What confuses me is the logic behind this system of pools for raid bosses. What is the advantage of having a system pick two numbers between 1/6, each corresponding to a pool instead of having a system pick two numbers between 1/12. In the second scenario, any two items can be chosen at random (depending on the "randomness" of the generator) instead of having the first chosen item drastically effect the second item.
If item1/pool1 is chosen, items2-6/pool1 can not be chosen but items1-6/pool2 can be. Under a truly random system, item1 comes on the first roll then any item2-12 can come on the second roll.
The only advantage would be to split up the pools according to DPS/healing/tanking. I see that, as far as the OP's lists, that is the general pattern. The problem is that it is only a general pattern and the pattern varies. Some are DPS/healing with tank gear thrown in. Some are melee DPS/caster DPS with healing and tanking thrown in.
What advantage does the system hold if it's pattern is quickly broken on the next boss?
But the Diablo II loot system was quite predictable at least in single player. It was simply based on the mobs position when it died.
What?
Originally Posted by Grayson Carlyle
Not even close. Did you even play the game, let alone write mods and loot tables for it? It's determined at spawn based on a cascading decision tree.
Exactly. It was pretty easy to see that even for a non-coder like me as soon as I started fiddling with the files.
Originally Posted by Grayson Carlyle
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.
-
It has always been and probably always will be the same as it was in Diablo II. It's an incredibly fast and versatile system, you can literally make loot table rules anything you can imagine.
Which is why they adopted it into WoW. The whole [ilvl to mlvl and clvl]* system is very good and allows for lots of customization. Items with random stats - greens and blues from Outland-rarespawns - follow strict rules embedded in this system while customized items, i.e. pvp/quest-items and instance-blues/purples, are made by the itemization team and thereby placed accordingly without regards for the system, but rather where they fit according to progress/power.
* D2-terms
ilvl = item level; clvl = character level required to use it; mlvl =mob/monster level required for it to appear on the table
Pre-edit:
This post didn't come out as I wanted it to, so I'll end it here. This is pretty much my stance too:
Originally Posted by Zifna
I'd say it adds suspense, to a certain point. If you know what you're getting no one's too excited about it one way or another, though the recipients may very well be pleased. But you wouldn't get everyone crowding around a corpse waiting for the ML to get rezzed and excitedly speculating about what *might* be on there.
The flip side to this is of course the irritation that people feel when a boss drops loots that were not what was desired by anyone. I would say a combination of guaranteed loots and randomness is optimal. See: the low number of people complaining after you drop Void Reaver, Kael, Vashj etc.
but you have PLATE ARMOR.
ITS PLATE. THAT YOU WEAR. PLATES OF METAL.
If a hundred thousand people roll that same die 6 times, and get one of each number, something is probably cooked. Yes, it *could* be random, but the less entropy your results contain, the less likely it is that your result is random.
Fair enough, though I'm sure a lot of people think that "random" implies "evenly distributed" in the first place. I would be willing to bet that if we did see results like that, the majority of players would actually be *less* inclined to think that the system is broken or can be manipulated.
People say they don't like a random number generator, but what they really mean is that they don't like a memoryless random number generator that has a statistically normal chance of streaky or clumpy behavior. It is possible to create a random number generator that normalizes against previous outcomes; I found out about Low-Discrepancy Sets while trolling wikipedia for monte-carlo integration. While it does make the gear drop significantly less random (obviously--no longer independent events), a LDS with a random or pseudo-random seed is more in line with what people expect a truely random sequence to be. Moreoever, it generates a more "fair" drop pattern by actively working against streaks.
There are ways to finnagle this to make it more random while still keeping the advantages: items could have two or three noncontiguous blocks across the loot set would allow for 2- or 3-long steaks but not much longer, or it could randomly seed the first 3-5 and start LDS after that.
The ironic thing is that, in order for loot seeds to persist across multiple raids the loot and loot seed would have to be tied to the raid ID or raid leader in some way; in other words, the fix to the problem is actually the commonly percieved cause of the problem. Which makes sense in a way... people think it's not random, but actually want less random instead of more.
