For magic items, we know from the Arreat Summit that:



So that in the long term, 1 in 4 magic items will spawn with both a prefix and a suffix, 1 in 4 only a prefix, and 1 in 2 only a suffix.

Rare items, however, can spawn with up to 3 prefixes and 3 suffixes. The chance of being a prefix or affix is 0.5, but what is the chance of the item getting an affix to begin with? I am specifically interested in the chance of an item rolling all 6 (max) affixes.

rare items have between 3 and 6 affixes, with an uniform distribution
so 25% chance to get each outcome


https://diablo2.diablowiki.net/Item_Generation_Tutorial#Affix_selection

The IGT has the wrong affix distribution for rare items in v1.10 (and subsequent versions); it is skewed towards 3 affixes being the least common and 5 affixes being the most common:

3 affixes: 1/8 (12.5%) [0P3S, 1P2S, 2P1S, 3P0S]
4 affixes: 2/8 (25.0%) [1P3S, 2P2S, 3P1S]
5 affixes: 3/8 (37.5%) [2P3S, 3P2S]
6 affixes: 2/8 (25.0%) [3P3S]

See Atair's D2Wissen (German only): http://d2wissen.d2chars.de (which is a fantastic resource for v1.09 and v1.10).
Possible affix configurations listed in [ ] above. Jewel is capped at a maximum of 4 affixes since v1.09 (as mentioned in the patch notes).

I do not recall what the distribution was before v1.10, so it might have been even (25% chance each).
Very-much-a-lot of item generation related stuff changed in that patch.

Since I can't read German and Google doesn't want to translate this page, does this person have a source?

I just noticed on the Arreat Summit this:



So they can have a minimum of 2, not 3, but the selection method is unclear. It appears to be uniform between 2 and 6 and mentioned in post #2, but again it's not clear.

As for the general method, my guess is that it first rolls for the number of affixes, and then each one is rolled 50/50 individually to be a prefix/suffix, followed by a counter that automatically makes the outcome a prefix or suffix if the opposite affix has reached its limit of 3.

This post was edited by bboynaki on Nov 25 2018 12:26pm

Its 3-6, arreat summit is wrong.
Selection method is not deeply explained. There are just the percentages listed for each amount of affixes.
And then the affixes are chosen randomly after the roll how many it will get.

the arreat summit is often incorrect
trust quorra (post #3)

+1 for don't trust Arreat Summit. The other wikis also have plenty of mistakes.
I would trust the Basin wiki, but I don't think it deals with item generation unfortunately.



just for reference, in 1.00 it was 1/3 chance for 4 affixes, 1/3 chance for 5, and 1/3 chance for 6. very simple, and maybe a little OP.
the item could still end up having less than 4 affixes, if its ilvl was low enough, but that only really happened in the earliest part of the game.

i presume this changed in 1.07 and then again in 1.09, but i haven't really looked at rare items in those patches since they were not really BIS ever.

This post was edited by Fruit on Nov 25 2018 05:05pm

In some cases the magic properties of two affixes combine into a single line display. See https://www.theamazonbasin.com/wiki/index.php?title=Minimum_Damage (below the Hurlbat of Transcendence image).
I presume that is how the '2' on the Arreat Summit website came about.

Yes, when a rare item is to be generated the number of affixes is diced first (using the distribution in #3 for v1.10 and later versions) and then the individual affixes (prefix: P, suffix: S).
Yes, while #P < 3 and #S < 3 each pick is a 50/50 toss between prefix (P) and suffix (S). Once #P = 3 or #S = 3, the remaining picks (if any) are automatically all S or all P, respectively.

To complicate matters even further, all affixes must be from different groups (the affix picker cannot select the same P/S group twice), and each affix within each group has its own weight (frequency).
Calculating chances for affix combinations on rare items is thus very complex. There is however an affix calculator for v1.09 and later versions (for instance google: ''Affixes for LoD-Items Patch 1.11'').


Thanks for chiming in with another cool old school info bit.

This post was edited by Quorra on Nov 25 2018 06:06pm

If one site is wrong then why should I believe any other site? Does anyone have any actual evidence of the correct answer? Like data from the MPQ files?

The correct answer to which question?