The Perfect Grail

[EasyG]

The Perfect Grail

Hi,

My name is EasyGijs, I played diablo at ladder1.11 europe from aug 2005 till jan 2006.

A few weeks ago i was looking at SPF and found out about ATMA, this inspired me to install diablo again and play SP. I was especcially interested in grailing and finding perfect items. So I've played some.
My sorc is lvl 93 now and I miss about 150 items for my grail, my best find is some tc87's and an griz honor with 332 defense.

Last week I decided to make an excel sheet which calculates how perfect your grail is. It works as following: Fill in the stats of your best items of each item and you'll see the results and graphs. For the allways perfect items, you have to fill in a 1 in the column left from the stats column.

I'm at 50.03 perfection now, self found.

Empty:
www.student.tue.nl/n/g.r.snieders/diablo/perfect grail empty.xls

Example:
www.student.tue.nl/n/g.r.snieders/diablo/perfect grail EasyGijs.xls

I hope you will love this sheet like I do.

PS it will take you like 2 hours to fill it in

 

[Phosheez]

looks interesting. will give it a try and see how close I am
Edit: unfortunately I've already found an error. In edition to ED being variable, % cold resist is variable on aldur's stony gaze. Not sure if there are more. You may want to check areat summit to make sure all of your stats are correct.

 

[Llathias]

Welcome to the forums! Check the stickies (I see you've already found out the almighty ATMA), get a free drink from the EMB, beware the shinkickers (*hands EasyGijs a pair of shinguards*) and tell us more about yourself, ie. your favorite character, where you're from, your hobbies etc. Hope you'll enjoy your stay.

That spreadsheet, by the way, looks really wonderful! At least, for those with the obsession for perfection, who are also crazy enough to attempt collecting the grail. Very nice tool, worthy of being linked from the first post in the grail thread! :thumbsup:

 

[jjscud]

Interesting, though it seems a bit odd to score a perfectly bad item the same as not having found the item.

Of course, our standard grail counts a Carthans Seal the same as a Tyreals Might so that isn't really anything new.

Anyways, very interesting and tempting to fill out.

 

[purplelocust]

Interesting, though it seems a bit odd to score a perfectly bad item the same as not having found the item.

Of course, our standard grail counts a Carthans Seal the same as a Tyreals Might so that isn't really anything new.

Anyways, very interesting and tempting to fill out.

Nicely done! Thanks for sorting that out!

How many of the perfect grail items are perfect simply because there is no variation in any of the parameters? I counted 67 in Quickdeath's A-R table in this thread and 33 in the S-Z table, giving 100 or so (my count was quick) out of 504, so that is about 20% of them that are automatically perfect, if you have them at all. Hopefully you were able to get Quickdeath's spreadsheet of variable stat items as a starting point, to save yourself some trouble and maybe cut down on errors.

Did you see the posting in that thread from ricrestoni which proposed a way of counting items quality not as "perfect or not" but instead gave a "perfectness percentage" that seems to be the way you chose to deal with it? It is odd that a perfectly bad item counts the same as not having it at all; I'm not sure a good way of addressing that. Maybe instead of doing (x-min)/(max-min), you could do (x-min+1)/(max-min+1) which would give partial credit even for the worst possible roll, which is better than not having it at all! For example, a 190 ED Heart Carver out of a 190-240 possible ED range would score (190-190+1)/(240-190+1) = 1.9%, which isn't much, but it is better than 0!

One thing is that if the quantity in "Found" is more than one, it messes up the perfection percentage. That is, if you have found 3000 Cathan's Seals (some of us no doubt have, probably not kept them all, though) that gives a perfection percentage of 3000000% in that line and makes the overall perfection more than 500%. I don't think that's the way you intended for things to work! Perhaps just putting an =max(?,1) before multiplying will solve that problem. Or maybe that wasn't meant to be a quantity, just a 1 or 0 for yes or no.

Your spreadsheet works fine on OpenOffice and NeoOffice/J, by the way.

 

[VCrow]

Nicely done! Thanks for sorting that out!

How many of the perfect grail items are perfect simply because there is no variation in any of the parameters? I counted 67 in Quickdeath's A-R table in this thread and 33 in the S-Z table, giving 100 or so (my count was quick) out of 504, so that is about 20% of them that are automatically perfect, if you have them at all. Hopefully you were able to get Quickdeath's spreadsheet of variable stat items as a starting point, to save yourself some trouble and maybe cut down on errors.

