What are the mechanics of affix generation onrares

[Schuey]

Ok I read Arreat Summit on choosing affixes for magical items and they dont have as detailed of information for the affix choice on rare items. I was wondering just how the choices are made for rares.

I thought maybe three magical choices were made (i.e. three choices that could be sufix, prefix or both). This would make sense, and would make it easy to limit to only 3 prefix/3 suffix max. Arreat Summit shows that in magical choices the percentages of affix choice are:

Both Prefix and Suffix: 25%
Only a Prefix: 25%
Only a Suffix: 50%

However, if three magical choices are made then how could a rare ever end up with only 2 affixes as is stated by Arreats Summit (i.e. 2-6 affixes on rare).

Using the percentages above and assuming that the rare routine uses 3 calls to the magic affix choice routine, it would seem that 3 affix rares would be the norm.

If this is how it works, what happens when an affix (group) has already been chosen, is the duplicate affix dropped and no affix for that choice added? Does the program call for another try at the same affix (prefix/sufffix). Does the program redo that entire call to the magic affix routine? If the affix is dropped is it not possible to have a one affix rare (theoretically)?

Well, all of this is pure speculation. Any of you code readers know what the real scoop on all of this is?

Dave

[adeyke]

The percentages you quoted are for magic items. Not rares.

With rares and crafts, it's:
Prefix (if there aren't already 3 prefixes): 50%
Prefix (if there are already 3 prefixes): 0%
Suffix (if there aren't already 3 suffixes): 50%
Suffix (if there are already 3 suffixes): 0%

This roll is repeated for the number of affixes it gets. With rares, you get 3-6 affixes, so it first determines how many you'll have and then decides if they'll be prefixes or suffixes.

Once it's decided if an affix will be a prefix/suffix, it'll look through the applicable prefixes/suffixes and pick on by weighted random. I'd assume that, if an affix group has already spawned on an item, that it'll just remove those from the list of applicable affixes.

This is also largely speculation on my part, though. I could be wrong.

[Schuey]

Any information on the % of 3, 4, 5 or 6 affixes?

If it choses, for example 4 affixes, then it could be anywhere from 1 prefix and 3 suffix to 3 prefix and 1 suffix I imagine, and given 50/50 nature you describe it is completely random.

If the choice is 5 affixes, then it will definately have five? Or could it have less if it chose 4 prefixes (i.e. dumping the 4 prefix choice?)

What did you mean by weighted random on the affix choice? Some are much less likely to occur?

Dave

[Thrugg]

As far as I know the number of affixes is (uniformly) random from 3-6. This is distinct from crafted items which have a minimum number of random affixes depending on ilvl.

Each affix has a weight, eg rare-enabled Cruel is 1, Ferocious and Merciless are 6, Savage and Massive are 7, etc. This means that on rare items capable of spawning Cruel, you will see only 1 Cruel for every 6 Ferocious or Merciless, or 7 Savage mods. The magic-only Cruel is weighted 5, so you see quite a few more Cruels on magic items.

When the game chooses eg a prefix, it adds together the weights of all eligible prefixes for that item, and the chance of any one is (its weight)/(total weight).

If you think you see a lot more Bronze charms than Shimmering ones, it isn't just your imagination :rolleyes:

[adeyke]

There's also the matter of the magic_lvl. This is a property of certain base items (specifically circlets, staves, orbs and non-elite wands). The magic_lvl modifies the alvl calculation, making the alvl much higher than it would be without the magic_lvl.

It also changes the weighting.

While, as you said, the probability is normally (weight)/(total weight), items with a magic_lvl instead have the probability (weight * (required alvl))/(total (weight * (required alvl))). That is, affixes with a high alvl requirement are weighted to become more common.