This would entail a gross, and specialized, overhaul of the loot system though, in a way that makes instances behave differently than normal mobs (unless you want mote farming to be a LDS too) so I wouldn't expect it. Still, it's something to think about. I'm trying to do my part to spread the word that in most non-cryptography applications, LDS is usually better than random or pseudorandom.
And for everyone who says that it's not possible, why wouldn't they just use a simpler more reliable method, there's always the ridiculous wi flag story as proof that these things can happen in the morass that is MMO code.
On the other hand.. the Wi Flag made sense. It was just a stupid math error, linked with the fact that characters tended to be processed in a standard order. The writeup, from srand, the developer who finally found it had the math making it clear. It was the sort of error I could see doing myself.
(Monster target selection was pick a person, with a chance inversely proportional to the distance the person was from the monster. So most of the time, the foe should pick the person next to them. But randomly, it was permitted to run after someone further away.)
The bug with the Wi flag was incredibly hard to find.
Apparently the person's character ID number was used in some sort of target selection table and rather than the number of people in the room and distance from the MOB being set to a number from 0-1, based on range, it was somehow being ADDED, so the total would very likely be greater than 1, which it shouldn't have been, and what made that so hard was that some people had no chance of being selected as a target by the MOB, (they came in above the 1 mark which wasn't considered) while others, thanks to their character's ID number, were always in the 0-1 range, and unusually likely to be chosen as a target.
Poor Wi seemed to always be selected over his compatriots, and wailed upon by said MOB.
Srand finally tracked down the bug and fixed it in one of those epiphanous "AHA!" moments, and thus the Wi Flag was exorcised from AC forever.
By the way, I think I still have my official "Wi Flag" card somewhere.
(So basically, you had a 1 roll table created on monster spawn. (Person A (far) = 1%, person B (middling)= 10%, person C (close) = 50%, ..... Person Q = 5%. Total sum= 100%) Roll a number from 0..99%, find who was picked..... Unfortunately, the table summed up to something like 800%, so the actually chances used was (Person A = 8%, person B = 80%, Person C = 12%, person D..Q = 0%) And people were just roughly processed by some sort of usernumber. So Person A with a low number tended to be at the start while Person Z with a high one, tended to the end.
Last edited by Fringe_Worthy : 10/17/07 at 6:25 PM.
People say they don't like a random number generator, but what they really mean is that they don't like a memoryless random number generator that has a statistically normal chance of streaky or clumpy behavior. It is possible to create a random number generator that normalizes against previous outcomes; I found out about Low-Discrepancy Sets while trolling wikipedia for monte-carlo integration. While it does make the gear drop significantly less random (obviously--no longer independent events), a LDS with a random or pseudo-random seed is more in line with what people expect a truely random sequence to be. Moreoever, it generates a more "fair" drop pattern by actively working against streaks.
There are ways to finnagle this to make it more random while still keeping the advantages: items could have two or three noncontiguous blocks across the loot set would allow for 2- or 3-long steaks but not much longer, or it could randomly seed the first 3-5 and start LDS after that.
The ironic thing is that, in order for loot seeds to persist across multiple raids the loot and loot seed would have to be tied to the raid ID or raid leader in some way; in other words, the fix to the problem is actually the commonly percieved cause of the problem. Which makes sense in a way... people think it's not random, but actually want less random instead of more.
This would entail a gross, and specialized, overhaul of the loot system though, in a way that makes instances behave differently than normal mobs (unless you want mote farming to be a LDS too) so I wouldn't expect it. Still, it's something to think about. I'm trying to do my part to spread the word that in most non-cryptography applications, LDS is usually better than random or pseudorandom.
I have long believed this is a great idea that will never happen -- and probably shouldn't happen.
Who's to say what the group really is that has the "loot string"? Is it the guild? A preponderence of the guild? What?
It introduces some complexity in "loot memory" and really would please many people. The problem is getting it right, determining when you are finally going to "get something" and worrying about exploiting. If a guild knows it hasn't gotten Superdeluxe Item XYZ for 12 weeks and it's going, it could sell the right to be there...
Is this worse than what goes one? I guess not.
But it does make for attendance issues and so many problems.
Like I said, a great idea, that should never happen.
the problem with implementing loot memory is simple. a) it would lead to abuse and b) whose loot memory do you choose?