Did you see the posting in that thread from ricrestoni which proposed a way of counting items quality not as "perfect or not" but instead gave a "perfectness percentage" that seems to be the way you chose to deal with it? It is odd that a perfectly bad item counts the same as not having it at all; I'm not sure a good way of addressing that. Maybe instead of doing (x-min)/(max-min), you could do (x-min+1)/(max-min+1) which would give partial credit even for the worst possible roll, which is better than not having it at all! For example, a 190 ED Heart Carver out of a 190-240 possible ED range would score (190-190+1)/(240-190+1) = 1.9%, which isn't much, but it is better than 0!

One thing is that if the quantity in "Found" is more than one, it messes up the perfection percentage. That is, if you have found 3000 Cathan's Seals (some of us no doubt have, probably not kept them all, though) that gives a perfection percentage of 3000000% in that line and makes the overall perfection more than 500%. I don't think that's the way you intended for things to work! Perhaps just putting an =max(?,1) before multiplying will solve that problem. Or maybe that wasn't meant to be a quantity, just a 1 or 0 for yes or no.

Your spreadsheet works fine on OpenOffice and NeoOffice/J, by the way.

It might be better to use two numbers. The first is the absolute number of grail items acquired, and the second is a percentage indicating how high the average roll of your grail variables are. If an item can roll with +100-200% enhanced damage as the only variable, for example, +100% ED would count as 0% (weighing down your overall grail percentage, but it will still count as +1 to your absolute number of grailers); and +200% ED would count as 100% in this case.

So, mathematically speaking, the average grail perfection percentage would be 50% for someone who only picked up the first copy of a particular unique and never attempted to find a better version.

This makes it more competitively interesting, so much so that I actually want to start a grail now with this idea in mind! This comes from someone who never got very interested in the grail idea before now. If you don't use two separate numbers, then you wind up having to assign arbitary percentage values to invariable items, and decide whether to give a bonus percentage for having a "worst roll" item. That is subject to the whim of the designer and it is not objective.

The only difficulty with this is that require a lot of effort to find all possible variables on uniques and find a convenient way for people to plug those in for calculation, with as little error as possible.

A program that can "read" what stats items have and is programmed to be able to calculate the perfection percentage of, say, a particular ATMA stash would be ideal. Unfortunately I can't program at all beyond the most basic HTML. More realistically would be an excel file that has a line for every single variable in every single unique and set item, and then calculates a percentage at the end. Unfortunately, that's easy to make mistakes with, and more importantly it requires far, far too much effort on the part of users to actually get many people interested in doing it (as opposed to downloading a utility, pointing the utility to your grail ATMA stash, and instantly being told the perfect percentage).

 

[Quickdeath]

Great spreadsheet!

The Perfect Grail has been an interest of mine, but I found that very few others in the forum seemed interested.

I entered my data into your spreadsheet -with a few gaps that I'll fill later - and I have 53 Perfect items and 35.5% -so, I'm substantially behind you.

A couple of glitches:

Natalya's Soul has multiple variable modifiers, but it is shown on the spreadsheet as having none.

Your range of values of Enhanced damage for Hulsoldo Evo has a problem - my item is outside your range. Ditto, I think, with your range of durability for Bverrit's Keep. Also, Aldur's Stony Gaze -as already noted.

There are I believe 101 S/U's that have only a single version -but I guess its fair to omit them from the calculation.

 

[VCrow]

Now that I think about it, ATMA actually has a really nifty feature that would extremely simplify making a "perfect grail" utility. It can dump a full readout of items in a stash into a text file. It is much more well-known and I believe easier to get a utility to "look" at a text file and determine what values are in certain blocks. For example, I took this out of my stash's ATMA full info dump readout text file:

43: Maelstrom
Yew Wand
One-Hand Damage: 2 to 8
Durability: 15 of 15
Required Level: 14
Staff Class - Slow Attack Speed
Item Version: 1.10+ Expansion
Item Level: 49
Fingerprint: 0x763236b3
+13 to Mana
Lightning Resist +40%
Adds 1-9 lightning damage
30% Faster Cast Rate
+3 to Amplify Damage (Necromancer Only)
+3 to Corpse Explosion (Necromancer Only)
+2 to Iron Maiden (Necromancer Only)
+3 to Terror (Necromancer Only)
150% Damage to Undead

Maelstrom varies in being able to have 1-3 to each of these skills. A utility program could look in the text file for "Maelstrom" or "Maelstrom" followed by a line break and "Yew Wand" (in order to prevent any rares that have a similar name to a unique from creating bugs). Notepad can do a text search like that so that should be easy. Then, it could look for where it is programmed to know where the variables are: lines 14-17 (if "Maelstrom" is line 1), on block 2 of each line. It can vary from 1 to 3, with 4-12 total +skills total, so:

+1 to each (+4 total) would be 0%,
+3 to each (+12 total) would be 100%,
and each value in between would be worth 12.5% (12-4=8, 1/8=0.125).

