Discussion:
Rune drop chance caculations?
(too old to reply)
Oliver Wenzel
2004-08-14 21:36:45 UTC
Permalink
Hi,

does anybody know this? In 75 NM countess runs, I got 153 runes - an
average of 2 per kill/run.
Now, is the drop chance calculated per drop or per rune?

For example, AMN drop chance is 1:15. In 75 runs, I got 4 AMNs, that's
close enough to 75:15=5.

But with 153 runes dropped, shouldn't it be 153:15=10 AMN? Or just a streak
of bad luck statistically?

Puzzled,

Oliver
Dave
2004-08-15 17:55:28 UTC
Permalink
Post by Oliver Wenzel
Hi,
does anybody know this? In 75 NM countess runs, I got 153 runes - an
average of 2 per kill/run.
Now, is the drop chance calculated per drop or per rune?
For example, AMN drop chance is 1:15. In 75 runs, I got 4 AMNs, that's
close enough to 75:15=5.
But with 153 runes dropped, shouldn't it be 153:15=10 AMN? Or just a streak
of bad luck statistically?
Puzzled,
Oliver
Hello, Not sure where you come up with a 1:15 drop chance. (ATMA?)
What follows gets a little confusing if you're not familiar with MPQ
files. (Basically Blizzard's proprietary ZIP file containing the data
tables used by the game)

The following is my attempt to explain the countess drops for NM
Countess, using LoD Expansion, with the 1.10 patch. I used the
D2DataEXP.MPQ file data to try this. As you will see, this disagrees
slightly with the www.d2data.net data on NM Countess. I believe this
comes from the fact that Champs and Bosses get a +1 or +2 boost to
their mlvl (which translates to monster class in the mpq files). If
anyone has further thoughts on the matter, please reply.
=================================================
From what I can tell, The Countess is treated as a Superunique monster
(like eldritch) which is a class 45 monster (From the Superunique.txt
file).

Class 45 is a Corrupt rogue, so we find the appropriate Treasure Class
"Act 1 (N) Unique C" (From the Monstats.txt file). It's interesting to
note here that the CountessRune Treasure Class doesn't exist under
1.10 (please correct me if I'm wrong, but it's not in the new mpq file
that I can see)

Next we look up that Treasure Class to see what drops (In the file
TreasureClassEx.txt).

The "DarkStalker" (Act 1 Lvl 45 monster) drops from three classes:
Act 1 (N) H2H C, Act 1 (N) Champ C, and Act 1 (N) Unique C.

The H2H class contains Act 1 (N) Equip B (covered below), Act 1 (N)
Junk, and Act 1 (N) Good (also covered below). Junk basically
contains gold, bolts/arrows, thaw/stamina/antidote pots, keys, TP/ ID
scrolls, and exploding/gas potions.

The Act1 (N) Champ C class contains Act 1 (N) Citem C, and Act 1 (N)
Cpot C (covered below).
CItem C contains Gold x 1280, Act1 (N) Equip B (covered below), and
Act 1 (N) Good (covered below).

The Unique C class yields Act 1 (N) Uitem C with a probability of 1,
and Act 1 (N) Cpot C with a prob of 2.

The CPot group yields pots ranging from hp4 to hp5, mp4 to mp5, and
rv1. (Basically the two biggest Health and Mana potions, and the
little purple pot.) You could double check this by seeing if NM
Countess (not minions) ever drops a thawing potion or one of the minor
health/mana pots.

The Uitem yields: other sub classes, which in turn have other sub
classes. I carefully tracked each one and came up with this:
Armo and Weap classes 42 and down.

Uitem also yields Act 1 (N) Good, which contains chipped through
flawless gems, Jewelry C, and Runes 8.

Jewelry C contains rings, amulets, and all 3 charm sizes.

Runes 8 contains r15 (HEL) and r16 (IO), and Runes 7. (Rune 7 is r13,
r14, and Ruse 6. The same pattern continues down to EL.
===================================================
Okay, look at www.d2data.net, you can see that Countess drops all of
the above, PLUS armor and weap 45, AND r17 (LUM) and r18 (KO).
I'm assuming that the folks that made this website know better than I
:). Thus I assume that the extra items are coming from a Champ/Boss
boost to the mlvl. (I recall reading this somewhere else too...)