Example: We have up to 3 kara Ids in a week, 1 the next week, then 2, whatever. I may not have an Id this week, yet be raidleader next week. Someone from id1@week1 maybe 0@2 and then 2@3. And another member was in 1@2, but also 2@2 with a twink.
Do you take my memory, or those of the other one? Or those of the twink? or what?
Not to mention that this would lead to situations where you could predict and sell the drops with increasing chance of success.
Hey, DST has not dropped for 3 weeks, this week it will drop. Hmm, should I take rogue1 my good buddy, or should I take along the upstart enhance shaman with more DKP? Hmm, let me see...
Even better, if this was implemented and had not added at least some random element into it anyway, you could be able to MULTIPLY the drops of the specific items. You have 20 people which can force a DST drop, you can sell those IDs to other people. How much would your rogues/enh shamen/retpalas/warriors pay for a certain DST?
We've gotten Talon from Al'ar way to many times now :| I do though think that the loot-distribution is completely random and that we will never discover the so called "seed". With tier-drops we've seen some nice congestion at times but then suddenly a swap, it happens. I'm well happy we got like 2 hero tokens on Vashj like three times in a month or something =) Only thing I'd wish for was generally an increase in item-drops from Bosses because we've only seen one Vestments of the sea-witch which I was lucky enough to pike.
My guild has 1 Vestments and 6 Krakken Hearts. Random is definitely not the way to go. Our luck has been so bad that we try to guess which offspec/shard plate/mail loot is going to drop. Raid itemization is incredibly bad for 25-mans.
Simple solution would be to have the raid ID stay for more than one week. Say you are tied to a raid ID for 4 weeks, but once a week everything in there respawns. This would more or less kill some PUG groups in Karazhan, but keeping loot in memory would be no problem.
After the 4 weeks, the LDS could reset, this allows for atleast some more varied loot. Not sure how this LDS systsem works, but it makes sense every drop should be able to come each week, but chance decreases if it has already dropped during that 4 week period. It would still be quite likely to get 2 of the same item during the same 4 week period, but chances for getting 3 or 4 would be close to zero (still possible though). This makes it hard to guess what will drop, just you can have some guess of what will not drop.
I don't see how this will cause any more attendance issues than today. If you know that X is not likely to drop, it would mean the same as you already got the item today and not coming cause of that. Or you are not coming on that specific boss since it has no loot wou want.
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.
Even if we assume loot is generated at spawn I highly doubt the periodic server updates store loot that was generated for live NPCs. To do so would store large amounts of worthless data with every update and provide no benefit.
And then we have another problem. *Your* raid doesn't want to see an item drop 3x in a row because you lack people that use swords/cloth pants that weren't tailored/wands/whatever who care.
*Their* raid craves the item 3x in a row because they have no tailors/everyone wants fist weapons/whatever.
Now are we supposed to pre-program the loot "winnower" to meet the needs of each guild. In Kara it's readily apparent this idea wouldn't even get a nanosecond of consideration from Blizzard (ditto ZA I'd guess). Distributing all the loot from the table isn't particualrly useful for a 10-man nor is denying the raid the same drop more than once.
In a 25-man it's more pleasing but unless they weight the drops to meet your needs, it's still fairly unpleasing. About the only way this could work is if there were a currency dropped by bosses that could (a) be turned into the items over time -- a guarantee of achieving the items evenutally or (b) some way of using the currency to turn Item X into Item Y when you have enough of it. I don't really like any of these ideas even as a write them.
And let me be clear, I want "my turn" to come up as much as the next guy instead of waiting 6 weeks for the right pants token to drop or whatever. I want the game to remember that we never got XYZ and give it to us. I just can't see how it's going to do this in a useful, agreeable manner such that people actually prefer it to the existing system.
My guild has 1 Vestments and 6 Krakken Hearts. Random is definitely not the way to go. Our luck has been so bad that we try to guess which offspec/shard plate/mail loot is going to drop. Raid itemization is incredibly bad for 25-mans.
I could not agree with that statement more. Its so ridiculous to get niche spec loot drops week after week. Or the same goddamn drop 5 times in a row.