On this scale, my Maelstrom would be graded as 87.5% perfect. ([11-4]/8) or, in words:

[ActualVariableValue minus MinimumVariableValue] divided by VariableRange equals ItemPerfectionPercentage.

VariableRange would of course be equal to the "12-4" above, or in words once again:

MaximumVariableValue minus MinimumVariableValue equals VariableRange.

This would be a huge amount of work for a programmer just to find each possible variable and calculate the values with as few mistakes as possible, but after it is done it would really stimulate interest in creating a perfect grail.

I wish I had the skills to make that utility myself. However, I can definitely do the monkey work of scrubbing the Arreat Summit for item variables. I do data monkey work for a living anyway. I just need to know what format to give the parsed data in so that a programmer can just punch it in very easily. IE: Chance guards' variables are 20-30% enhanced defense, 25-40% magic find, and 27-28 defense (don't forget this part!!). So the range is:

[30+40+28] - [20+25+27] = 26

This is actually really easy to figure out, since all numbers are integers, we weigh them equally regardless of what exactly it modifies, and the Arreat Summit actually has (varies) after every single variable!!

So if a Chance Guards has 25% enhanced defense, 32% magic find, and 28 defense, the perfection percentage is 50% using our formula above:

{[25+32+28] - [20+25+27]} / 26 = 50%

Therefore this:

23: Chance Guards
Chain Gloves
Defense: 27
Durability: 9 of 16
Required Strength: 25
Required Level: 15
Item Version: 1.10+ Expansion
Item Level: 49
Fingerprint: 0x2c68706b
+21% Enhanced Defense
+25 to Attack Rating
+15 Defense
200% Extra Gold from Monsters
40% Better Chance of Getting Magic Items
+2 to Light Radius

Would have:

{[21+40+27] - [20+25+27]} / 26 ~= 61.54%

It's easy. Who knows C? Who's with me!?! I'll send you all of the possible variable's data in any format that you want. Within a week. I just don't know the programming language to do this. I guarantee that every single person with a grail stash would download the utility in order to see how their grail stacks up in perfection. It would even give the complete grail guys something new to do. Come on, now!

 

[EasyG]

Interesting, though it seems a bit odd to score a perfectly bad item the same as not having found the item.

Yes, I know this is odd, but a perfectly bad item is perfectly bad, so should add 0 % in my opinion.

Nicely done! Thanks for sorting that out!

How many of the perfect grail items are perfect simply because there is no variation in any of the parameters? I counted 67 in Quickdeath's A-R table in this thread and 33 in the S-Z table, giving 100 or so (my count was quick) out of 504, so that is about 20% of them that are automatically perfect, if you have them at all. Hopefully you were able to get Quickdeath's spreadsheet of variable stat items as a starting point, to save yourself some trouble and maybe cut down on errors.

Did you see the posting in that thread from ricrestoni which proposed a way of counting items quality not as "perfect or not" but instead gave a "perfectness percentage" that seems to be the way you chose to deal with it? It is odd that a perfectly bad item counts the same as not having it at all; I'm not sure a good way of addressing that. Maybe instead of doing (x-min)/(max-min), you could do (x-min+1)/(max-min+1) which would give partial credit even for the worst possible roll, which is better than not having it at all! For example, a 190 ED Heart Carver out of a 190-240 possible ED range would score (190-190+1)/(240-190+1) = 1.9%, which isn't much, but it is better than 0!

One thing is that if the quantity in "Found" is more than one, it messes up the perfection percentage. That is, if you have found 3000 Cathan's Seals (some of us no doubt have, probably not kept them all, though) that gives a perfection percentage of 3000000% in that line and makes the overall perfection more than 500%. I don't think that's the way you intended for things to work! Perhaps just putting an =max(?,1) before multiplying will solve that problem. Or maybe that wasn't meant to be a quantity, just a 1 or 0 for yes or no.

