Massive lootdrop_entries for a single lootdrop_id

[gej302]

Hybrid DB/Dev issue...

from observation the lootdrops are handled sequentially, if you have several items on the list. I noticed that some classes (bards/clerics) were getting more drops because I had inserted their gear close to the top of the list.

This does limit large lists from working effectively. Even at 1% chance of dropping, with 50 items on the list the 50th item has a true chance = 1%*.99^49
(.6%) because if anything before that one drops, that's it for that list.

Anyone else seen this happen?

[Drawde]

That would explain why skeletons, etc. only ever seem to have 1HS weapons, hardly ever 2HS, blunt etc. The 1HS weapons are at the start of the "rusty weapon" loot table in my DB (since they occur first in the item DB) so are much more likely to get picked by the random selection code.
This problem didn't seem to occur in earlier EQEmu versions, I recall.

[gej302]

I can't speak to earlier versions, but from the table and drop systems, the table processes all chances, but the drop stops when the first item drops... the fix to this can be done in the db as a temporary measure, but it's somewhat time consuming. If you have x items and want equal probability for each, the probability for item n = 1/(x-n+1) so for item 1 of a 10 item list, you have 1/(10-1+1)=1/10=10% for item 2 you get 1/9, ...item 9 = 1/2, item 10 = 1/1. Note that it seems counter intuitive since it feels like you have a combined 175% probability, but each probability is a separate entity. if you get to item 10, then you missed the chances on the other 9, so by definition you are in the 10% case. Ok I've hit the limit of my math proof skills there, I know it's right and my Math Professor wife agrees it is, so that's enough for me. Note that the granularity for these tables is integers, so we may not be able to level out tables completely, but we can get closer. I can't come up with a way to do this in the code that isn't resource intensive, but I'm working on it... it might be possible to do a single probability roll, then check all that match that roll and do a straight up random on the number of matches, but for large tables, that could result in a good chunk of extra memory consumption during zone bootups/repops.

[Drawde]

The way it seemed to work originally was that the drop chance was PER ITEM.
So if you had a loot table with 20 different weapons each with a 10% drop chance, each
one had a 10% chance of dropping, and you could potentially get more than one weapon drop
at once.
This is the system that I based the loot drop percentages in my DB around. For example, a dragon
might have a 30% chance of dropping any one of the items in their inventory (higher chance or 100% chance
for quest items etc.) and with 12 or so items in their list this means they drop 4-5 randomly-chosen items,
more or less like EQLive I think.
Currently, though, it seems that only one item per table can ever drop (unless the loot group
has a multiplier), and which item drops is determined by its chance (and, unintentionally, by how high up on the list it is).
When was this changed, and why? It's not that there's anything wrong with the new system (apart from the problem with
long lists of items), it's just that the values in my DB (and I suspect other people's too) are based around the old system,
and it'll take quite a bit of work to change everything (adjust drop chances and use multipliers instead to give multiple item drops)