Thesis: How the loot tables work - Page 3

Howard Roark · 10/13/07, 7:00 AM

I remember another interesting example of outdoor mob's loot coding from level 60. It deals with the Winterfall Vilage Furbolgs.

They will actually drink the winterfall firewaters that are in their inventory after a seemingly random amount of time. This is very easy to see just by letting them spawn and watching them, if they grow, they've used the firewater. If you kill them right when they spawn, the firewater drop rate is significantly better, but if you watch them drink it and then kill them, it will never ever drop (or so was the way I experienced it from my enchanting grind).
As far as I'm concerned, it's a very good indicator that loot must be calculated on spawn.

redimp · 10/13/07, 3:03 PM

I remember way back in MC, I was guild and raid leader, and would always be the 'seed' for the raids. People were convinced that there was not enough loot variety and that somebody else should seed the raid.
For the following 3-4 weeks I made it known to everybody that I would get somebody else to seed the raid. Lo and behold for those weeks the loot was suddenly varied and useful! What I didn't tell any of the guild save for a few members was that I continued seeding the whole time, and always did except for a couple rare occasions.
Examining the loot from before that period and comparing to within that period and afterwards there was no pattern change at all and it was always random.

People will convince themselves of anything

PSGarak · 10/13/07, 9:50 PM

At the risk of a Useless Post infraction, I just want to say that you are a true american hero. I don't suppose you have screenshots or logs so The Internet will believe you?

Mideci · 10/14/07, 5:05 AM

Well at the further risk of responding to your self-proclaimed useless post, who cares?

His story could be aporcryphal or it could be 100% true. Whatever the case he makes a point that is critical: People will convince themselves of anything.

In the case of streaky loot or loot that is "bad" for them, they concoct impressive fantasies about Rube Goldberg-esque ways of determining boss drops that fail to pass a basic rationality test.

redimp · 10/14/07, 5:52 AM

Originally Posted by PSGarak

At the risk of a Useless Post infraction, I just want to say that you are a true american hero. I don't suppose you have screenshots or logs so The Internet will believe you?

I'm overjoyed, but I am neither a hero or an American.

Feel free to check out the history on www.junglefortress.com (look at the DKP link and check through and you tell me if you can find the 3-4 weeks in which our molten core drops were radically different from the rest...)

I'm not an expert and don't proclaim to be, I can only speak from experience and what I do actually know, which is what I ended my last post with: People will convince themselves of anything.
If you already KNOW that the same person is seeding the instance and KNOW that this is going to cause the same loot to drop all the time, then you will notice those occurences which support this theory. Anything outside of your theory flies under your radar because basically you just aren't interested enough (ie. the priests happen to get a rare drop they never had before, but you're a mage who cares).
When the seed is changed all of a sudden you are actually interested and looking for anything new to drop, and you don't notice the same shit dropping in between.

It's all been said before, I'm not the first person to come up with this theory and I don't claim to be, but I do enjoy discussions even though I doubt anyone will actually truly believe a thing unless Blizzard fully releases the details of their loot system...

redimp · 10/14/07, 5:57 AM

Sorry, I just checked and they don't have the old info available anymore, only BC.

Emeraude · 10/14/07, 6:52 AM

I think the real question on everyone's mind isn't whether the loot is random or not, but if the loot is predictable or not. What raid leader wouldn't kill for the opportunity to know how to get x-drop to happen eh, or know if there was a way to figure out the "randomness" of the system.

You have drop A, and you have drop B.

You know when you get drop A drop B must follow, and perhaps a drop C(Non-Tier item), and maybe even drop D(Legendary).

A->algorithm-result(B)->algorithm-result(C)->algorithm-result(D) the result is based on whatever rules the algorithm makes, if it has to be 1-5, or 1-1,000,000, it'll end up there.

"The generation of random numbers is too important to be left to chance."

That was said, because it's pretty difficult to make a working, and accurate random number generator, that's truly random. Computers take an algorithm, and spit out a random number, the random number is determined by the seed, find out the seed and you can work backwards to figure out what's about to be spit out first. That MIGHT be the only one you can predict, but then the question is when does the random number generator start?

Is it determined by something simple, like raid make-up? Maybe server time? Does it go on an individual basis, with each person having a shadow number attached that it uses, based on past drops? It gives you a headache when you really start thinking about it, but how complex of an algorithm do you think WoW uses to determine all this crap?

