Doom drops turn off during the event?

[Herman]

The problem with probability is that when looking at a single event everything looks like a rare event it is only when comparing that event to every other possible events you see if it is a rare or common

Example you flip a coin 3 times and get 2 tails and 1 head

There is 37,5% chance you get that combination

37,5% chance 2 heads one tail
12,5% all heads and 12,5 all tails

 

[Herman]

Yes, please explain how your coin should find out that it had 40 times heads streak before, so it must give more tails later.

For what it is worth I think you understand probability very good

It is just as rare to get 5 tails 1 heads and then 34 tails in exactly that order as it is to get 40 tails in a row there is ony 1 chance of all possible combinations in both those examples
it is only when you allow the 1 heads to get any position in that combination you get a different results


so judging how rare an event was in hindsight everything will look damn rare it is just an endless row of damn near impossible to get events

it is as Basara and stinky pete said you need to check for patterns if some combinations appear far too often then it could be a problem with the RNG does not have to be but it could be

 

[skett]

I thought they (the devs) had a system in place to guarantee drops after so many rolls

i realize this is not the topic but wouldn’t that fix a lot off the rng problems that so many give up
Games supposed to fun and enjoyable not a complicated ****ed up grind

 

[gwen]

I thought they (the devs) had a system in place to guarantee drops after so many rolls

i realize this is not the topic but wouldn’t that fix a lot off the rng problems that so many give up
Games supposed to fun and enjoyable not a complicated ****ed up grind

There are different types of content. Some have guaranteed drop over time (doom gauntlet) , have turn-in artifacts where you can pick reward you need (TOT or Blackthorn) , and content where you just grind with no guarantee you will get what you need (Cora, Medusa or Navery)
You just choose what you like.

 

[mihali]

There are different types of content. Some have guaranteed drop over time (doom gauntlet) , have turn-in artifacts where you can pick reward you need (TOT or Blackthorn) , and content where you just grind with no guarantee you will get what you need (Cora, Medusa or Navery)
You just choose what you like.

What is TOT content, and is it still on, and how do you do it - any information on that?

 

[Anon McDougle]

What is TOT content, and is it still on, and how do you do it - any information on that?

Tot or treasures of is the old new system it's currently not running but we should get a new version after the first of the year

 

[The Zog historian]

For what it is worth I think you understand probability very good

It is just as rare to get 5 tails 1 heads and then 34 tails in exactly that order as it is to get 40 tails in a row there is ony 1 chance of all possible combinations in both those examples
it is only when you allow the 1 heads to get any position in that combination you get a different results

so judging how rare an event was in hindsight everything will look damn rare it is just an endless row of damn near impossible to get events

it is as Basara and stinky pete said you need to check for patterns if some combinations appear far too often then it could be a problem with the RNG does not have to be but it could be

Actually, as I pointed out multiple times, she has no clue when it comes to probability. Her spreadsheet is one example. And if you think she "understand" it, well...

There shouldn't be anything "rare" about an event that is supposed to have a significant chance on any roll of the dice (e.g. 12.5% chance of each Yukio quest reward), or an event that despite having a minimal chance will see many rolls of the dice (e.g. killing hundreds of small spawn in Tokuno or a dungeon). No one has yet to answer my simple question satisfactorily: if the UO RNG is working properly, then why were guaranteed drops put in?

Your example is correct only if you're talking about a desired set of results. If we're to look at any results overall, 40 tails in a row has half the probability of 5T/1H/34T. But I'm not talking about a desired set of results, i.e a precise percentage over 100 or 1000 or 1 million rolls of the dice. I'm talking about certain parts of the game having clear streakiness, typically with clear deficiencies in dungeon drops and other things.

Dungeon drops, Axem's quest, imbuing, and especially refinements. Players have talked for years about the RNG not seeming to work like it should, and those four parts of the game are clear demonstration.

 

[The Zog historian]

Well that was in an intersting read so far. I think you guys are arguing a bit different cases though.

I'm arguing basic probability, which they don't understand. Roll two dice a couple of times, and you might get the same result twice. But do it ten times, and the same result a majority of the time should make you wonder about the randomness.