Your spreadsheet works fine on OpenOffice and NeoOffice/J, by the way.

I found quickdeaths posts and wrote them over in my excel sheet, while taking out a lot of errors, I'm sure there are still errors, so feel free to post them here, s i can fix them.

Yes I used that system, which seems good to me.

Found should just be a 1 or a 0.

It might be better to use two numbers. The first is the absolute number of grail items acquired, and the second is a percentage indicating how high the average roll of your grail variables are. If an item can roll with +100-200% enhanced damage as the only variable, for example, +100% ED would count as 0% (weighing down your overall grail percentage, but it will still count as +1 to your absolute number of grailers); and +200% ED would count as 100% in this case.

I think you want to calculate the average perfection of all the items you have, so inclusive double items and exclusive items you did not found. That would also be cool, but I wanted to calculate the perfection of your grail.

Offcourse it would be cool to implent perfection calculation in ATMA. But I can't.

Great spreadsheet!

The Perfect Grail has been an interest of mine, but I found that very few others in the forum seemed interested.

I entered my data into your spreadsheet -with a few gaps that I'll fill later - and I have 53 Perfect items and 35.5% -so, I'm substantially behind you.

A couple of glitches:

Natalya's Soul has multiple variable modifiers, but it is shown on the spreadsheet as having none.

Your range of values of Enhanced damage for Hulsoldo Evo has a problem - my item is outside your range. Ditto, I think, with your range of durability for Bverrit's Keep. Also, Aldur's Stony Gaze -as already noted.

There are I believe 101 S/U's that have only a single version -but I guess its fair to omit them from the calculation.

I fixed the errors:
Natalya's soul
Stony Gaze
Bverrit keep (i used th Dur+ instead of Dur, while Dur is more easy)
Crushflange (found out this item also has ED)

I did countblank(column C) and it gave me 100 always perfect items. I also added them to this sheet, cause you have to collect them too for a perfect grail

 

[VCrow]

I think you want to calculate the average perfection of all the items you have, so inclusive double items and exclusive items you did not found. That would also be cool, but I wanted to calculate the perfection of your grail.

I didn't even think about the fact that what I described would calculate averages for every single item, not just the best item of each type which is what I intended. Good point.

 

[Tigereye1]

Hi EasyGijs, nice thing you have going here, as you might have seen I have filled out your excel-file and added myself to the table, in the other thread. I hope you got my corrections/suggestions and they are correct (I could be wrong on some, so please check them).

About checking the informations, what are you using? I am mostly using Arreat Summit and D2Data.net, as they, in my experience, are correct. I am asking this because I see that you have updated Natalya's Soul again and added a Stamina 100-120. According to Arreat Summit and D2Data.net this is not correct, assuming v.1.11 or v.1.10 which is the same. If you need some help in checking the variables, I can try to help. Just let me know.

 

[Sir Lister of Smeg]

I've been looking at it, and am happy to confirm that for once an (insert expletive of choice here) excel sheet works in open office.
I did wait with adding myself to the table in the other thread though for a couple of reasons.
If an item has multiple variable items, what is the best one? For an example take the Nagelring, I yesterday found my best one yet, the first on with 30% magic find, still, although it's the best one in my eyes, the attack rating was quite low, so it would surely not be the most perfect one. Now I have somewhere between 15 and 30 Nagelrings, how do I quickly decide on the most perfect one in this case? And the Nagelring is a relatively easy example, iirc, the two mods I mentioned above are the only two variable mods, there are those with four variable mods, or even more maybe.
Another point is, I'll take Civerb's Ward as an example here, since we've got to find a perfect one in the Scavengerhunt III, how important is the variation relatively? It has three variable mods, defense, hidden set bonus to mana and hidden set bonus to poison resist. Both the latter have two possible values, still I'd count a 22 to mana 25 to poison resist as more perfect than a 21 to mana 26 to poison resist (assuming both have an equal defense value), since the relative perfection of 21 out of 22 and 26 out of 26 is lower than 22 out of 22 and 25 out of 26. So in other words, I want the imperfection to be weighted relative to the value to which it applies (one could take either the minimum, or the maximum, the average or the sum of the two, it does not matter). I did not see this being taken into account, only the difference between minimum and maximum, not how that difference compares to the value, but I might be wrong.