Grayson Carlyle · 10/14/07, 7:31 AM

Originally Posted by Junpei

The old list is inaccurate in places with the loot change in SSC/TK

A quick browse around for Solarian (including our own data for tonight)

Table A - Star-Soul Breeches, Worldstorm Gauntlets, Star-Strider Boots, Girdle of the Righteous Path, Vambraces of Ending
Table B - Ethereum Life-Staff, Wand of the Forgotten Star, Trousers of the Astromancer, Heartrazor
Table C - Solarian's Sapphire, Boots of the Resilient, Void Star Talisman, Greaves of the Bloodwarder

Maybe I'm posting a bit late, but our first Solarian kill post-patch was Sapphire, Bloodwarder and Resilient, all from your pool C. All 3 kills since the patch have awarded us a Sapphire, Bloodwarder + 1 other.

On a general note, if you want to know how WoW's item generation and loot tables work, I'd suggest you start with Diablo II.

Macblade · 10/14/07, 10:10 AM

Originally Posted by Bazazu

I realize this is straying from the topic a bit, and of course, it's been discounted a million times, but I'm still convinced that "seeding" is 100% legit and completely broken.

Typically for our raids I start invites. Going back to when we started Karazhan, SSC/TK, and even T6 content, we noticed *insanely* obviously streaks in loot. Basically anytime I started raid invites, we would see the same 4 items per boss. it was like clockwork. When another officer would do invites (Trouble, some of you probably know him on these boards) we would see a distinct shift in loot. Not much, but some.

So about a month ago we changed things up. For 3 weeks, we ran with completely different raid leaders. Grabbed a random paladin to "seed", had our hunter officer seed, and there was a 3rd who I forget now, probably one of our other shadow priests.

The difference? Suddenly after months of seeing the same shit, we started seeing loot tables that were complete flips.

Examples include... Seeing 80% defender token drops for the first 3-4 months of farming ssc/tk. Soon as we switched raid leaders, our mages were praising jebus for finally getting their tokens.

Azgalor dropped the exactly same loot forever. Swapped raid leaders, got a ton of new loot.

And to reinforce this, for the past 2-3 weeks I've been starting raids again and it went right back into our old loot tables that we used to see 2-3 months ago.

So I guess to wrap this up, yes, we see the same thing with items holding a specific loot slot in the table. But in addition to that, I'm pretty pissed off that the loot system is so fucked up that you end up getting the same shit all the time because they can't figure out a creative way to distribute loot evenly.

The fact that some guilds are working on their *third set* of warglaives, and guilds such as ours, EJ(at least, this is what I've heard), and many others have yet to see a single legendary, is complete bullshit.

It is human psychology to find patterns in random data that do not in fact exist. So are you really going to argue that for some reason blizzard tied looting to whoever starts the raid, and that on top of that they rigged it so that this person would be screwed and be more likely to have their raids see the same loot over and over again. OR, is it more likely that you've simply been unlucky. It is possible to had someone a die and have them roll a 1 many times in a row without those dice being loaded or rigged. On top of that, you only see the loot off of a given boss, at most, once a week, which just makes trends like seeing the same loot SEEM much stronger than they probably ARE.

Trey · 10/14/07, 11:21 AM

For Blizzard's purposes, their loot is random. There is some guy looking at a database in California and the numbers they see show that it is indeed random. The numbers appear random because they have the numbers for every single boss kill. You can go through the DKP sites of the top 50 guilds in the world and see that some items drop more often than others and claim it's not random. But how many kills is that? Several hundred? A thousand maybe?

As the sample size gets larger, the numbers come to be more and more of what you'd expect. Are people really so stupid to believe that the Rogue/Mage/Druid token off Illidan drops twice as often as the Paladin/Priest/Warlock? (Illidan Stormrage - NPCs - World of Warcraft) Because the sample size is so small, it appears that because your raid leader's wife has a level 37 druid named Illy, that your guild magically gets more Druid tokens. But when you increase the sample size to a thousand times the number of Illidan kills, the loot distribution is much closer to what you'd expect I.E. Prince Malchezzar: Prince Malchezaar - NPCs - World of Warcraft

Jeffrey · 10/14/07, 11:38 AM

Originally Posted by Howard Roark

I remember another interesting example of outdoor mob's loot coding from level 60. It deals with the Winterfall Vilage Furbolgs.

They will actually drink the winterfall firewaters that are in their inventory after a seemingly random amount of time. This is very easy to see just by letting them spawn and watching them, if they grow, they've used the firewater. If you kill them right when they spawn, the firewater drop rate is significantly better, but if you watch them drink it and then kill them, it will never ever drop (or so was the way I experienced it from my enchanting grind).
As far as I'm concerned, it's a very good indicator that loot must be calculated on spawn.

From my exalted grind, I haven't noticed this at all. Firewater is just a really rare drop and I've had it drop from every creature, buffed or not.