Everything should be tokenized IMO and/or they just dramatically shrink loot tables.
I'm going to have to agree that random is just not the way to go. A pseudo-random generator with long-term memory just bandaids loot distribution issues - getting fewer Talons of Al'ar wouldn't help. The majority of loot just isn't well-suited to being distributed randomly. The occasional item that is somewhat more powerful, sure, but being stuck several tiers of content behind gear-wise because the RNG hates you is not a good thing, even if it's only a few slots.
I could not agree with that statement more. Its so ridiculous to get niche spec loot drops week after week. Or the same goddamn drop 5 times in a row.
Everything should be tokenized IMO and/or they just dramatically shrink loot tables.
Well, see your problem is not loot randomization necessarily. It's possible that the proportions should be altered so that oddball niche loot is less common than more uniformly desirable loot.
I personally wouldn't mind everything being tokenized but under many dkp systems unaltered, that would basically lead to "he with the most points gets everything first". Those problems are far bigger than bad luck loot streaks, from my experience.
Shrunken loot tables is also reasonable, especially if serendipity is still preserved by the "third drop" off a boss where the oddball loot might come from a somewhat longer table.
That said, your opinions and Blizzard's about loot is obviously different with the generic you often having a much better idea about what's interesting. I mean we've all seen for 2 years epic boss drops that are just not appealing. And yet a Blizzard itemizer thought they were.
Would you want Vendorstrike to be one of only 4 drops off the boss?
I personally wouldn't mind everything being tokenized but under many dkp systems unaltered, that would basically lead to "he with the most points gets everything first". Those problems are far bigger than bad luck loot streaks, from my experience.
I agree, though it depends on how the guild handles loot. If bosses drop 2-3 "marks" instead of actual items, then the marks are redeemed at a vendor on 1:1 price for the conventional armor pieces there would be no problem with not getting the right items. Then it would come down to the guild saying you can only get 1 mark per boss and the top 3 people in DKP get a mark. I could see it working but it is much more dependent on effective loot management in guild.
However, problems arise such as voids. Crappy loot funds enchants, and if you get no crappy loot until everyone has everything they could possibly need from that boss then that shiny new staff may be shy a Sunfire until 150-200 (assuming crystal prices increased) g for the mats on top of what isn't acquired in raids like the ungodly amount of LP shards needed for high end enchants.
I see the current loot system as very annoying and could face improvements but they shouldn't be too much. If it is entirely redesigned in to a token system, we'd end up dealing with many more flaws then the current system has. It's the lesser evil but random loot is still a pain.
I personally wouldn't mind everything being tokenized but under many dkp systems unaltered, that would basically lead to "he with the most points gets everything first". Those problems are far bigger than bad luck loot streaks, from my experience.
My suggestion to this would be to basically keep the t4+ token system, but
a) more tokens! Especially Kara is so annoying as that you most likely won't have one of each class and spec, hell it's not even remotely possible (*moan* Grasp of the Dead *moan*)
or
b) being able to trade them in for an additional "fee" - so all your warriors/druid/priests already got their glove token and it drops again? give it to a shaman and he has to pay some voids (or whatever) to trade it in for the Shammy token. This would counteract the absolute abundance of having to AH your Voids after a certain point as you don't get enough upgrades to reapply Soulfrost/Sunfire/Mongoose. And I suppose many of the people unlucky with drops would be quite thankful to be able to trade that 10 Kara items they saw drop and be disenchanted repeatedly to trade in for their loot, although it may cost a bit. At least I would have done it all the time.
I agree, though it depends on how the guild handles loot. If bosses drop 2-3 "marks" instead of actual items, then the marks are redeemed at a vendor on 1:1 price for the conventional armor pieces there would be no problem with not getting the right items. Then it would come down to the guild saying you can only get 1 mark per boss and the top 3 people in DKP get a mark. I could see it working but it is much more dependent on effective loot management in guild.
However, problems arise such as voids. Crappy loot funds enchants, and if you get no crappy loot until everyone has everything they could possibly need from that boss then that shiny new staff may be shy a Sunfire until 150-200 (assuming crystal prices increased) g for the mats on top of what isn't acquired in raids like the ungodly amount of LP shards needed for high end enchants.