Taking all that into account, it still seems like countess has a
better drop chance for runes than is suggested. Can anyone confirm or
deny the existence of a special "no drop" conversion to runes? (I read
that somewhere else too... but it might have applied only to 1.09 or
earlier.)
David Carson
2004-08-15 21:45:05 UTC
Permalink
Post by Dave
From what I can tell, The Countess is treated as a Superunique monster
(like eldritch) which is a class 45 monster (From the Superunique.txt
file).
Class 45 is a Corrupt rogue, so we find the appropriate Treasure Class
"Act 1 (N) Unique C" (From the Monstats.txt file). It's interesting to
note here that the CountessRune Treasure Class doesn't exist under
1.10 (please correct me if I'm wrong, but it's not in the new mpq file
that I can see)
Err, no, the Countess has her own Treasure Class: "Countess", "Countess
(N)", "Countess (H)". They're the last three lines in TreasureClassEx.txt.

Even if she didn't, she wouldn't use the Corrupt Rogue's treasure class
- she'd use one of the superunique classes, like "Act 1 Super C".
Post by Dave
Okay, look at www.d2data.net, you can see that Countess drops all of
the above, PLUS armor and weap 45, AND r17 (LUM) and r18 (KO).
I'm assuming that the folks that made this website know better than I
:).
Don't assume that, that website is horrifically incorrect and has been
ever since the release of 1.10.
Post by Dave
Taking all that into account, it still seems like countess has a
better drop chance for runes than is suggested. Can anyone confirm or
deny the existence of a special "no drop" conversion to runes? (I read
that somewhere else too... but it might have applied only to 1.09 or
earlier.)
I'd explain it, but I have to go to work now. You can find the
Countess's special drop documented in many places, but if you have no
luck, let me know, and I'll write up on it this evening..

Cheers!
David...
Dave
2004-08-17 16:49:03 UTC
Permalink
On Mon, 16 Aug 2004 07:45:05 +1000, David Carson
Post by David Carson
Post by Dave
From what I can tell, The Countess is treated as a Superunique monster
(like eldritch) which is a class 45 monster (From the Superunique.txt
file).
Class 45 is a Corrupt rogue, so we find the appropriate Treasure Class
"Act 1 (N) Unique C" (From the Monstats.txt file). It's interesting to
note here that the CountessRune Treasure Class doesn't exist under
1.10 (please correct me if I'm wrong, but it's not in the new mpq file
that I can see)
Err, no, the Countess has her own Treasure Class: "Countess", "Countess
(N)", "Countess (H)". They're the last three lines in TreasureClassEx.txt.
Have you extracted the TreasureClass.txt file since 1.10? 1.10 brought
about some significant changes in The last lines in mine are for
Smith, and countess is nowhere to be found. (She also doesn't appear
in Monstats.txt.) Maybe I'm just not extracting my MPQ files
correctly. (Using MPQView.)
Post by David Carson
Even if she didn't, she wouldn't use the Corrupt Rogue's treasure class
- she'd use one of the superunique classes, like "Act 1 Super C".
I can buy this, but since she shows up in Superuniques.txt, I figured
the game uses the class listed here.
Post by David Carson
Post by Dave
Okay, look at www.d2data.net, you can see that Countess drops all of
the above, PLUS armor and weap 45, AND r17 (LUM) and r18 (KO).
I'm assuming that the folks that made this website know better than I
:).
Don't assume that, that website is horrifically incorrect and has been
ever since the release of 1.10.
That's why I tried to dissect her drops from the MPQ files. :)
I have found the website to be pretty accurate though, and a good
pointer towards where to MF for certain items. You do have to balance
some of the data against monster density for an accurate idea of where
to MF. (Cows for example, are listed with lower probabilities than
most other monsters. But don't forget there are a ton of cows
available in a small area, so your odds go up for finding an item in a
certain amount of time, if you can kill them fast enough.)