What I have however noticed is that after grouping up a green will drop on the first kill the vast majority of the time. I used to get one every 30 minutes or so but one would always drop after grouping up.
This didn't happen when trying to make it happen (repeating the cycle of grouping, killing, looting, ungrouping) and I haven't seen this behavior since 2.0. But I find it very odd that it happened so often to me.

Kytrarewn · 10/14/07, 11:50 AM

Originally Posted by Grayson Carlyle

Maybe I'm posting a bit late, but our first Solarian kill post-patch was Sapphire, Bloodwarder and Resilient, all from your pool C. All 3 kills since the patch have awarded us a Sapphire, Bloodwarder + 1 other.

On a general note, if you want to know how WoW's item generation and loot tables work, I'd suggest you start with Diablo II.

I don't know why people are convinced that loot went to a 3-table system with the 2.2 change for "3 items to randomly drop".

My honest guess would be that it's

Loot Table [A] = [A1,A2,A3...]
Loot Table [b] = [B1,B2,B3...]

Loot Table ["C"] = [A] whatever the stupid sign is for combination of sets [b], other than the item that had previously dropped.

For example, with Al'ar this week: Phoenix Ring, Claw of the Phoenix, Talisman of the Sun King. That's A, B, A, for comparison.

Hydross: Blackfathom Warbands, Band of Vile Aggression, Living Root.

A,B,A yet again.

I'm not convinced that they've split the loot tables up into an additional group. Be much easier just to put in code to "reroll if non-token double-drop" and go from there.

Hozz · 10/14/07, 3:57 PM

On a related topic, why is PVE loot so...random? Because its always been that way? Is that the way it always should be?

One of the draws with the PVP reward system (both Arena and BG), I am sure, is that it gives the player the power to decide their own reward. You spent your honor or your Arena points how you see fit and you get the reward you want. How popular would the Arena be if when you spent your points you got a random item from the vendor? Not very.

I think raiding would see a little bit of a resurgence if they would make the reward component of PVE more like PVP. Set loot tokens are a small step in that direction but I think they should do more.

Junpei · 10/14/07, 7:06 PM

Originally Posted by Grayson Carlyle

Maybe I'm posting a bit late, but our first Solarian kill post-patch was Sapphire, Bloodwarder and Resilient, all from your pool C. All 3 kills since the patch have awarded us a Sapphire, Bloodwarder + 1 other.

On a general note, if you want to know how WoW's item generation and loot tables work, I'd suggest you start with Diablo II.

Aye, as I said I used a ridiculously low sample to show an idea that things might have changed to a 3 table system. It was always going to be incorrect in many many ways : b

ildon · 10/15/07, 1:08 AM

Originally Posted by Howard Roark

I remember another interesting example of outdoor mob's loot coding from level 60. It deals with the Winterfall Vilage Furbolgs.

They will actually drink the winterfall firewaters that are in their inventory after a seemingly random amount of time. This is very easy to see just by letting them spawn and watching them, if they grow, they've used the firewater. If you kill them right when they spawn, the firewater drop rate is significantly better, but if you watch them drink it and then kill them, it will never ever drop (or so was the way I experienced it from my enchanting grind).
As far as I'm concerned, it's a very good indicator that loot must be calculated on spawn.

This is just a myth. I killed enough of them during Naxx days to know for a fact that whether or not they drank a firewater potion either before or during combat had absolutely no statistically noticeable effect on whether or not they would drop one. I remembered hearing about it, and after that I paid very close attention, and it was definitely not true.

This is coming from someone who got honored with Timbermaw ONLY from killing Winterfall mobs (starting at unfriendly when I wanted to get through the cave without ruining my rep at level ~55).

Trust me, there is absolutely no relation between the mob drinking (or not drinking) a fire water potion and them dropping one.

kervi · 10/15/07, 3:02 AM

Long time ago, around 2 years or so, in MC we killed golemagg and looted him, only to have server crash and rollback for around 15mins. When we came back online, golemagg was alive again and loots have disappeared. We killed him again, and surprise, got exactly same loot. Of course it might have been coincidence, but seemed little too good to be such.

That would speak on behalf that loots are predetermined by something.

Emeraude · 10/15/07, 3:58 AM

Originally Posted by kervi

Long time ago, around 2 years or so, in MC we killed golemagg and looted him, only to have server crash and rollback for around 15mins. When we came back online, golemagg was alive again and loots have disappeared. We killed him again, and surprise, got exactly same loot. Of course it might have been coincidence, but seemed little too good to be such.

That would speak on behalf that loots are predetermined by something.

In contrast to that there was a well known working Karazhan exploit that allowed for Aran to respawn in your instance, and loot would be different upon each kill.

