Experimentation theory: how it works?

Discuss any general aspect of crafting in this forum.

Moderators: Crafting Forum Moderators, Forum Moderator

User avatar
panino76
Intermediate Crafter
Posts: 184
Joined: Mon May 19, 2008 1:26 pm
Location: Italy

Re: Experimentation theory: how it works?

Post by panino76 » Tue Jun 17, 2008 4:01 pm

Wasahi wrote:I had a similar question regarding assembly/experimentation and the resource bonus, rather than satrt a new thread I figured I'd add it here.

Say an experimentation line used an average (CD/OQ) 50/50.
And you had a resource that had 972 CD and 930 OQ

When computing the average after the resource bonus, would the formula look like:

(1012 + 970)/2

Or would CD be capped before the average was taken and so the formula would look like:

(1000 + 970)/2

This is assuming that it is a strait +40 bonus for the resource quality expertise.

It sounds like any work done previously in this area may have been lost, but I figured I'd ask.

Wasahi

Edit: Disregard the above after pouring over the Begginers guide it seems the reource bonus is applied AFTER the averages are computed.
Exactly. In a recent update, in an attempt to modify the user interface of the crafting system they changed the coloured bars (the one that is red, yellow and green depending on the resources stat) so that they were reflecting the real value of a resource (see, when you have multiple lines of experimentation, your resource could be uber for one of them but crappy for the other ones, the coloured bars give you an overall idea).

The problem was that they were forced (due to some design constraints it seems) to cap the resource stat to 1000, when the bonus exceeded that cap (or the hard cap like 650 for CD on steel), before the average was taken.

Obviously, you can imagine what was the reply of the most experienced crafters that knew what was happening, so that they abandoned the idea of modifying the UI.

Artax Wavingwalker
__________________________________________
The NEW Jenner and Fooman Tool

User avatar
Savacc
Architect & Shipwright Forum Moderator
Posts: 1211
Joined: Mon Mar 26, 2007 9:45 pm
Location: Central Oregon

Re: Experimentation theory: how it works?

Post by Savacc » Wed Jun 18, 2008 12:27 am

Wasahi wrote:Edit: Disregard the above after pouring over the Begginers guide it seems the reource bonus is applied AFTER the averages are computed.
Yes, as far as we know, until someone like Panino76 proves otherwise, we believe, the expertise bonus is appled at the very end of the equasion, after everything else is taken into account.

Bulba
Novice Crafter
Posts: 21
Joined: Fri Jun 20, 2008 5:29 am
Contact:

Re: Experimentation theory: how it works?

Post by Bulba » Fri Jun 20, 2008 7:13 am

A guy called FahoAckad posted this on the SOE forum, I use it in my crafting spreadsheet to get the exact starting %, NarfBlinko also uses it in his Shipwrights Little Helper spreadsheet. I've edited it a bit for clarity:
Start % is :

Max value with expertise bonus * ( 0.000015 * max value without experise bonus+ 0.015)

If you have calculated a max value of 97% and you have the 4% expertise bonus you get:

100 * (0.000015 * 97 + 0.015) = 29.55% <-- Note that the max is 100, not 101 (97+4).

With a max value 86.4% and 4% expertise:

90.4 * (0.000015 * 86.4 + 0.015) = 25.27584%
Nobody can tell me where the 0.000015 and 0.015 come from, but I think it came from testing with crafting tools. The effectiveness rating of crafting tools increments by divisions and multiples of 0.15 according to the CD rating of your Copper, that's why I think they're related.

User avatar
panino76
Intermediate Crafter
Posts: 184
Joined: Mon May 19, 2008 1:26 pm
Location: Italy

Re: Experimentation theory: how it works?