As to the programming Vcrow mentioned, taking the readout from ATMA is indeed an easier way of doing it, but I must add that Easygijs already did the work of putting in the variable ranges into the excel sheet. One could easily export that to a text file, and then compare each item found in the ATMA readouts to those values. The only problem being of course that we'd need to know exactly which fields are in what order in the ATMA readouts for each item. Although grepping/sedding/awking without that might work. His example of checking for both the name and the item type is not enough though. In most cases, the name of the set/unique item won't appear as the name of a rare item so there are no problems. I did however find already quite a few Stone Crushers, all rare items, I don't think any of them was a legendary mallet, but I do think it could happen. I think that what is needed is to go over all the unique items that have names that can appear as rare (or crafted) items (I don't think there are set items that can have this happening), and check whether they have mods that set them apart from rare/crafted items, even then, it might be that there are a few that have only mods that can appear on rare/crafted items of the same type, so there would be no way of automatically distinguishing between the uniques on the one hand and the rare/crafted ones on the other. It might however be that that's not the case, and in that case everything could indeed easily be automated.

As to the programming (supposing there are nor problematic items as I mentioned above), it would basically come down to this:
1. Make full dumps of all characters and stashes. (Done easily automatically, for BSD, for example Mac, users, can you use ATMA using Wine? This excludes the drop calculator since that crashes under Wine (at least under linux, it might be different under BSD), just the muling part, and thus the dumping of info). Windows users can either use Flavie for this, or use cygwin for easy scripting which they will need for the rest I think, windows being unusable without it. (also with it due to windows' inherent instability but that's another matter))
2. Make a list of all those that you'd like to include (can be done before step 1)
3. Either make a script to compare all those items to a list (the list being based on Easygijs' excel sheet, why do the same work twice? Or alternatively taken straight from SetItems.txt and Uniques.txt (I might not have named the two files correctly, I cannot check it atm)) Or concatenate all the dumps to one file and then do that comparison. In either case, each item in the dump(s) is put into a new file containing only the name of the item and it's perfection value. If the original dumps were not concatenated earlier, then that is done now.
4. Sort the so obtained file. (optional step: uniq the result hereof)
5. Choose the best versions of each item (the uniqing above will fasten this step, but it might not be necessary)
6. Give results as perfection percentage and number of perfect items found.

Note: in most of the steps above, the creation of files is not actually necessary, assuming you do have enough memory in your computer (the value of enough will depend on the total size of the dumps though), instead everything can be piped to the next step.
Note: for the comparison, it might be handy to rewrite the dumps (or the concatenated dump) to have one record per line instead of one field per line. This would probably ease the grepping to throw out all the junk (this is meant to be the first part of 3 above). (such as character stats, number of items in the stash, rares with the names of uniques)

 

[Tigereye1]

I have to agree with Sir Lister of Smeg about it is necessary to find an easy method of deciding which Nagelring is the best, if you have a lot, then you have to try each of them to find the one, which gives the highest percent.
It is possible to just look at the Nagelrings and narrow it down to a couple of candidates and then try these. This is possible when the item have two variable stats, but if you have something like Natalya's Soul having four variable stats you have to try a lot (maybe all) of the Natlya's Soul's in your stashes.
This is where a program like Flavie will make this much easier, also it can be difficult to update the excel-file without looking at all the items in your stashes.

I will not comment on the programming, as I have no experience in doing something like this.

I do have a couple of comments about how the percent of perfection is calculated.
First I think all variable stats on an item must be equal in the calculation. So a Nagelring with 30% MF and 50 AR is equal to a 15% MF and 75 AR. Both give a perfection percent of 50%. I think we all can agree that it is better to find the first one (with 30% MF), but in this context it is not possible to find a better way. Let me try to explain this by listing a couple of potential problems in doing this.
If MF is more important than AR, then how much more important is MF?
Natalya's Soul have Cold resist and Lightning resist as variable stats, is LR more important compared to CR, or are they equal? I don't think we can find a way of doing this so all people agree with the calculation of the perfection percent.
Thereby I think the current system, where all variable stats have the same weight in the calculation, are the best options. (This is actually just like Tyreals' Might and Nagelring count equally in the quest for the Grail.)

