'Anything Goes Lounge' posts worth preserving

art_vandelay

Well-known member
Mar 16, 2020
339
582
93
This is a place for 'Anything Goes Lounge' posts you want to preserve because you found them useful or they contain valuable info or maybe just because you put some effort in them and don't want them to vanish with the old site.

I'll make a start with some of my posts as well as a few fruit posts, feel free to add to them!

To keep things easy I will maintain an index in this post:


Index

- fruit explains how champions, hardcoded bosses and evil urns count towards the boss limit of an area. (link)
- fruit explains how gold drops work. (link)
- art_vandelay compares the rune drop chances of different monsters. (link)
- art_vandelay proves that the chance to pass the unique roll in quality selection never bottoms out. Contains a table of MF values at which set and rare chances bottom out. (link)
 
Last edited:

art_vandelay

Well-known member
Mar 16, 2020
339
582
93
@art_vandelay @T72on1

A bit of a late reply, but.. I got some bad news. :mad: In that post I had said that hardcoded spawns don't count towards the spawned_packs counter, but they do. Furthermore, each Champion individually increments this counter! Hardcoded packs will spawn regardless of the counter however, so there's at least that.

So if you know the location(s) of hardcoded champions/uniques (including superuniques!), then you will want to try and reach those locations last.

Each area has its own counters, so Pit 1 and Pit 2 are handled separately. I believe Pit 2 always has one hardcoded unique pack, but I'm not too familiar with the map files and can't say for sure. :E

If an Evil Urn spawns a pack, then that will also increment the spawned_packs counter, so if you're out to min-max your monster packs, pop those urns at the end. Also, these urns will only spawn a pack if the urn dropped at least one item, so you will increase the chance of an urn spawning a pack by increasing the player count (these urns run 1 cycle of the appropriate Chest TC). Silly Blizzard mechanics. By the way, it is not a bug that those two hardcoded urns spawn spiders; all Evil Urns spawn these spiders. The reason that those two urns never spawn a pack is because the game does not allow monsters to be spawned that close to those entrances. (Hardcoded spawns ignore this rule, which is why that Thresh Socket guy for example is allowed to guard that entrance.)


Sorry for that mistake! I had some stuff mixed up. :(
 

art_vandelay

Well-known member
Mar 16, 2020
339
582
93
Does anybody know what the ranges on gold drops are, i.e. when the game drops the item 'gld' what is the min and max value?

Can't seem to find anything on this :confused:
The standard calculation is simply mlvl + random(mlvl*5), which is then multiplied by (100+gold_find) and divided by 100. And just like with Magic Find, if your mercenary made the kill, the Gold Find from both is added. I'm sure this is not the complete story though -- I think certain monsters/treasureclasses get a bonus. For example the "Baal (H)" TC has a "gld,mul=2048", which sounds like he can drop a whole lot more. I don't know where or how this is used, though. Kinda looks like all TCs used by champions, uniques & bosses have a multiplier like that.
@Fruit Thanks for the base formula, that's exactly what I was looking for!

As for the mul thing I guess it's like the quality selection in TCex.txt, the 2048 is something like bits and you have to divide by a suitable power of 2.

I can only make an educated guess at that though:

Council has mul=2048 and in the Piles of Money tournament mentioned by Pb_pal people roughly have 2400% GF on Hork and have reported stuff like 110000 gold in one pile. Dividing that by 510=85*5+85 (council has mlvl 85) and then by 25 for the GF we get 8,6 and incidentally 2048/256=8.

I think some other stuff is measured in 1/256's, like poison damage, so it seems reasonable.

That would mean that highest possible Gold Pile from council is 118728 at 2810% GF if you make the kill (highest possible GF, classic items with 120% GF taken into account)
Yep, looks like you're right. So first is the mlvl+random(mlvl*5), then multiplied by this "mul", then divided by 256, and then the gold find part. This "mul" only comes in 1280, 1536 or 2048. (I think that the random number includes 0 and excludes the last number, btw. So the lowest should be mlvl, and the maximum should be mlvl*6-1.)

That should be all, unless there's something special hardcoded somewhere. The basic calculation also seems to be used for those gold drops that spawn in areas on the ground, and they'll use the area level. :p
 

art_vandelay

Well-known member
Mar 16, 2020
339
582
93
I've been going through the 'TreasureClassEx.txt' a lot lately and one of the things I was interested in was wether Wraiths are the only 'regular' monsters (i.e. not champions/uniques/SUs/bosses) with better chances at runes. So I took the TCs for all the regular monsters and calculated the chances to drop a rune for all of them on all player settings.

They all have one pick except for the council which have three. Here I took the chance to get at least one rune but it doesn't matter much since we already know how good they are for runes.

The only two contenders which we don't know yet are 'Swarm' and 'Quill' which are the TCs for Swarms and Quill Rats respectively. There are also TCs for Flying Scimitars but you can't kill those en masse so I omitted them. (They have the exact same chance to drop from their rune TC as a Cow if anyone cares)

In the following table the columns indicate the TC and the player bonus (1 for p1/2, 2 for p3/4,...).
runecomp.png
This gives the ratio of the rune drop chance for a given TC on a given psetting versus the rune drop chance for the H2H/Cast/Miss TC on the same psetting. For example on players 1 the Quill TC has 1,45 times as high a chance to drop a rune than H2H/Cast/Miss.