I think it's unquestioned that the UO RNG is streaky a lot. Streakiness in those old games often comes from using those bad LCGs, as @Basara seems to mention. They have been covered ad nauseam. To me this always looked pretty plausible.

As I keep posting:

OTOH does that alone explain what we often perceive (well - perceived lol) as "bad luck"? I bet there is another layer of abstraction build on top the basic RNG simply for having an convinient "UO style rolls" API. So there would be plenty of room for engineers that didn't understand the RNG in the first place to introduce even more side effects.

It isn't impossible to roll snake eyes a few times in a row. It isn't impossible to roll it 10 times in a row, but that's so highly improbable that you have to question the dice (like in the movie of the same name I finally got around to watching).

If I go to Axem and often need 10 tries (once 20) to get the particular quest I want, something's very wrong. When the devs put in a guaranteed drop algorithm with Tokuno, they were basically admitting that something was wrong with the RNG, which their apologists claim is just "bad luck."

The main takeaway for me is that they prenteded for 20 years to be unable to rework those like maybe couple hundreds lines of code. How helpless are they lmao.

Like with the database company I've mentioned, does BS have anyone who would understand where to go? If there is one person, is the person afraid to touch it, and they'd rather try to make new content?

 

[The Zog historian]

Yes, please explain how your coin should find out that it had 40 times heads streak before, so it must give more tails later.

I never said anything of the kind. You're still operating under the absurd presumption that I want exact percentages turning out.

 

[The Zog historian]

You might want to make clear why true randomness matters in this context?

This is not a mathematical discussion about stochastics, but rather about applied computer science.

Why shouldn't players want good randomness in a game they pay to play and invest time in? This isn't just a crappy app by a CS undergrad which can be uninstalled once it turns out to have bad algorithms.

When two dozen players are slaying a boss generating a couple dozen items per player, the server needs to make a couple hundred or a couple thousands rolls with the RNG only for this instance. Do you want the server to block for half a second on each kill because after an hour of fighting it eventually ran out of entropy?

Like the other guy fancying himself a computer/electronics expert, you don't seem to realize how fast computers are today. A couple of thousand rolls is nothing with modern CPU cycles. And if we're talking about making a new seed, then no biggie, it can be done at server up, or even once an hour.

Any true randomness is completely off the table, idk why you keep bringing it up.

Old 8-bit computers (6502 family) were able to do effectively true randomness. There's no excuse to not have it today.

You can keep talking in terms of that idealized model of a dice or coin used in stochastics. Yet the poor engineer implemting the model will have to resort to fairly less ideal means to make it happen in the real world.

I guess you found a new word you're trying to use every day, but UO's RNG problem is a lot simpler and needs no "model" (don't make gwen's mistake in thinking I want precisely distributed outcomes in the end), nor any analysis beyond simple probability measurements, so your last sentence here is bereft of anything relevant. I'll repeat myself: when computers 40 years ago could do effectively true randomness, even built into Atari BASIC, there's no excuse to not have it today.

 

[The Zog historian]

There are different types of content. Some have guaranteed drop over time (doom gauntlet) , have turn-in artifacts where you can pick reward you need (TOT or Blackthorn) , and content where you just grind with no guarantee you will get what you need (Cora, Medusa or Navery)
You just choose what you like.

When the guarantee has been removed, by all means, tell us how to "choose" here. You're sounding like the old Dreads who'd tell players they "choose" to get PKd by going to a dungeon or mining outside town. There was no good reason a player pre-Trammel needed to adopt a limited playstyle because of griefers, and no good reason today a player should give up on a certain monster spawn because the RNG is broken.

And one more time: the guaranteed drops are just the Devs' tacit admission that the RNG doesn't work properly.

 

[petemage]

Old 8-bit computers (6502 family) were able to do effectively true randomness. There's no excuse to not have it today.

And they would never run out of entropy? I don't have any benchmarks, but I doubt the chip could provide large amounts of random numbers over a short time without running into some sort of entropy issues. You would probably only pull true random numbers once in a while to feed something more performant and more determistic (an LCG for example) and thus inherit the same issues that come with it.

