I don't think all affixes could be equally likely. Anyone want to check my work?

[NReynolds]

I haven't seen anyone give an answer about how affixes are chosen from the list of available ones. Until I did the following, I assumed they chose the affixes from all of the possible ones, then decided whether it was the most powerful version or a lower one, and then if it were a range of values, roll the final result. But I no longer think it is that simple.

Lets assume you're crafting armor with 6 affixes and there are 12 possible choice not counting stats, plus 4 stats plus 6 doublestats (str/vit, or somesuch)

The number of possible items without a doublestat affix (one affix that provides two stats) would be:
16!/(10! x 6!)

The number of possible items with one doublestat and 5 of the other 14 affixes (12 plus the two stats not in the double stat) would be:

6*14!/(9! x 5!) = 6*10*6*14!/(10! x 6!) = 1.5*16!/(10! x 6!)

so 60% of all the 6 affix items would have a doublestat if they were equally likely.

This number is so out of keeping with what I've observed that I've either screwed up my math, made a really false assumption, or the affixes aren't equally likely.

Which is it?

My ultimate goal is to set up a spreadsheet that calculates how much a 4, 5 or 6 affix ilvl62 mighty belt (I'm also interested in boots and helms) is worth on average if I were to make hundreds. If I've missed anyone elses work on theorycrafting the odds of rare item creations, I'd love to know about it.

[Dorjan]

I haven't seen anyone give an answer about how affixes are chosen from the list of available ones. Until I did the following, I assumed they chose the affixes from all of the possible ones, then decided whether it was the most powerful version or a lower one, and then if it were a range of values, roll the final result. But I no longer think it is that simple.

Lets assume you're crafting armor with 6 affixes and there are 12 possible choice not counting stats, plus 4 stats plus 6 doublestats (str/vit, or somesuch)

The number of possible items without a doublestat affix (one affix that provides two stats) would be:
16!/(10! x 6!)

The number of possible items with one doublestat and 5 of the other 14 affixes (12 plus the two stats not in the double stat) would be:

6*14!/(9! x 5!) = 6*10*6*14!/(10! x 6!) = 1.5*16!/(10! x 6!)

so 60% of all the 6 affix items would have a doublestat if they were equally likely.

This number is so out of keeping with what I've observed that I've either screwed up my math, made a really false assumption, or the affixes aren't equally likely.

Which is it?

My ultimate goal is to set up a spreadsheet that calculates how much a 4, 5 or 6 affix ilvl62 mighty belt (I'm also interested in boots and helms) is worth on average if I were to make hundreds. If I've missed anyone elses work on theorycrafting the odds of rare item creations, I'd love to know about it.

From what I understand, all affixes are treated as one. So if you're able to have 10 levels of a single affix there's a 1/10 chance of any one of them.

I believe also there are further restrictions like RANGE. So you can't have affixes that are TOO low (some sort of cap). Most of it is just theroies.

[NReynolds]

Yes, I was commenting on the likelihood of having an affix that provided both str and vit (for example) versus the likelihood of getting an affix with just str.

There have since been other tests using crafting that show that certain affixes rarely appear while others are more common, which is what I was finding here. It will be a while before we know how likely a certain type of affix will occur, unfortunately.

[djIgneo]

I'm having a hard time following your math. Could you make it a bit more explicit?

Let's assume your assumptions are correct, that there are 12 affixes, 4 stat affixes, and 6 double-stat affixes.
Let's also assume (not sure if you did, but I would expect the item logic to work this way) that the item generation logic will first decide if the affix is a stat or double-stat, then roll for the type of stat. This would make the number of effective affixes 14.

This would mean the odds of getting at least one stat on a 6-affix item would be (100% - odds of not getting any stat affix at all):
1 - 12/14 * 11/13 * 10/12 * 9/11 * 8/10 * 7/9 = 69.2%

The odds of getting both stat mods on a 6-affix item would be:
1 - 13/14 * 12/13 * 11/12 * 10/11 * 9/10 * 8/9 = 42.8%

Let me know if you follow my logic.