I will say this: using the MPQ files, I was able to come pretty damn
close to what the website listed, without trying hard. It's just
little nit-noid details that I have to guess at which keep me from
getting spot-on.
Post by David Carson
Post by Dave
Taking all that into account, it still seems like countess has a
better drop chance for runes than is suggested. Can anyone confirm or
deny the existence of a special "no drop" conversion to runes? (I read
that somewhere else too... but it might have applied only to 1.09 or
earlier.)
I'd explain it, but I have to go to work now. You can find the
Countess's special drop documented in many places, but if you have no
luck, let me know, and I'll write up on it this evening..
Cheers!
David...
David Carson
2004-08-17 20:49:07 UTC
Permalink
Post by Dave
On Mon, 16 Aug 2004 07:45:05 +1000, David Carson
Post by David Carson
Err, no, the Countess has her own Treasure Class: "Countess", "Countess
(N)", "Countess (H)". They're the last three lines in TreasureClassEx.txt.
Have you extracted the TreasureClass.txt file since 1.10? 1.10 brought
about some significant changes in The last lines in mine are for
Smith, and countess is nowhere to be found. (She also doesn't appear
in Monstats.txt.) Maybe I'm just not extracting my MPQ files
correctly. (Using MPQView.)
Aha, I think I see the problem - you don't want to be looking at
TreasureClass.txt, you want TreasureClassEx.txt!
Post by Dave
Post by David Carson
Even if she didn't, she wouldn't use the Corrupt Rogue's treasure class
- she'd use one of the superunique classes, like "Act 1 Super C".
I can buy this, but since she shows up in Superuniques.txt, I figured
the game uses the class listed here.
But the classes listed in Superuniques.txt ARE "Countess", "Countess
(N)", and "Countess (H)" ?
Post by Dave
Post by David Carson
Don't assume that, that website is horrifically incorrect and has been
ever since the release of 1.10.
That's why I tried to dissect her drops from the MPQ files. :)
I have found the website to be pretty accurate though, and a good
pointer towards where to MF for certain items.
I probably exaggerated by saying "horrifically" incorrect. But it
totally ignores the monster TC scaling to area level in 1.10 (so it's
only useful at all for bosses and special superuniques with their own
treasure classes), for a start.

But they're also getting their percentages wrong.. e.g. look at
Skeleton, act 1, normal. One of the simplest things in the game. They
have light healing potion and minor mana potion both at 8.82%, but if
you trace their TC from "Act 1 H2H A" to "Act 1 Junk" to "Potion 1",
you'll see that those two should NOT have the same chance?! I don't know
where d2data.net get their numbers, but they're just wrong. And if the
simple things are wrong, I have no faith in the complicated ones.

The drop calc in ATMA seems to be the best, but even that has a few
things I'm a little dubious about, I've been a bit slack with my bug
reports though..

Cheers!
David...
Oliver Wenzel
2004-08-17 21:30:25 UTC
Permalink
Hi,
Post by David Carson
But they're also getting their percentages wrong.. e.g. look at
Skeleton, act 1, normal. One of the simplest things in the game. They
have light healing potion and minor mana potion both at 8.82%, but if
you trace their TC from "Act 1 H2H A" to "Act 1 Junk" to "Potion 1",
you'll see that those two should NOT have the same chance?! I don't
know where d2data.net get their numbers, but they're just wrong. And
if the simple things are wrong, I have no faith in the complicated
ones.
The drop calc in ATMA seems to be the best, but even that has a few
things I'm a little dubious about, I've been a bit slack with my bug
reports though..
I'm not so much into these specifics, but my initial question was if
those drop chances were per kill or per single rune drop? So if countess
decides to drop 3 runes do I have 3 chances for IO at 1:85?

Anyhow, I've just finished 250 runs and got the following runes:

15 AMN, 8 SOL, 5 SHAEL, 6 DOL, 2 HEL, 3 IO and a LUM (1:15000). 2-3 AMN
were from Cubing, though.

Best was a HEL and a IO in one drop and that single LUM.

As I average 2 runes per kill, the numbers tend to per kill and not to
per rune...