◊ Rob · 10/15/07, 4:40 AM

No contrast there -- those could both be consistent with a "loot determined upon spawn" theory.

Vhad · 10/15/07, 4:43 AM

If the server crashed and Golemagg respawned it would probably mean a very short rollback wouldn't it? Meaning it would be the same Golemagg that already spawned prior to the crash - thus the same loot as it's apparently determined on spawn time.

Dustwhisper · 10/15/07, 5:08 AM

Well relevant or not, in blasted lands up near those summoning/necrolyte circles you could see on the mobs weapons if they would drop a Krul blade or whatever the epic one was. They would have it on their belt if they'd drop one :P

Polleke · 10/15/07, 5:24 AM

Originally Posted by Emeraude

That was said, because it's pretty difficult to make a working, and accurate random number generator, that's truly random. Computers take an algorithm, and spit out a random number, the random number is determined by the seed, find out the seed and you can work backwards to figure out what's about to be spit out first. That MIGHT be the only one you can predict, but then the question is when does the random number generator start?

Is it determined by something simple, like raid make-up? Maybe server time? Does it go on an individual basis, with each person having a shadow number attached that it uses, based on past drops? It gives you a headache when you really start thinking about it, but how complex of an algorithm do you think WoW uses to determine all this crap?

Considering the huge ammount of random numbers needed, every swing and every spell of every mob and every player, loot is just a small thing to determine randomly. My guess is that there is a single seed of the entire realm (or every continent/instance server, whatever, it does not matter) upon start of the server. From that moment every random event is determined by the pesudo random algorithm. It does not matter that a pattern is repeated over billions of iterations, because the actions taken on the server do not follow a pattern.

◊ Cadfael · 10/15/07, 5:58 AM

Originally Posted by Polleke

Considering the huge ammount of random numbers needed, every swing and every spell of every mob and every player, loot is just a small thing to determine randomly. My guess is that there is a single seed of the entire realm (or every continent/instance server, whatever, it does not matter) upon start of the server. From that moment every random event is determined by the pesudo random algorithm. It does not matter that a pattern is repeated over billions of iterations, because the actions taken on the server do not follow a pattern.

A blueposter long ago said once that they are using IBAA as RNG and each process has its own RNG, sharing it with everyone running on this process. However it may very well be that each instance is its own process and thus you really have your own RNG which gets instantiated when the instance spawns. If there is a problem with seeding at this point, you really can get into some trouble from a statistical viewpoint.

However I don't know if instances all have their own process and if that is still this way (or that information was accurate in the first place).

Xandria · 10/15/07, 8:39 AM

Originally Posted by Emeraude

In contrast to that there was a well known working Karazhan exploit that allowed for Aran to respawn in your instance, and loot would be different upon each kill.

I can confirm this, though we didn't do it as an exploit. (Nor was I aware there was one.)

We killed Aran once right as he was sitting down to drink, and his door remained closed. A GM was unable to open the door and told us our only option was to let the instance soft reset and kill him again, so we came back the next night and did just that. The loot was different each time.

We like to joke about loot seeds in our guild but this incident showed that, whatever seeds the RNG, it isn't something as simple as who starts the raid ID or what time the first person zones in. Otherwise, the drops would've been the same both times.

Jaete · 10/15/07, 2:07 PM

One thing that I've been thinking about are the tails of the probability distributions in WoW. If a mob has, say, a 10% chance to drop the quest item you need, then the probability that you have to kill more than 50 mobs to get that item is 0.9^50 ~= 0.5%. This means that if there are 200 people doing that quest, you can expect one of them to need 50 kills or more to complete it; or put another way, if there are 200 servers each with 50 of the quest mobs, on one of the servers none of the 50 mobs will drop that quest item.

So I was wondering whether WoW in fact cuts these tails off somehow, or whether for some reason, such as generally high probabilities or the fact that new mobs will spawn to replace those that were killed, such cutting isn't needed. The optimal design is obviously one that needs no extra adjustments as those most likely aren't trivial.

Of course any answer to this probably can't be much more than speculation.

Riallatar · 10/15/07, 2:46 PM

Originally Posted by Xandria

I can confirm this, though we didn't do it as an exploit. (Nor was I aware there was one.)

We killed Aran once right as he was sitting down to drink, and his door remained closed. A GM was unable to open the door and told us our only option was to let the instance soft reset and kill him again, so we came back the next night and did just that. The loot was different each time.

We like to joke about loot seeds in our guild but this incident showed that, whatever seeds the RNG, it isn't something as simple as who starts the raid ID or what time the first person zones in. Otherwise, the drops would've been the same both times.

If it is based on time --- 7:00 PM Wednesday is different from 7:00 PM Thursday.