Secondly I think that a perfectly good item obviously has to have a perfection percent of 100, but also that a 'perfectly bad' item has to have a perfection percent of 0.
If the calculations have to have these endpoints then the current method of calcultion is the only one (as far as I can see).
Example (Naj's Light Plate):

x = variable stat (here Defense 721-830)
y = perfection percent

The endpoint conditions:
y(721) = 0
y(830) = 100
The solution must be a straight line:
y(x) = a*x+b
y(721) = a*721+b = 0
y(830) = a*830+b = 100
The only solution is:
a = 100/109 = 0.917
b = -100/109*721 = -661.468
After a little rearranging this becomes the expression used in the excel-file:
y(x) = 100/(830-721)*x-100/(830-721)*721 = (x-721)/(830-721)*100

Thereby I think that the current method of calculating the perfection percent is optimal.

Edit: Hope it is making sense, it is not very coherent but hopefully my points are shining through :scratch:. Please disagree with my points if you find them not understandable or wrong.

 

[Sir Lister of Smeg]

I have to agree with Sir Lister of Smeg about it is necessary to find an easy method of deciding which Nagelring is the best, if you have a lot, then you have to try each of them to find the one, which gives the highest percent.
It is possible to just look at the Nagelrings and narrow it down to a couple of candidates and then try these. This is possible when the item have two variable stats, but if you have something like Natalya's Soul having four variable stats you have to try a lot (maybe all) of the Natlya's Soul's in your stashes.
This is where a program like Flavie will make this much easier, also it can be difficult to update the excel-file without looking at all the items in your stashes.

I will not comment on the programming, as I have no experience in doing something like this.

I do have a couple of comments about how the percent of perfection is calculated.
First I think all variable stats on an item must be equal in the calculation. So a Nagelring with 30% MF and 50 AR is equal to a 15% MF and 75 AR. Both give a perfection percent of 50%. I think we all can agree that it is better to find the first one (with 30% MF), but in this context it is not possible to find a better way. Let me try to explain this by listing a couple of potential problems in doing this.
If MF is more important than AR, then how much more important is MF?
Natalya's Soul have Cold resist and Lightning resist as variable stats, is LR more important compared to CR, or are they equal? I don't think we can find a way of doing this so all people agree with the calculation of the perfection percent.
Thereby I think the current system, where all variable stats have the same weight in the calculation, are the best options. (This is actually just like Tyreals' Might and Nagelring count equally in the quest for the Grail.)

Secondly I think that a perfectly good item obviously has to have a perfection percent of 100, but also that a 'perfectly bad' item has to have a perfection percent of 0.
If the calculations have to have these endpoints then the current method of calcultion is the only one (as far as I can see).
Example (Naj's Light Plate):

x = variable stat (here Defense 721-830)
y = perfection percent

The endpoint conditions:
y(721) = 0
y(830) = 100
The solution must be a straight line:
y(x) = a*x+b
y(721) = a*721+b = 0
y(830) = a*830+b = 100
The only solution is:
a = 100/109 = 0.917
b = -100/109*721 = -661.468
After a little rearranging this becomes the expression used in the excel-file:
y(x) = 100/(830-721)*x-100/(830-721)*721 = (x-721)/(830-721)*100

Thereby I think that the current method of calculating the perfection percent is optimal.

Edit: Hope it is making sense, it is not very coherent but hopefully my points are shining through :scratch:. Please disagree with my points if you find them not understandable or wrong.

It does make sense, however, I feel that a modifier having a possible variation of let's say ten (I'll stay away from particular examples from the game this time) is not always equal. If the mod in question has a range of 800 to 809, then imo an item with 800 is not very imperfect even if it has the lowest possible value. On the other hand, if the mod has a range of 0 to 9 however, then I feel that the item with 0 is totally imperfect with regards to that mod. (this all assumes that the higher value is better). Also note that all imperfect items will still be noted as such.
Now suppose an item has both the above mods, an 800/9 and an 809/0 item will be equal in the method used so far, however, imo the former is far closer to perfection than the latter.

I also think that simply using (per mod on the item) value/(maximum value) and then averaging per mod (including those always perfect) on the item is wrong. I do however think that in some way the range compared to the base or maximum value should come into the equation. Note that I do not have an answer, just a question, what would be the best way to measure it?

 

[Quickdeath]

Some good points, a 10 point variation in a large number is worth less than a 10 point variation in a small number. However, AR and Defense are really uninterstuing because they are adders onto what already is a large number because of contributions from other sources .

If I we were going to weight the imptortance of modifiers, I think I would adopt a 5 point scale, where 5 was most important and 1 was least. Then do it something like this.