Regards,

Oliver
Post by David Carson
Cheers!
David...
Dave
2004-08-18 00:02:28 UTC
Permalink
On Wed, 18 Aug 2004 06:49:07 +1000, David Carson
Post by David Carson
Post by Dave
On Mon, 16 Aug 2004 07:45:05 +1000, David Carson
Post by David Carson
Err, no, the Countess has her own Treasure Class: "Countess", "Countess
(N)", "Countess (H)". They're the last three lines in TreasureClassEx.txt.
Have you extracted the TreasureClass.txt file since 1.10? 1.10 brought
about some significant changes in The last lines in mine are for
Smith, and countess is nowhere to be found. (She also doesn't appear
in Monstats.txt.) Maybe I'm just not extracting my MPQ files
correctly. (Using MPQView.)
Aha, I think I see the problem - you don't want to be looking at
TreasureClass.txt, you want TreasureClassEx.txt!
My typo - I meant TreasureClassEx.Txt. (Ex for LoD Expansion, no Ex
for Classic D2). But still, in my TreasureClassEx.txt file, last lines
are for Smith... no sign of Countess at all.
Post by David Carson
Post by Dave
Post by David Carson
Even if she didn't, she wouldn't use the Corrupt Rogue's treasure class
- she'd use one of the superunique classes, like "Act 1 Super C".
I can buy this, but since she shows up in Superuniques.txt, I figured
the game uses the class listed here.
But the classes listed in Superuniques.txt ARE "Countess", "Countess
(N)", and "Countess (H)" ?
The Superuniques.txt file I'm staring at doesn't list TreasureClasses
(like Monstats.txt does...)
Instead I have lines that look like this:
Name Class Mod1 Mod2 Mod3 MinGrp MaxGrp EClass
AutoPos Stacks
The Countess 45 9 0 0 6 6 0
1 0

Thats it. Under 1.09, these files looked very different - there was
indeed a CountessRune class. (Remember the old "TC87" classes? Gone
now; replaced with a "better" tree-based class system, with up to 10
"sub classes" per class, where each sub class can either be an item
class, like "r07" (for rune #7, or another Treasure Class.)
Post by David Carson
Post by Dave
Post by David Carson
Don't assume that, that website is horrifically incorrect and has been
ever since the release of 1.10.
That's why I tried to dissect her drops from the MPQ files. :)
I have found the website to be pretty accurate though, and a good
pointer towards where to MF for certain items.
I probably exaggerated by saying "horrifically" incorrect. But it
totally ignores the monster TC scaling to area level in 1.10 (so it's
only useful at all for bosses and special superuniques with their own
treasure classes), for a start.
True, www.d2data.net does NOT have a party size option, which
certainly changes some percentages, and also doesn't include the
area-level overide.

(Side note to readers: each map has an "area level", like NM Flayer
Jungle is alvl 50. If a lvl 47 monster spawns in this map, it's mlvl
is adjusted to alvl 50. Chests are also set to alvl. Uber chests are
also alvl, but larger(double?) drops.There is a related feature for
"bosses": a Champ is either mlvl or alvl+2, which ever is greater. A
pack boss is set to alvl+3 or mlvl, whic ever is greater. Act boss
mlvls are always far above the alvl, and aren't affected by this
override)

But how could you build a drop calculator by specific monster type
(not counting bosses that spawn in known alvl maps), since you can
have "Guest" monsters from earlier acts now just about anywhere. My
personal guess is that the www.d2data.net website simply lists the
base mlvl for all given monsters, and ignores any possibility of it
being a guest in a later map. ATMA does a decent job by showing the
normal locations where a monster can spawn. I don't think it does true
random guest appearances though.

Notice that Guest monsters seem to get better drops than they would if
they appeared in their home maps. (caused by the alvl override)

My personal theory is that Superuniques are now treated as Champs, and
thus they get alvl+2 as their mlvl. The tricky part comes with trying
to guess how Blizzard figures out Treasure Classes from these
overrides. I think I see how this happens: For each monster type there
are 4 Treasure Classes listed. I believe TC1 is the regular monster,
TC2 is for Champs, and TC3 is for Pack Bosses. (The data for TC4 is
blank for norm, NM, and hell) Now my only question is does a Champ get
just TC2, or does he get TC1 and TC2.
Post by David Carson
But they're also getting their percentages wrong.. e.g. look at
Skeleton, act 1, normal. One of the simplest things in the game. They
have light healing potion and minor mana potion both at 8.82%, but if
you trace their TC from "Act 1 H2H A" to "Act 1 Junk" to "Potion 1",
you'll see that those two should NOT have the same chance?! I don't know
where d2data.net get their numbers, but they're just wrong. And if the
simple things are wrong, I have no faith in the complicated ones.
This is kind of fun. (For those who think this post is WAY over your
head, don't worry, you can enjoy Diablo2 at many many levels of
complexity. It took me 9 months to work my way down to the MPQ files
and stand in awe of the inner workings of Blizzard's programs.)