Post by panino76 » Fri Jun 20, 2008 10:37 am

Bulba wrote:A guy called FahoAckad posted this on the SOE forum, I use it in my crafting spreadsheet to get the exact starting %, NarfBlinko also uses it in his Shipwrights Little Helper spreadsheet. I've edited it a bit for clarity:
Start % is :

Max value with expertise bonus * ( 0.000015 * max value without experise bonus+ 0.015)

If you have calculated a max value of 97% and you have the 4% expertise bonus you get:

100 * (0.000015 * 97 + 0.015) = 29.55% <-- Note that the max is 100, not 101 (97+4).

With a max value 86.4% and 4% expertise:

90.4 * (0.000015 * 86.4 + 0.015) = 25.27584%
Nobody can tell me where the 0.000015 and 0.015 come from, but I think it came from testing with crafting tools. The effectiveness rating of crafting tools increments by divisions and multiples of 0.15 according to the CD rating of your Copper, that's why I think they're related.
Thanks for posting m8, now it's just uo understand how the bonus works.

Artax Wavingwalker
__________________________________________
The NEW Jenner and Fooman Tool

velrahnkoon
Novice Crafter
Posts: 16
Joined: Sat Nov 24, 2007 2:19 am
Location: Blacksburg, VA
Contact:

Re: Experimentation theory: how it works?

Post by velrahnkoon » Fri Jun 20, 2008 2:26 pm

panino76 wrote:
Bulba wrote:A guy called FahoAckad posted this on the SOE forum, I use it in my crafting spreadsheet to get the exact starting %, NarfBlinko also uses it in his Shipwrights Little Helper spreadsheet. I've edited it a bit for clarity:
Start % is :

Max value with expertise bonus * ( 0.000015 * max value without experise bonus+ 0.015)

If you have calculated a max value of 97% and you have the 4% expertise bonus you get:

100 * (0.000015 * 97 + 0.015) = 29.55% <-- Note that the max is 100, not 101 (97+4).

With a max value 86.4% and 4% expertise:

90.4 * (0.000015 * 86.4 + 0.015) = 25.27584%
Nobody can tell me where the 0.000015 and 0.015 come from, but I think it came from testing with crafting tools. The effectiveness rating of crafting tools increments by divisions and multiples of 0.15 according to the CD rating of your Copper, that's why I think they're related.
Thanks for posting m8, now it's just uo understand how the bonus works.
I'm not quite sure you got the forumula right when you edited it.

0.000015 * 86.4 = 0.001296

0.001296 + 0.015 = 0.016296

0.016296 * 90.4 = 1.4731584, which is nowhere near 25.27584.



Edit:I downloaded the tool, and you should be using the Max value with and without expertise bonus based out of 1000, not 100.

Therefore, in the above example:

0.000015 * 864 (not 86.4) = 0.01296

0.01296 + 0.015 = 0.02796

0.02796 * 904 (not 90.4) = 25.27584

I would agree with your guess as to the reason for the 0.000015 and 0.015 being based on the fact that the crafting tools increment in steps of 0.15.

Bulba
Novice Crafter
Posts: 21
Joined: Fri Jun 20, 2008 5:29 am
Contact:

Re: Experimentation theory: how it works?

Post by Bulba » Fri Jun 20, 2008 3:54 pm

Well spotted :)

I should have proofread it first.

Whist34
Novice Crafter
Posts: 9
Joined: Sun Jun 15, 2008 1:35 am

Re: Experimentation theory: how it works?

Post by Whist34 » Wed Jul 02, 2008 4:10 pm

Wasahi wrote:Say an experimentation line used an average (CD/OQ) 50/50.
And you had a resource that had 972 CD and 930 OQ

When computing the average after the resource bonus, would the formula look like:

(1012 + 970)/2
...
Edit: Disregard the above after pouring over the Begginers guide it seems the reource bonus is applied AFTER the averages are computed.
Wasahi,

You will get the same result if you add the flat 40 bonus in before the average or after. Mathematically speaking the quality will come out to be the same using either method (and it doesn't matter what the percent weights are).