Pretty much as expected, we see the familiar 3,59 factor for p8 ghosts and how good Cows/Council are for runes. As for Swarms and Quillrats:
  • Swarms suck. On top of their worse odds for runes they are also capped at Ber no matter where they spawn.
  • Quillrats actually are better at dropping runes than Cows. However except for Minion Quillrats in the 'Ancient's Way' and the 'Halls of Pain' they are capped at Vex. The two exceptions are capped at Ber though.
Apart from that the only thing I learned here is that H2H (melee monsters), Cast (Caster type monsters) and Miss (ranged monsters) have the same chances at runes.

(All of the above results are true for 'runes' replaced by 'good' where the good TC has gems, jewelry and runes. However the only monster type in the above table with a bonus to dropping uniques/sets/rares are the councilmembers, so they're the only regular monsters with high chances for facets or rare rings)
 

art_vandelay

Well-known member
Mar 16, 2020
339
582
93
I'm currently in the process of doing a writeup on Classic 1.00 MF with the help of fruit and this means I'm calcing all kinds of MF related stuff, such as breakpoints where 'more MF' actually hurts.

In the spirit of this I wanted to share something I learned about modern patch MF which also has to do with the whole 'more MF doesn't always help' thing. In one of his videos @drmalawi conjectured that while Set and Rare chances bottom out at some value of MF (depending on mlvl and qlvl and other stuff) the same is probably not true for Uniques. Well, I did the math and I can confirm that this indeed NEVER happens for unique quality.

To be precise: I show that if 'bottoming out' occurs for Uniques it is not due to the comparison of the reciprocal chance for the unique roll with a minimum value (6400 for unique). Also this result is only applicable to monsters that use the usual quality selection, i.e. it doesn't apply to racks, super chests, sparkly chests, dclone and ubers (and any other oddball you might come up with).

Maybe there is some combination of mlvl, qlvl and an MF treshhold (no, I'm not talking about 250*249 MF) which makes the unique reciprocal chance bottom out before the 'comparison step' is ever reached but this is not what I'm talking about.

The quality selection calculates the reciprocal chance for a quality to be selected in a number of steps. The step that is responsible for the Set and Rare chances sometimes bottoming out compares the current chance with some minimum reciprocal chance which is dependent on the quality in question. To see whether this leads to the chance being set to the minimum it is enough to consider the caculation only up to the point where the comparison takes place.

The reciprocal chance at the time the comparison takes place is given by:
Code:
chance(x,q,m)=[128*(rar-[(m-q)/div])/EMF(x)]
where x is MF, q is qlvl and m is mlvl. The value EMF(x) is the effective MF divided by 100 with 1 added. The values div and rar are constants depending on the quality in consideration and whether it is class specific. Also [] means rounding.

We want to know whether the chance is at or above the minimum chance 'min' in all cases, i.e. we care about finding an easy condition equivalent to:
Code:
For all x,q,m: chance(x,q,m)≥min

I'm hoping there's no mistake here (explanations for equivalences in brackets):
Code:
For all x: chance(x,q,m) ≥ min
iff
For all x: [128*(rar-[(m-q)/div])/EMF(x)] ≥ min
iff (the right hand side is an integer)
For all x: 128*(rar-[(m-q)/div])/EMF(x) ≥ min
iff
For all x: rar-[(m-q)/div] ≥ min*EMF(x)/128
iff (maxEMF = maximum of EMF(x) over all x; this is always attained for U/R/S quality, namely by x=a*(a-1)where a is 250/600/500 respectively)
rar-[(m-q)/div] ≥ min*maxEMF/128
iff
rar ≥ min*maxEMF/128 + [(m-q)/div]

So we already eliminated MF out of the question, now on to m and q: The maximum value of m-q is 98 and is for example attained by Hell Baal dropping a Ring.
Code:
For all q,m: rar ≥ min*maxEMF/128 + [(m-q)/div]
iff (98 = maximum of m-q over all m,q)
rar ≥ min*maxEMF/128 + [98/div]
iff
rar-min*maxEMF/128 + [98/div] ≥ 0

Let's call the right side of this test. Then using the values for rar, min and div from itemratio.txt and the maxEMF values we get this table:
uniquecantbottomout.png
Hence chance(x,q,m) never dips below min for unique quality (regular or class specific).

I'm sorry for the the cancer inducing non-TeX formulas and yeah I realize that one could just calc all values for chance in excel (all 2,128,500=250*99*86 of them ;)) and see whether they're bigger than 6400, but this is a lot slicker :D
While I was at it I calculated the amount of MF needed for chances to bottom out on rare and set rolls for each difference mlvl-qlvl.
raresetbottom.png
This also has the 170 MF Gripphon found for rare rings from Council.