I see the current loot system as very annoying and could face improvements but they shouldn't be too much. If it is entirely redesigned in to a token system, we'd end up dealing with many more flaws then the current system has. It's the lesser evil but random loot is still a pain.
As long as the tokened loot can be D/E'ed, I see no problem with it. The guild disenchanter can just pick up the token.
But if you mean there would be more of a shortage (at least in the short term) than there is now of Void Crystals, yeah I can see that. But isn't it better to have the item you want and have it unenchanted than to have all the Void Crystals in the world and no item?
The loot distribution system is most of the problem, but it's not all of the problem, and not necessarily even the best way to go about solving it. The fact that there is shitty loot on the drop table in the first place is a problem, as is the fact that certain drops are definitely more desirable than others. Sharding your tenth breastplate of annihilation is aggravating, but sharding your very first vendorstrike says it shouldn't be on the loot table in the first place. It's similarly out of line that there are no upgrades for DST, or even nelth's tear, for a long-ass time.
To an extent, there's also the issue of certain items being very specialist pieces of gear, so you need them to drop some but not much. Having better itemization, and class mechanics that allow for better itemization, can account for a lot. For example, the change to druid mechanics making them favor agi and have no preference for str vs ap means they now share the same dps itemization with rogues--this eliminates the need for specialist str leather, allowing them to tighten up the loot tables a bit.
Has anyone had any luck with getting loot that you haven't had drop in like the first 20 kills (excluding legendaries). Like a theory (Who is ML, First to zone in, First to tap, etc.) I know there is no way of knowing exactly how the loot system works. I remember pre BC switching the person who ML'd, started the raid, and zoned in first every week and getting complete new items and a more even spread of the total loot table. Has anyone else experienced a theory that seems like it works?
Has anyone had any luck with getting loot that you haven't had drop in like the first 20 kills (excluding legendaries). Like a theory (Who is ML, First to zone in, First to tap, etc.) I know there is no way of knowing exactly how the loot system works. I remember pre BC switching the person who ML'd, started the raid, and zoned in first every week and getting complete new items and a more even spread of the total loot table. Has anyone else experienced a theory that seems like it works?
Please read practically the entire portion of the thread up to now.
About the only way this could work is if there were a currency dropped by bosses that could (a) be turned into the items over time -- a guarantee of achieving the items evenutally or (b) some way of using the currency to turn Item X into Item Y when you have enough of it. I don't really like any of these ideas even as a write them.
Sounds suspiciously like Heroic Badges off of Kara and ZA bosses. Now who would be crazy enough to do that? Oh wait...
Interesting read regarding percieved streaks written by Shermer, M. about a study conducted by Tversky, Amos in 1985 of 'hot streaks' in pro basketball.
In fact, what [the researchers] found is so counterintuitive that it is jarring: the number of streaks(successful baskets in sequence) did not exceed the predictions of a statistical coin-flip model. If you conduct a coin-flip experiment and record heads or tails, you will shortly encounter streaks. On average and in the long run, you will flip five heads or tails in a row once every 32 sequences of five tosses. Because Tverky was dealing with pro basketball players, however, adjustments in the formula were made to account for ability. If a player's shooting percentage is 60 percent, for example, chance dictactes that he will sink six baskets in a row once every 20 sequences of six shots attempted. When average shooting percentage was controlled for, Tversky found that there were no shooting sequences beyond what was indicated by chance. Players might feel "hot" or "in flow" when they have games that fall into the high range of chance, but science shows that nothing happens beyond what probability says should happen.
So if even the guys making millions buy in to the belief that they are lucky or not lucky, it's no surprise gamers experience the same feelings. Basic need to recognize pattern is at the heart of it.
For any true sports fans out there, a similar study was done regarding Joe DiMaggio's 56 game hitting streak and it was so many standard deviations away from the mean that, as physicist Ed Purcell put, it "should not have happened at all." The more you know.
That doesn't look like official proof of anything, that looks like someone on the official forums trying to stir up a discussion. Futhermore, it devolved into a European existential crisis and after the last one in '83 I've made it a habit to stay far away from European existential crises and/or meltdowns.
10/17/07, 8:10 AM
Similar Threads