I believe as Artax (panino76) mentioned the values can go over the cap (I don't have any statistical data for this, but it seems to be generally accepted or known that it does).


Some questions to consider:

Question 1:
  • Is it a flat +40 or is it +4% based on the theoretical cap of the resource attribute?
    • - A flat +40 on a resource attribute with a cap of 300 turns out to be a +13.33% increase to the quality so this question does merit some concern.

      - The alternative is if you add a flat 4% to the percent quality of the resource. If you use a proportion where 300=1000 then it turns out 4% of 300 = 4% of 1000 or the flat +40 which is really a flat +4%.
      • An Example (no weighted percents to simplify)
        Resource A has HR of 950 with a cap of 1000 (95% quality)
        Resource B has HR of 760 with a cap of 800 (95% quality)

        Resource A Bonus = 4% of 1000 = 40
        Resource B Bonus = 4% of 800 = 32

        Resource A with Bonus applied = 950 + 40 = 990 (same as 95% + 4% = 99%)
        Resource B with Bonus applied = 760 + 32 = 792 (792 is 99% of 800 - so again same as 95% + 4% = 99%)

        * The hypothesis is that the resource quality increase is 4% of its cap which is the same as adding 4% to its percent quality.
        * This is not the same as adding +40 to the existing attribute value regardless of the cap.
        * It is the same as adding +40 to the attribute value once it has been converted to a 1000 scale based on percent quality.
        * Note: This might explain why Artax is seeing different values (+40, +32, etc) for different ranges.
Question 2:
  • Is the 4% based on the cap or the existing attribute value? Or is it just a flat +4% to the experimental effectiveness?
    • - Most guides I have read state it is a flat +40. This value agrees with the hypothesis from Question 1. This supports the idea that the bonus is based on the cap and not the existing attribute value.

      - I did do a basic comparison of percent qualities between both methods and found that the average difference between the two was 0.264. (ie. If you calculate a percent quality of 97.5% using the cap method then, using the average, the existing attribute method would result in a percent quality of 97.236%). Note this was only for my randomly generated chart to try to get an idea of how much difference there could be. The differences ranged from a min of 0.196 and a max of 0.332. (Please don't expect a more in-depth statistical analysis from me :P )

      - The method that makes the most common sense to me is the one that just adds a flat 4% to the experimental efficiency after the weighted average of the resources is calculated. Eventhough doing it before makes no difference on the calculation outcome it would seem to be more efficient code.

      - I would love to look at the data that generated the +40 flat rate and see if the existing attribute method could also fit the data. (I hate generating this data on my own - so for now I will just be philosophical about it all :D )

I love these kinds of discussions and hope we can figure this out.

Thanks

Zimoon
Forum Moderator
Posts: 4817
Joined: Mon May 14, 2007 6:55 am
Location: Stockholm, SE
Contact:

Re: Experimentation theory: how it works?

Post by Zimoon » Wed Jul 02, 2008 6:46 pm

Have missed this thread during my vacation. I agree with Savacc, do it first without Experrtize and then with. Remember that whether you get Good or Great/Amazing Success also has its impact on the result, you only want to take notes on the Great/Amazing I think (Great and Amazing has the same value with Assembly).

I read Lunariel's formula the other week, or month, where can that be??? :?

Whatever, I seldom bother too much, all I bother with is to have perfect resource that yields 29% to start with :P

/Zimoon

Zimoon
Forum Moderator
Posts: 4817
Joined: Mon May 14, 2007 6:55 am
Location: Stockholm, SE
Contact:

Re: Experimentation theory: how it works?

Post by Zimoon » Wed Jul 02, 2008 6:53 pm

Isn't it this formula you refer to, Savacc?

Code: Select all

Y = X * (0,000015 * X + 0,015)
where X is the final weighted average for the resources involved, computed as read in the Beginners Guide for Traders.

