Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Question about Convex combinations


narsille
 Share

Recommended Posts

Has anyone been experimenting with convex combinations of crafting materials?  In particular, does anyone know whether the resulting stats is a linear combination of the raw inputs or whether something more complicated is being used?

 

WAZ6Fov.png

"The cinnabar is a lie"

Link to comment
Share on other sites

What do you mean by convex?

 

Sorry, I sometimes fall into domain specific shorthand.

 

https://en.wikipedia.org/wiki/Convex_combination provides some useful material.

 

At the most basic level, lets assume that I was making copper bars which require 3 x (3 stacks of ore)

 

During the previous version of big world, I did a bunch of testing trying to determine item quality when I used the same quality ore in each slot.  (For example, what does an amazing success look like with (three stacks of three green copper) or three stacks of three white copper.

 

Now I am curious what happens if I start using combinations like (one stack of three green copper + two stacks of white green copper) which is a convex combination of (three stacks of three green copper) and (three stacks of three white copper)

WAZ6Fov.png

"The cinnabar is a lie"

Link to comment
Share on other sites

I haven't ran enough to be 100% certain, but I think the system is built to simply allow experimentation up to lowest quality item used on any single experiment run.

 

 

Each upgrade of base material appears to add more experimentation points.  If and experiment on green materials fails, and you used all your dots, you get the exact same result as if your experiment fails with base materials.

 

I think the resultant "bar" is still colored green for future experiment purposes, (meaning more dot's when you make a blade for example), but the base stats prior to experimentation for all materials seems to be the same regardless of material quality.

 

Would have to run some more specific tests, but I think that's the behavior I have seen.

Edited by KrakkenSmacken
Link to comment
Share on other sites

Actually in creation or rather "make" the final item quality is determined by the quality of the raw material...  if you have 3 input spots and use green in 2 and white in one you have ??66% chance of a green end result giving you 4 pip spots for sheen and durability (or even more stat properties of the item).  i haven't checked direct correlation but 3 greens nearly always gets a green product and 3 whites have a very small chance (but it does happen) of producing a green result.  The thing I have found about using green mats is that with a good result you do not have enough total pips to spend for all the pip input spots you get...  if you do 4 pips though (which is the most we get for now) and the result is amazing the stats are boosted by 50% (3 pip amazing is 25%, 2 pip Amazing is 16%, one pip amazing is 12%)  If you do all 4 pips at once you gain no more than the best you can get doing 4 pips one at a time...   you are just putting your eggs in one basket.   The odds say do them one at a time but getting 4 amazing in a row is far less likely.

6FUI4Mk.jpg

                                                        Sugoi - Senpai

Link to comment
Share on other sites

Actually in creation or rather "make" the final item quality is determined by the quality of the raw material...  if you have 3 input spots and use green in 2 and white in one you have ??66% chance of a green end result giving you 4 pip spots for sheen and durability (or even more stat properties of the item).  i haven't checked direct correlation but 3 greens nearly always gets a green product and 3 whites have a very small chance (but it does happen) of producing a green result.  The thing I have found about using green mats is that with a good result you do not have enough total pips to spend for all the pip input spots you get...  if you do 4 pips though (which is the most we get for now) and the result is amazing the stats are boosted by 50% (3 pip amazing is 25%, 2 pip Amazing is 16%, one pip amazing is 12%)  If you do all 4 pips at once you gain no more than the best you can get doing 4 pips one at a time...   you are just putting your eggs in one basket.   The odds say do them one at a time but getting 4 amazing in a row is far less likely.

 

I wouldn't say that.

 

Assuming a 5% "Amazing Success" rate, you will get 4 in a row Amazing Success results 1/160000 times.  That is a 00.000625% chance, and it will take longer to build. Even a 10% Amazing success is only a 1/10000 chance of pulling it off 4 times in a row.

 

When conserving resources and preventing getting nothing, yea maybe one at at time will produce a positive result more reliably, but you will almost never get the maximum possible result as taking a chance on maximum risk.

 

BP's are all going to be based off of maximum risk.  Even at 1/100 chance of Amazing Success on 100% risk vs 1/625 with a 20% odds 4 times in a row with 12.5% risk would be better odds.

Edited by KrakkenSmacken
Link to comment
Share on other sites

In particular, does anyone know whether the resulting stats is a linear combination of the raw inputs or whether something more complicated is being used?

 

As far as i know, the stats (like critical hit chance) are simply the result of your basic object quality/basic crafting result (like green arrows) multiplied by your experimentation success (like +25%).

 

The object quality depends on the quality of the ingredients you use and a specific chance to upgrade or downgrade your basic crafting result. The basic result is what you get after you click combine and before you start experimenting. You see the icon of the object and it's quality. Blair said that's the reason why there is an abort button at this stage of crafting, so you can just let go and save time and try next.