5: All +X to skills, # of sockets,

4: MF%, Resist All, EDamage %, All damage adders, IAS, FCR

3. LL, ML, Resist to any elemental damage, LAEK, FHR, Faster Blocking, Block %, FRW, PDR, DR, MDR

2. EDefense, +AR, MAEK, %Dam to Mana, +x to defense, base defense, Replenish Life

1. =XGold%, EDam and AR% to Demons or Undead, +X to light radius, +X to Missile Defense

 

[Sir Lister of Smeg]

Some good points, a 10 point variation in a large number is worth less than a 10 point variation in a small number. However, AR and Defense are really uninterstuing because they are adders onto what already is a large number because of contributions from other sources .

If I we were going to weight the imptortance of modifiers, I think I would adopt a 5 point scale, where 5 was most important and 1 was least. Then do it something like this.


5: All +X to skills, # of sockets,

4: MF%, Resist All, EDamage %, All damage adders, IAS, FCR

3. LL, ML, Resist to any elemental damage, LAEK, FHR, Faster Blocking, Block %, FRW, PDR, DR, MDR

2. EDefense, +AR, MAEK, %Dam to Mana, +x to defense, base defense, Replenish Life

1. =XGold%, EDam and AR% to Demons or Undead, +X to light radius, +X to Missile Defense

I would not use such a list as you mention, since there is no universal value of which mod is more important than another. It depends from character to character. An example, +skills for a character that already kills everything with one cast/hit will most of the time be less important than fcr/ias. But, for perfection of an individual item one cannot take those other sources into account, only those on the item. For a truly perfect item all those mods need to be perfect, but to measure the perfection level of an imperfect item mods with a relatively large variation should be more influential than mods with a relatively small variation. In addition, if an item has affixes that stack with one another, there order should be taken into account.

As to defense being unimportant modifiers since it's on an already large number, I just started looking at the Arreat Summit just to see whether that statement is correct. The first item I saw was Ormus' Robes. This item (dusk shroud) has a huge variation in base defense rating, and in addition, this unique item has a modifier of +10 to +20 defense. Now of course, the latter is very small compared to the base defense rating, but on the other hand, it's one of the modifiers that it gets as a unique item, and 10 difference on a maximum value of 20 is rather a lot. The next item on the list (of elite unique armours), is The Gladiators Bane, again an item that has variation in defense, this time 150% to 200% enhanced defense, again a rather big variation albeit relatively smaller than the 10 to 20 on Ormus' Robes. In this case that big variation also modifies the effectiveness of the already big variation in the base defense of the armour (wire fleece), so is very influential. On the other hand, the variation of 150% to 200% is seemingly on par with the variation on the same item in damage reduction and magic damage reduction, both have a range of 15 to 20. The only difference in this case being that the 150%-200% translates on the item to 250%-300% times the base defense (which itself is variable as well). My point here is that I do think that variable modifiers giving a percentage should be counted as 100% plus that percentage (of course relatively to the maximum of that percentage range both with that 100% added).
To this there are exceptions however. (Magical) damage reduced by a percentage and absorbtion percentages should not be treated in this way, since the base value here is not 100%, but 0%. There might be others, this is of the top of my head.
Add to this that some items have a few very similar modifiers, one absolute and another relative. In this case, I think the first one that is applied should be the one being seen as more important. I'll give an example. Say an item has 15-20 damage reduction (property A) and 15%-20% damage reduction (property B).
If A is applied before B, then A influences how well B functions and thus a 15/20% item should be seen as less perfect than a 20/15%. (although with some numbers it might be here that the former would be more effective at reducing damage).
On the other hand, if B is applied before A, due to the nature of the modifiers neither influences one another. Still, I think that in this case a 15/20% item should be seen as more perfect than a 20/15%. (although again with some numbers it might be that the latter would be more effective at reducing damage). I know that it's known what the order of these modifiers is, it's just hidden by amnesia atm.
Other items of this type are those with +mana/+energy/+mana percentage/mana regeneration (does this also have a straight and a percentage variant?), and similar items for life. In all those cases the perfection percentage calculation for imperfect items should take the order of application into account.

 

[EasyG]

Hi,

I've updated the empty excel file, I don't know how you can import your dat from the old one to this one easily, i guess some copy paste and add work.
I've updated the points tigereye mentioned (thanks for that) Except for the HBLvL at boneslayer blade. I removed this stat since it is allways 20.
I also removed the stamina at natalya's soul.
I also found some wrong stats myself which I fixed. I do not really remember which, at least IK's Will has got some stats, which makes the total of allways perfects 99.