/Zimoon

User avatar
Savacc
Architect & Shipwright Forum Moderator
Posts: 1211
Joined: Mon Mar 26, 2007 9:45 pm
Location: Central Oregon

Re: Experimentation theory: how it works?

Post by Savacc » Wed Jul 02, 2008 7:25 pm

Yes Z that would be it.

Whist34, I think you may have found the clue, good work.

User avatar
omnitree
Novice Crafter
Posts: 8
Joined: Fri Jun 20, 2008 10:09 pm

Re: Experimentation theory: how it works?

Post by omnitree » Fri Jul 04, 2008 12:37 pm

Having read this thread and finding it quite interesting, you see I've never delved so far into the systems of a MMO or any game for that matter. There is a part of what Whist said that doesn't feel right. If I understand what he is saying is that +40 and +4% are used on the resources cap. He then provides equations to back up with theory. Which works as he describes, so there are no issues there. Where my question comes from is the +40 and +4%. It has always been my interpretation that SWG uses a 10 to 1 scale (points to percentage) to determine the percentage gained by any given recipient (combat skill, entertainer buff and or trader resources expertise). Basically this means that if you have +35 points then you will receive +3.5% on, in this case, Resource quality. In the end, when it comes down to the calculations the growth of your points are not used directly in the equation, so much as they are converted prior into their percentage form.

Have I finally stumbled on common knowledge or made a leap that has thus far been overlooked? And if this is common knowledge, just ignore this post and keep watching from the corner.

Zimoon
Forum Moderator
Posts: 4817
Joined: Mon May 14, 2007 6:55 am
Location: Stockholm, SE
Contact:

Re: Experimentation theory: how it works?

Post by Zimoon » Fri Jul 04, 2008 2:04 pm

omnitree wrote:Having read this thread and finding it quite interesting, you see I've never delved so far into the systems of a MMO or any game for that matter. There is a part of what Whist said that doesn't feel right. If I understand what he is saying is that +40 and +4% are used on the resources cap. He then provides equations to back up with theory. Which works as he describes, so there are no issues there. Where my question comes from is the +40 and +4%. It has always been my interpretation that SWG uses a 10 to 1 scale (points to percentage) to determine the percentage gained by any given recipient (combat skill, entertainer buff and or trader resources expertise). Basically this means that if you have +35 points then you will receive +3.5% on, in this case, Resource quality. In the end, when it comes down to the calculations the growth of your points are not used directly in the equation, so much as they are converted prior into their percentage form.

Have I finally stumbled on common knowledge or made a leap that has thus far been overlooked? And if this is common knowledge, just ignore this post and keep watching from the corner.
Regarding crafting and bonuses some schematics provide quite good visibility to the outcome, with lots of decimals.

Our assumption was that the "4%" read at the expertize should be used as it historically was used, as X*1.04=final_outcome, where X is the weighed average with caps considered according to the rules about caps. That should have made 962 the golden limit, as 1000/1.04=961.54.

However, I think it was NarfBlinko that found out and reported at the official Structure forum that the Devs rather just added 40 to X, making 960 the golden value we now aim at. Yes, 40 is 4% of 1000, but 40 is certainly not 4% of 960 (40/960=4.17%) and the lower the X the higher the % when 40 is used*. It was easily proved that 40 is indeed what is added to X rather than 1.04 is multiplied with X.

Does that make sense to you?

/Zimoon

EDITED: * Not that it matters that the bonus becomes the higher the lower the X is, it will still not help the crafter getting closer to 1000, it is just silly that the expertize text reads 4% when it is not. But, yes, there are many other situations where the wording involving % is far from clean from a mathematical point of view.

Onyx
Chef Forum Moderator
Posts: 199
Joined: Sun Mar 25, 2007 1:33 am
Location: USA
Contact:

Re: Experimentation theory: how it works?

Post by Onyx » Thu Jul 17, 2008 1:46 am

Hmmm, no clue how I missed this thread :(

Back in March, it was confirmed and verified to me by a dev that yes, the "4%" resource quality bonus merely adds +10 per point for a max of +40.