E: I like old computers and gave the 6500 family datasheet a look. The chip unsuprisingly doesn't feature dedicated TRNG circuitry (it's 1975 and this is not Intel lol). I know people did some stuff with e.g. sampling audio channels for entropy, but even this (as all TRNGs) is limited by physical means more than raw clock speed. Anyway, would be welcome if you could elaborate on that 6502-family thing a bit more.

Why shouldn't players want good randomness in a game they pay to play and invest time in? [...]

A couple of thousand rolls is nothing with modern CPU cycles. And if we're talking about making a new seed, then no biggie, it can be done at server up, or even once an hour.

Well, there you say it: "good randomness". As in "good enough". That's what even a bad LCG is about. A tradeoff between "good" and "true" that gives engineers less headaches. That's why PRNGs exist. Every engineer arguing that a game with a $10 monthly sub needs a true random number generator clearly needs some days off to get his **** together.

OTOH you proudly admitted to not even having looked at the improvements @Stinky Pete mentioned. But I guess that's the difference between historians and actual game developers. The former love their thought models of coin flips, while the latter realize there are no people flipping coins at AWS.

 

[Herman]

Actually, as I pointed out multiple times, she has no clue when it comes to probability. Her spreadsheet is one example. And if you think she "understand" it, well...

There shouldn't be anything "rare" about an event that is supposed to have a significant chance on any roll of the dice (e.g. 12.5% chance of each Yukio quest reward), or an event that despite having a minimal chance will see many rolls of the dice (e.g. killing hundreds of small spawn in Tokuno or a dungeon). No one has yet to answer my simple question satisfactorily: if the UO RNG is working properly, then why were guaranteed drops put in?

Your example is correct only if you're talking about a desired set of results. If we're to look at any results overall, 40 tails in a row has half the probability of 5T/1H/34T. But I'm not talking about a desired set of results, i.e a precise percentage over 100 or 1000 or 1 million rolls of the dice. I'm talking about certain parts of the game having clear streakiness, typically with clear deficiencies in dungeon drops and other things.

Dungeon drops, Axem's quest, imbuing, and especially refinements. Players have talked for years about the RNG not seeming to work like it should, and those four parts of the game are clear demonstration.

IMO quaranteed drops was put in because people in general do not like random they want to be able to see their goal or prize



You always have to confine your experiment in some way be it % of all the events or between 2 events or whatever
Probability is just that % of all events that can possible happen some are more common others are rare while it is rare to get a single rare event it is not so uncommon to get one of all the rare events
Because there is in most cases are so many possible combinations even the most common one will come in well under 50% so it is still more common to get any of the others than the most common one

so while it cannot explain everything about the UO RNG it can explain why people that feel unlucky blame the UO RNG
So this is from my own UO experience
There are many times in this game i felt cursed
Lets say there is a 1/20 dropp and it took 50 tries for me to get the artie that is very unlucky
But i would say the same thing that I was unlucky if i got it in 49,48,47.........30 so there is actually not that uncommon that i will hit 1 of those
This is why it is so bad to judge the rng on only 1 event if the rules say it can happen at some point it will

And while this has very little correaltion with your post it shows how silly things can become if you do not confine your experiment
EX you will hit an infinite numbers of heads in a row an infinite number of times if there is an infinite number of attempts

 

[Tjalle]

What is TOT content, and is it still on, and how do you do it - any information on that?

Treasures of Tokuno - UOGuide, the Ultima Online Encyclopedia

www.uoguide.com

Treasures of Tokuno is what ToT references to. Some of the new dungeon events use this type of system.

 

[petemage]

I see there is a discussion about probabilities and stochastics in general going on. I enjoy following that but I would like to talk about a different angle for once:

So UO is hosted on AWS, much likely in some sort of Linux VM. So how do you get random numbers in Linux in 2021? Well, Linux offers two random number generators: A pseudo-RNG and a true-RNG. So far so good. All the higher layers have to go there through various abstractions. Let me quote the official documentation on it:

The character special files /dev/random and /dev/urandom [...] provide an interface to the kernel's random
number generator.

The random number generator gathers environmental noise from
device drivers and other sources into an entropy pool. The
generator also keeps an estimate of the number of bits of noise
in the entropy pool. From this entropy pool, random numbers are
created.