To answer to your questions TigerEye. I started this sheet by using the data of QuickDeath. After that I fixed the errors mentioned in that thread. Then I fixed some more errors. Then I fixed errors I found while filling in my stats. Then I fixed all EDe errors at Normal Uniques. I used purediablo.com item stats, since it's ez to make them offline available and i don't have internet at home. I know they contain errors sometimes.

I hope most of the errors are removed now. If anyone finds some more, feel free to post. From now on I will clearly post every update here. So you're able to update in your own sheet, that might be more pleasant.

Then the discussion here.
I think this is a very nice discussion and I understand the point.
Off course a bonus to all skills is better as a bonus to one skill tree for example.
But to make these differences is very objective and wasn't really my goal while making this sheet. So for ormus robes, a more perfect one can have worse defense as a more perfect one. It's strange, but it is the way it is.

I have a nice example here. I had this gheeds:
35MF
160GF
14 PReduced

Then I found this one:
37MF
103GF
15PReduced

The first one was more perfect, but the second one is "better". So I had a moral question. Drop the more perfect one, or drop the better one?
I decided to drop the more perfect one and go behind on my stats in the perfect grail. I will use the extra 2 MF to find a perfect gheeds ) :beer:

 

[Tigereye1]

Hi,
Except for the HBLvL at boneslayer blade. I removed this stat since it is allways 20.

About the Boneslayer, from Arreat Summit:
Two-Hand Damage: (50-57) To (196-224) (123-140.5 Avg)
Required Level: 42
Required Strength: 115
Required Dexterity: 79
Range: 3
Durability: 50
Base Weapon Speed: [-10]
+180-220% Enhanced Damage(varies)
+ (5 Per Character Level) 5-495 To Attack Rating Against Undead (Based On Character Level)
+ (2.5 Per Character Level) 2-247 % Damage To Undead (Based On Character Level)
50% Chance To Cast Level 12-28 Holy Bolt When Struck (varies) <------------------------------ variable
Level 20 Holy Bolt (200 Charges)
20% Increased Attack Speed
35% Bonus To Attack Rating
+8 To Strength
(Only Spawns In Patch 1.10 or later)

From D2Data.net:
Two Hand Damage53-60) - (198-227)
Durability:50
Required Dexterity:79
Required Strength:115
Required Level:42
Base Weapon Speed:[-10]
+(180-220)% Enhanced Damage
+8 to Strength
20% Increased Attack Speed
35% Bonus to Attack Rating
50% Chance To Cast Level *31* Holy Bolt When Struck <----------------- What does this mean?
+ (2.5 Per Character Level) 2-247% Damage To Undead (Based On Character Level)
+ (2.5 Per Character Level) 2-247 To Attack Rating Against Undead (Based On Character Level)
Level20 Holy Bolt (200 Charges)

From purediablo.com:
+180-220% Enhanced Damage
+2.5% Damage Against Undead per Clvl
+5 to Attack Rating Against Undead per Clvl
+8 to Strength
20% Increased Attack Speed
35% Bonus to Attack Rating
Level 20 Holy Bolt (200 Charges)
50% Chance to cast Slvl (12 - 20) Holy Bolt when struck <----------------- Strange

And it is variable, my finds are HBlvl 20, 20, 20, 18 and 13, so I dont know what the range is on this one, as all my sources are saying different things.
I probally have some more corrections as I go through the list again, but I was wondering about this the first time looking through all the items. Maybe we should make a thread for people to show us Boneslayers found in game to settle this.

 

[EasyG]

Oh, my mistake, I looked at the charges at arreat summit. I found three of them and all of them had CTC HBLvl 20, but since you also found different ones it might vary. Maybe i should change it back to 12-28 since that is what arreat says.

Might it be possible that you've found your 13 and 18 version in older versions af diablo?
interesting...

 

[Tigereye1]

Oh, my mistake, I looked at the charges at arreat summit. I found three of them and all of them had CTC HBLvl 20, but since you also found different ones it might vary. Maybe i should change it back to 12-28 since that is what arreat says.

Might it be possible that you've found your 13 and 18 version in older versions af diablo?
interesting...

I am pretty sure that all are found in v.1.11b, but offcourse I could be wrong, I play selffound so there are no hacked or anything like that. According to my finds it could also be purediablo.com which is correct (12-20).