After receiving confirmation, I posted two threads on the main SOE forums, one in the senate forums and the other in domestics. In the domestics one I made several suggestions on clearing up the confusion. So far, we haven't seen any changes but I think most people agreed that we should leave it how it is and fix the inaccurate text in the expertise window.

Thread is here:
http://forums.station.sony.com/swg/post ... _id=619269

I'll paste the text of it here for convenience:

Several weeks ago it was noticed that the resource quality bonus seems to be applying a static number rather than a percent bonus. I did some testing and confirmed this was indeed the case. After having the devs take a look at this, it is still unclear as to whether it's always functioned this was or whether this is a change that cropped up somewhere along the way.

While this is certainly not a game-breaking issue, it is something that isn't working properly.

W'eve always thought the 4% resource quality bonus was calculated by adding 4% of the final weighted average for the resources. That meant you would determine the weighted average of all the resources/stats for an experimentation line as follows:

Nutrition, based 66% on PE and 33% on OQ:

960 PE / 960 OQ would yield a weighted average of 960

Take that and add 4%, the end result:

960 * 1.04 = 998.60

998.60 / 1000 = 99.86%

So, on that particular item, the experimentation could get you to 99%. In order to hit 100%, the combined weighted average had to be 962 or higher (technically 961.6 but it was always rounded down).

However, now we see that instead of adding 4%, it is simply adding 40 (which would be 4% of 1000). Basically treating all resource averages as though they were 1000 when calculating the 4%.

To test this, I first crafted an item where the weighted average was exactly 960. Instead of getting 99% like I should have, I achieved 100%.

960 + 40 = 1000

1000 / 1000 = 100%

To make absolutely certain that it is indeed adding 40, I crafted an item where the weighted average was 199.8.

It should have been:

199.8 * 1.04 = 207.79

207.79 / 1000 = 20.779% rounded down to 20%

Instead getting an end result of 20% like I should, I achieved 24%. So not only is it simply adding 40, it's also rounding up.

199.8 + 40 = 239.8

239.8 / 1000 = 23.98% rounded down to 23%

Instead we get:

239.8 / 1000 = 23.98% rounded up to 24%

Yes, this makes it slightly easier to cap things. Yes, it is an error on the side of increasing rather than decreasing what the actual result should be. But it is also not the way it is supposed to work. At the least, it's adding almost 1% more than it should. At the worst it's adding 5% more. Should it be fixed/adjusted?



There are four possibilities for "fixes" here--and I'd like to hear recommendations from everyone on what we ask the devs to do:

1) Apply the % increase to each of the resource attribute on a 1000 scale. This was what DevH did in the 10/3/2007 "fix" that later was reverted before it went live. This would be done before the weighted average is calculated and the resource stats would cap at 1000 and we'd lose the ability to "go over" 1000 for purposes of compensating a lesser resource for a higher one. Basically, it would be adding +40 to each stat but not exceeding 1000.

2) Apply the % increase to each of the resource attribute based on the value of that resource attibute (i.e. 4% increase to OQ 956 if the resource OQ value is 956). Again, this would be done before the weighted average is calculated and again, the resource stats would cap at 1000. This would be even more drastic than #1, and again we would still lose the ability to compensate for lesser resources by using better ones.

3) Apply the % increase to the final resource experimental average (i.e. 4% increase to 956 if the final resource value is 956). This is the way we've thought its worked.

4) Leave the bonus calculations alone and change the wording of the expertise boxes to read "Increases resource quality by adding +10 per point spent".
Onyx, Chef Forum Moderator

Whitley's Fine Food & Drinks
Open for business since June 2004 -- Twilight's End, Naboo (-3036, 5300), Kettemoor

Official Domestic Goods Senator

Locked

Who is online

Users browsing this forum: No registered users and 2 guests