Ok: Starting with a Norm Skeleton (LoD Exp of course), and no boss or
champ upgrade to worry about.
I assume that D2 adds up the probability numbers, then generates a
random number from 1 to X, and chooses the appropriate class. Not sure
how or when "no drop" is calculated.

From Monstats.txt, TC1 should be his starting class. That is "Act 1
H2H A".

Act1H2HA gives us gld-21 (item type, probability), Act1EquipA-16,
Act1Junk-21, Act1Good-2. (35% Gold, 27% Equip, 35% Junk, land 3% of
Good)

<<<First TC sub class>>>
Not sure how max gold value is determined. Maybe random from 1 to
mlvl)

<<<Second TC sub class>>>
Act1EquipA gives us weap3-7 and armo3-7. (50% weapon, 50% armor.)

Act1Junk gives us Potion1-8, Misc0-4, and Ammo-4. (50% Potion, 25%
Misc, 25% Ammo)

Potion 1 gives us hp1-12, hp2-4, mp1-5, mp2-4, rvs-1, vps-4 (40% minor
health, 13% light health, 17% minor mana, 13% light mana, 3% minor
rejuv, 13% stamina)

Misc 0 gives us key-3, tsc-3, isc-3, opl-2, gpl-2 (23% key, 23% town
scroll, 23% id scroll, 15% minor exploding potion, 15% minor gas
potion)

Ammo gives us aqv-1,cqv-1 (50% arrows, 50% bolts)

<<<Third TC sub class>>>
Act1Good gives us JewelryA-5, ChippedGem-5. (50% Jewelry, 50% Chip)

Jewelry A gives us rin-8, amu-4, jew-1, cm3-2, cm2-2, cm1-2 (42% ring,
21% amulet, 5% jewel, 11% small charm, 11% medium charm, 11% large
charm.)

<<<Compiling end probabilities>>>
Here's what I come up with taking percents of percents:
35.00000% gold
13.33333% armo3
13.33333% weap3
7.24138% hp1
1.81034% hp2
3.01724% mp1
2.41379% mp2
0.60345% rvs
2.41379% vps
2.01923% key
2.01923% tsc
2.01923% isc
1.34615% opl
1.34615% gpl
4.37500% aqv
4.37500% cqv
0.70175% rin
0.35088% amu
0.08772% jew
0.17544% cm3
0.17544% cm2
0.17544% cm1
1.66667% chippedgem

If you try just adding up the probability number values, you get:
20.38835% gold
6.79612% armo3
6.79612% weap3
11.65049% hp1
2.91262% hp2
4.85437% mp1
3.88350% mp2
0.97087% rvs
3.88350% vps
2.91262% key
2.91262% tsc
2.91262% isc
1.94175% opl
1.94175% gpl
0.97087% aqv
0.97087% cqv
7.76699% rin
3.88350% amu
0.97087% jew
1.94175% cm3
1.94175% cm2
1.94175% cm1
4.85437% chippedgem

Neither method comes up with what www.D2Data.Net lists.

And also, there is no "No drop" probability listed, but we all know
that taking a bare lvl 1 char into the den and only wacking skeletons,
you do get a lot of no-drops.

I couldn't get ATMA to display a similar simple drop percentage.
Post by David Carson
The drop calc in ATMA seems to be the best, but even that has a few
things I'm a little dubious about, I've been a bit slack with my bug
reports though..
Cheers!
David...
It would really be interesting to see a flow chart of the actual
program code to see how drops actually get calculated. Without a
doubt, the math behind Diablo2 is mind bogglingly complex. Of course,
we don't see the time spent developing these tables and algorhythms
except the end result: "Oooh looky looky, a unique Shako!". And
that's all that matters in the end.

Best of drops to y'all
The other Dave....

Loading...