Is there something wrong with some items' affix tendency?

basically in the loot tables there is a bunch of digits corresponding to the loot weights or “bias” of the affix, which is then part of a total pool
so say affix 1-2-3-4 each has 15000 as value assigned to them out of a total pool of 100.000 combined for all affix drops,
so, if affix 1 had 1500 set instead, that would then drastically change the outcome,
^something similar was discovered recently iirc

Looks like Shattered Reality and Aetherfire were missing their bias for caster weapons at high levels. Fixed in v1.2.1. Thanks.

Boneblade never had Bleed bias, but it can be added. Sure.

10 Likes

Gratifying news. Thanks a lot😸

1 Like

May add pet bias to Lunal’Valgoth’s Girdle and Death-Watcher Pendant? Lunal has pet conversion and is useful for vitality pets. Death-Watcher Pendant also fits vitality pets and aether pets.

1 Like

those are also useful to non pet builds tho, so it being neutral is probably better than pet bias

Yes. But pet prefix and suffix are much less than non-pet, so you would only get one pet after over 20 non-pet. I don’t mean to change it to 1 pet : 1 non-pet, but maybe around 1 pet : 3 non-pet or 1:4.

belts are neutral, so adding pet bias would indeed skew it to pets compared to non pets
*i’m also not sure it’s fitting to add player bias to it, since it would “upset” the neutrality status compared to other items

To add a few more to this list:

Ugdenbog Flamestrife:
Flamestrife

I understand, why there is some chaos bias on the item, but combined with the other non-elemental types, the tendency to roll the useless (for casters) “Void-Touched” or “Searing”, plus the 1-in-3 chance to drop the right dagger to begin with, this one is quite hard to pick up a decent one of.

Riftclaw Slicer:
Converts most relevant damage types to fire, but the bias is all over the elemental types. Stormcharged is actually the most frequent rare prefix, followed by Frostborn, Glacial and Thunderstruck, all of which are pretty irrelevant.

Chosen Epaulets:
Since 1.2, they roll with an increased bias to physical affixes, which displace pierce affixes by sheer number, yet the only use I have seen for them is the pierce-conversion. Physical builds seem to not use these shoulders, but I might be wrong.
It’s also very difficult to find one with a defensive rare affix like “Menhir’s Wall” etc.

Rime-Rotten Wand:
This one rolls with a massive cold-bias, yet it converts all cold damage to acid for it’s only relevant use.
(Can’t check this in modal’s DB, but I crafted ~50 of them and its bias is very obvious)

I think this is the 3rd bug now that’s been discovered because of my sheet? I wonder if you all could find a use for Row Zero in your development process :slightly_smiling_face:

This is standard Caster dagger loot roll behavior.

It has pierce and physical bias, as intended.

This was a mistake, looks like a note in the loot table was left saying “cold” which is inaccurate. Will fix in 1.2.1.

Thank @modal for his sheet.
I went through all weapons, off-hands and shields and checked their rare Affixes to see whether there were other mistakes left out.
And here are the results.
GD Affix statistics 1.2.0
I highlighted what were strange and what might be mistakes or errors in the last column.

Following are some excerpts:

  1. Wrong bias:
  • Wraithbound (weapon) is regarded as Vitality and Aether Prefix.
  • the Venomclaw (weapon) is regarded as Aether Suffix.
  • Riftclaw Slicer is a Fire Weapon, and Korvaak’s Storm-Blade is a pure Lightning Weapon. But their bias is Elemental (mix of Fire, Cold and Lightning).
  1. Missing bias (drop rate < 2%):
  • Physical Weapon: Earthbound, Barbaric
  • Fire Weapon: Arcane Balance
  • Cold Weapon: Rimefrost, Arcane Balance
  • Lightning Weapon: Skyfallen (non-Caster), Celestial Wrath (Caster)
  • Vitality Weapon: Essence Drinker’s (Caster), Insanity (Caster)
  • Aether Weapon: The Abomination, Celestial Wrath, Shattered Reality (all Caster)
  • Aether and Chaos Off-hand: Shattered Reality
  • Mixed Elemental Affixes (like Runecarved and Scorched Runes) are only common on Caster Weapons and rare on non-Caster Weapons. I think Enchanter’s, Runecaster’s, Incanter’s, Runecarved and the Sage are designed for only Caster Weapons, as they have only Cast Speed, no Attack Speed, or provide Spirit. The Elements and Scorched Runes provides both Attack Speed and Cast Speed, making the two be more like for both Caster and non-Caster, but they are still rare on non-Caster Weapons.
  1. Bleed mixed items:
    Like Boneblade, there are other items that have Bleed Modifiers but do not have Bleed bias.
    Boneslicer, Scorpius Bludgeon, and Troll Bonecrusher.