 

The basic crafting result determines the number of fields you can fill up with experimentation points (number of experimentation points depend on your skill). At this stage it doesn't matter if you used 2 green and 1 black stack. If the basic result is grey, you will get the number of experimentation fields according to this qualtity level.

 

I don't know though, if the basic stat might be different, according to different quality levels of the used ingredients. I think it is more likely that they just depend on the basic quality result. Regarding this i only did a few experiments with arrows and they had always the same basic stats depending on the basic quality result. But that might have been a coincidence.

Link to comment
Share on other sites

I wouldn't say that.

 

Assuming a 5% "Amazing Success" rate, you will get 4 in a row Amazing Success results 1/160000 times.  That is a 00.000625% chance, and it will take longer to build. Even a 10% Amazing success is only a 1/10000 chance of pulling it off 4 times in a row.

 

When conserving resources and preventing getting nothing, yea maybe one at at time will produce a positive result more reliably, but you will almost never get the maximum possible result as taking a chance on maximum risk.

 

BP's are all going to be based off of maximum risk.  Even at 1/100 chance of Amazing Success on 100% risk vs 1/625 with a 20% odds 4 times in a row with 12.5% risk would be better odds.

Yes.. when you have access to BPs and a thrall station AND a significant quantity of identical high quality mats then sure experiment for max results which is 4 pips at once...   I would do that for sheen first and if it is an amazing result I may do one pip at a time for durability to get the best "average" result and accept what I get as probably as good as I expect...   you could waste a lot of mats doing 4 pips sheen followed by 4 pips durability until you get both as amazing...   this just isn't going to be possible for a crafter for quite awhile when the crafting pots go away.

6FUI4Mk.jpg

                                                        Sugoi - Senpai

Link to comment
Share on other sites

As far as i know, the stats (like critical hit chance) are simply the result of your basic object quality/basic crafting result (like green arrows) multiplied by your experimentation success (like +25%).

 

I believe there is also a bonus/malus modifier attached to the math here based on what quality your resources are.  When you multiply your experimentation increase with your base stats, the numbers don't quite add up.  I tried testing the quality level theory, but the results we inconsistent.  For now, I'm blaming rounding on these tiny stat numbers we have at the moment (+.04, etc).  I think I'll try again at a later date when we have actual whole numbers to work with. ;P

Link to comment
Share on other sites

Yes.. when you have access to BPs and a thrall station AND a significant quantity of identical high quality mats then sure experiment for max results which is 4 pips at once...   I would do that for sheen first and if it is an amazing result I may do one pip at a time for durability to get the best "average" result and accept what I get as probably as good as I expect...   you could waste a lot of mats doing 4 pips sheen followed by 4 pips durability until you get both as amazing...   this just isn't going to be possible for a crafter for quite awhile when the crafting pots go away.

 

That's probably one of the best parts about the crafting lines. RNG tuning strategy.  Take your example above, say you get that first Amazing Success, then bounce to single pips, make your BP. Then you could try again, having a certain BP, and wait for double Amazing Success. (Really need to know the actual math on that for good odds.)

 

Looking at the video 

 

Here he hits "Great Success" on 8 pips.  Probably fully pot buffed, but it shows just what can happen even with 100% risk.  Of course like everything else, I think the whole crafting RNG system is going to need quite a few passes.  All these round large numbers in the risk  and difficulty tell me they have not ran the numbers through proper odds regression testing.  

 

Right now it feels way way way to easy to hit Amazing Success, while at the same time way way way to frequent on the critical failures.

 

It should be noted, that the difference between a high pay out, and low payout slot machine is about 6%.  

 

The difference between 90%-96%  ( http://www.casinocenter.com/loosest-slots-2016/ ) is actually a HUGE difference in odds, especially for something that has a high event repeat model like BP creation is going to have.

Link to comment
Share on other sites

THESE POSTS ARE ALWAYS THE MOST INTERESTING TO READ. STOP IT.

 

Edit: please do not stop it.

Edited by goose

https://en.wikipedia.org/wiki/Software_release_life_cycle#Pre-alpha <--this is where we are. If your complaint is that the game don't not works good, come back later.

Link to comment
Share on other sites

 

Right now it feels way way way to easy to hit Amazing Success, while at the same time way way way to frequent on the critical failures.

 

 

Agreed.  Come the actual game, I am hoping that Amazing Successes will be much more rare.

 

One of the things that I am actually excited about is trying to figure out a good design matrix once the game goes live.

 

What combinations of materials do we need to test to understand how the crafting system has been parameterized (And along with this, how many samples do we need to generate reliable estimates)

WAZ6Fov.png

"The cinnabar is a lie"

Link to comment
Share on other sites

 Share

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...