[SnickerSnack]

I'm having a hard time following your math. Could you make it a bit more explicit?

Let's assume your assumptions are correct, that there are 12 affixes, 4 stat affixes, and 6 double-stat affixes.
Let's also assume (not sure if you did, but I would expect the item logic to work this way) that the item generation logic will first decide if the affix is a stat or double-stat, then roll for the type of stat. This would make the number of effective affixes 14.

This would mean the odds of getting at least one stat on a 6-affix item would be (100% - odds of not getting any stat affix at all):
1 - 12/14 * 11/13 * 10/12 * 9/11 * 8/10 * 7/9 = 69.2%

The odds of getting both stat mods on a 6-affix item would be:
1 - 13/14 * 12/13 * 11/12 * 10/11 * 9/10 * 8/9 = 42.8%

Let me know if you follow my logic.

He was not assuming that affixes roll as either stat or double stat; I think he was separating them on the basis that one is clearly better than the other.

It looks like you're using a naive approach to the calculation, whereas the OP use an n-choose-r approach. I'm too tired right now to look closer, but maybe that will help you until the OP can comment. :P

[krischan]

I wouldn't use rares as a base for statistics about affix distribution. If a rare item adds to to e.g. str and vit, it might be two separate affixes or one of those which adds to both at the same time. I would use magical items because they have the prefix and/or suffix in their name. You also find much more magical items than rares and the more samples you have, the better funded will your statistics be. Besides, you could make statistics about prefix/suffix combos in the same sweep.

Affix distribution wasn't even in D2 as well, so wouldn't be a surprise if it's the same in D3.

[djIgneo]

Hm, well my curiosity was bothering me so I quickly perused the d3 database. I think the OP's numbers don't look right because the number of affixes available should be much larger than 12 excluding stats. I'm counting somewhere on the order of 65 including stats, excluding affix rank. I'd be terribly interested if anyone has factual knowledge about the way affixes are chosen because if the distribution is flat among all affixes as we are discussing in this thread, I think we would see many more rares without any stats. I propose we nail down the number of affixes we think are available to a specific ilvl62 item, such as a mighty belt. Once we've done that we can begin to talk about how probable a particular affix is.

[wumbabum]

1. Count affix groups rather than affixes. You cannot role more than one affix from one group. Stats is one group, so you cannot have two combo modifiers.

2. 12 is far below what any armor piece has available to choose from. Here are the affix groups for boots:

ArcaneResist, ColdResist, Defense, Dex, Stats, DR, Experience, FireResist, Gold, GoldPickUpRadius, HealthGlobeBonus, Indestructible, Int, LightningResist, MF, PhysicalResist, PoisonResist, REQ, Regen, ResistAll, ResistFreeze, ResistRoot, ResistStun, ResistStunRootFreeze, Str, Thorns, Vit, HitProc, and Run.

Only some of these are used in game, but I count at least 24 that I know are definitely on the AH.

3. Use blues, like krischan said above. This is the only way to make sure you know what affixes are on an item.

[Dahmer]

I wouldn't use rares as a base for statistics about affix distribution. If a rare item adds to to e.g. str and vit, it might be two separate affixes or one of those which adds to both at the same time. I would use magical items because they have the prefix and/or suffix in their name. You also find much more magical items than rares and the more samples you have, the better funded will your statistics be. Besides, you could make statistics about prefix/suffix combos in the same sweep.

Affix distribution wasn't even in D2 as well, so wouldn't be a surprise if it's the same in D3.

Using magical items is a good and simple way to go but anyone with a fair amount of mf will end up with plenty of rares. To figure out the stats you could just get the itemlink and figure out the affixes that way *unsure*

[Karth]

Using magical items is a good and simple way to go but anyone with a fair amount of mf will end up with plenty of rares. To figure out the stats you could just get the itemlink and figure out the affixes that way *unsure*

From what I understand, the item link doesn't distinguish between dual-stat affixes and single-stat affixes.