  2. Unfair weights:
    Some afffixes are so common while others are too rare .I’m not sure whether these were intentional or mistakes.

I hope it can be helpful. @Zantai

true, I’ve just noticed that yesterday.

thank you for your hard work to find all of this. nice work, and hope they will improve the affixes🐱

This is great but you should try doing the analysis in Row Zero; you can have all the data right there in your sheet! Easier to show your work. :slight_smile:

Also, it looks like you haven’t enabled link sharing on the google doc you shared. I couldn’t open it.

Sorry for that. I have reset the sharing settings and it’s OK now.

Row Zero is good, but I’m not yet proficient in using it.

Or… just drop the bias altogether and let RNG handle this. Such manual intervention will always leave something to be desired.

Hmm by the way, would you take this into consideration?:face_in_clouds: Add Ascended sets (Casque and Shoulderplates) to some Bosses’ loots?

My friends and I have just uploaded and farmed Blaze Herald and Heart Of T.M… It seems that you increase the bias of “Of Sanctification”, instead of “Of Shattered Reality”? Is that a mistake?

By the way, “Of Insanity” still has too high bias. This is not a very big issue, though.

i’m kinda curious if magick pet prefixes has a higher prevalence than magick player affixes, on non-pet caster/accessory items specifically, on 1.2.1 :thinking:
obv potential confirmation bias being possible, seems like ex stuff like subjugator appears more often than stalwart/aggressive/mystic on ex player dmg caster offhands

I’m so suprised that many biases have been changed in 1.2.1.

For prefixes, unusual prefixes, like Earthbound, Barbaric and some mixed elemental prefixes, appear much common. And in each damage type,different prefixes have the similar probability. I think the prefix bias is very nice now.
The only one possible mistake is that Vitality Caster Weapon misses the bias of Essence Drinker (drop rate < 1%).

However, for suffixes, things are not as well.
Piercing Darkness, Conflagration, Ferocity, Insanity, Voracity and the Abyss still have a significant higher probability than others in the same damage type.
For example, LIghtning Weapons (like Storm Halberd) have 10% chance to meet Ferocity, but only 3.5% chance to get other Lightning suffixes and 2.6% chance for Elemental suffixes.
Aether Weapons (like Rut Blaster), another example, have 15% chance to get Insanity while only 4% for other Aether suffixes. The gap is huge

2024-06-20_131239

2024-06-20_132219

Another problem for suffix is that the Sage is abnormally common for Fire, Cold and Lightning Weapons, and Sanctification is so common on all Caster Weapon of all damage type.

(The Sage has a much higher probability than all Cold suffixes on Cold Weapons. It is also more common than most Fire and LIghning suffixes except Conflagration and Ferocity.)
2024-06-20_132947

(Sanctification has higher probability than Poison suffixes on Poison Caster Weapons. And as similar common for all damage type.)
2024-06-20_133234

I hope suffix bias can be adjusted better in future patches or builds. @Zantai

It is the nature of how affix tables are laid out. Suffixes on weapons, for example, are split among 2 tables, and it was largely chance how those probabilities ended up. For Aether, one of the suffix tables only has 1 rare suffix in it that is weighed towards Aether damage, but the other has 5. This largely came about from the affix tables being much smaller back in the day, where all of the affixes would not have fit into one.

Reworking the tables to smooth out the probabilities is not really worth it at this point.