times more raw legendary from recycler, but you also have other qualities that you can upcycle again and get more legendary from 1 million normal ores. So ultimately quality condenser on default settings is more that 20 times worse than brute force upcycling.

Summary:
Changing number of module slots from 1 to 2, yields about 244 ores. It's still less than theoretical 375 from recycling, but it's a fair price to pay for convenience and UPS save. Setting base quality to 1.3% will get you closer to recycler to 362 legendary ores.

Now I'm not mathematician, so you may want to verify it. But it seems my practical tests do match theoretical calculations.

ps. And thanks for great mod. After solving upcycling once in my life, I don't really want to go back to it anymore xD

interesting, i figured that since a recycler has 4 module slots and a 1 in 4 chance to output that having just 1 module slot and a guaranteed chance is fairly identical, but i suppose i made a mistake somewhere in that thought process?

Yeah. I think thought error was that, recycler upgrade by ONE TIER ONLY. So if you put common ore, then you can ONLY get up to uncommon and nothing higher. But you get in about 75% common item back, and the rest 25% is split with other qualities. So you have 22.32% chance to get uncommon, 2.232% for rare and 0.0248% for legendary FROM A SINGLE PASS. With condenser you can ONLY get to uncommon in single pass. Then things just compound into huge error, since recycler "bonus" is applied to every upcycle.

Then there still may be some cases where this will fail. I mean, your current condenser will work just right when you feed machine with epic items. You will get 6.25% of items back as legendary, and with recycler you will get 0.25*0.25 - so same value. I think my math and 2 module only will work well if we assume we always feed machine with common and get back legendary. If you recycle up to say rare, and then you feed into condenser with "my" math, I think you will get higher yield of legendary. Which means 2 module condenser is cheesable.

But then again, if you gonna cheese like that, you can just straight away get some "cheat" mods lol.

Came to post the same thing. Per the wiki, quality outcomes normally follow this distribution for normal to legendary respectively:

1-Q, Q9/10, Q9/100, Q9/1000, Q1/1000

So a recycler with 24.8% quality (4 x legendary QM3) from 1000 inputs produces:

188, 55.8, 5.58, 0.558, 0.0558 items vs quality condenser generating 62 uncommon and nothing else.

These then typically get recursively upcycled until void or legendary which throws out the ratio even further. For instance even if just those 188 items are recycled again it will already outpace the uncommons from quality condenser and this just gets worse with each recursion.

Modelling this loop to completion in YAFC yields ~3.67 legendaries per 10k inputs while quality condenser requires ~67k inputs to deterministically output a single legendary. While it would probably be fair to balance in favour of not quite being as 1:1 efficient as the recycler because its more convenient for items that don't recycle into themselves the default is undershooting by being dozens of times worse (~24x).

This is also before considering that brute force upcycling something that recycles into itself is the worst possible outcome and all the processes people actually use are dozens of times more efficient so in essence this is actually 100s of times worse than any process someone would typically implement in unmodded SA.

There are some knobs to tune this obviously but I'm not sure exactly how it'll go yet.

My solution is to simply increase number of modules to 2, and set base quality to 1.3%. And then feed ONLY common items, and extract legendary.

This will give you roughly same results as brute force upcycling.

Putting anything higher than common, will results in you getting more high tiers items than brute force method.

Just doing some math but seems about right, by my reckoning 2 mods is slightly worse & 3+ just becomes exponentially better / is equivalent to entire multi-stage upcycling processes which is becoming unbalanced in the other direction. Figures below assume 0 base quality and legendary QM3s.

Brute force reference rate = 1k:0.37
1 mod = 0.062 = 1/16^4 = 1/67675 = 1k:0.015
2 mod = 0.124 = 1/8^4 = 1/4230 = 1k:0.24
3 mod = 0.186 = 1/5.4^4 = 1/836 = 1k:1.20
4 mod = 0.248 = 1/4^4 = 1/264 = 1k:3.78