When read, the /dev/urandom device returns random bytes using a
pseudorandom number generator seeded from the entropy pool.
Reads from this device do not block (i.e., the CPU is not
yielded), but can incur an appreciable delay when requesting
large amounts of data.

The /dev/random device is a legacy interface which dates back to
a time where the cryptographic primitives used in the
implementation of /dev/urandom were not widely trusted. It will
return random bytes only within the estimated number of bits of
fresh noise in the entropy pool, blocking if necessary.
/dev/random is suitable for applications that need high quality
randomness, and can afford indeterminate delays.

So let's ask ouselfs: Does Ultima Online need high quality randomness and can afford indeterminate delays over it? Do you want the server to block in the middle of the night needing poor @Bleak to get out of bed for it? Because god knows why this poor AWS VM ran low on entropy?

Or maybe you can email those beginner engineers at IBM and google that they could just buy some 1975 6502 chips to solve all their entropy issues @The Zog historian *scnr* I'm sorry if it sounds a bit harsh, but as much as your theoretical understanding of stochastics goes, as little do you seem to be in touch with how it's practically done and it's limitations up to this very day.

I don't even have to go into bad players weaponizing access to the RNG for denial of service (i.e. crashing the server). I tried reading 100 integers from /dev/random on one of my servers and entropy dropped straight to zero, even blocking the loop for the last rolls. It's so easy to run a system out of entropy given you have unlimited access to the TRNG. What you are arguing for is straight harmful.

I certainly see why engineers don't think it's needed.

E: Ah well those numbers are too funny too not throw them into the mix. 2021 Linux server. Reading 100 integers (i.e. 3200 coinflips) from true- vs. pseudo-RNG:

$ time for i in {1..100}; do dd status=none if=/dev/random bs=4 count=1 of=/dev/null; done
real    0m29,314s

$ time for i in {1..100}; do dd status=none if=/dev/urandom bs=4 count=1 of=/dev/null; done
real    0m0,184s (<- even with depleted entroy pool lol)

 

[petemage]

EX you will hit an infinite numbers of heads in a row an infinite number of times if there is an infinite number of attempts

Afaik that would contradict the "Law of large numbers".

If you indeed go to infinity, the mean of your draws will converge towards the mean of it's distribution. I.e. unless the coin has two head sides (an fairly unfair distribution to begin with), it will eventually even out towards the expected value at some point. My old prof used to joke around this topic a lot, recalling the poor gamblers only seeing the implication of numbers evening out without paying the "large" part proper attention. "As long as they got all the time in the world", he joked.

You are right in a sense that you can go very far (large) with your example, but you as far as I think went one step too far

 

[Herman]

Afaik that would contradict the "Law of large numbers".

If you indeed go to infinity, the mean of your draws will converge towards the mean of it's distribution. I.e. unless the coin has two head sides (an fairly unfair distribution to begin with), it will eventually even out towards the expected value at some point. My old prof used to joke around this topic a lot, recalling the poor gamblers only seeing the implication of numbers evening out without paying the "large" part proper attention. "As long as they got all the time in the world", he joked.

You are right in a sense that you can go very far (large) with your example, but you as far as I think went one step too far

But do I realy break it in that last sentence or do I support it ?
Have to think about that for a moment (thx for the brain gymnastics combine great with a bottle of Gin)

The rest of that post you quoted for sure you could do that claim if you think the sample size is not big enough to support abnormalities but you probably already know where i go after that
You can not prove Randomness without a realy big sample
and round and round we go

I will not tell you the rest of that infinite joke but i can tell you how it ends
.
.
.
INFINITE X 10 YOU ARE IT!!!!!!!!

 

[gwen]

When the guarantee has been removed, by all means, tell us how to "choose" here. You're sounding like the old Dreads who'd tell players they "choose" to get PKd by going to a dungeon or mining outside town. There was no good reason a player pre-Trammel needed to adopt a limited playstyle because of griefers, and no good reason today a player should give up on a certain monster spawn because the RNG is broken.

And one more time: the guaranteed drops are just the Devs' tacit admission that the RNG doesn't work properly.

RNG works properly. Otherwise Krampush or UW boss will have same loot. It doesn't. But we are people, not computers. While it is ok for me to generate thousands of values in Excel, killing Navery 10 thousand times will be boring. Or sinking and killing 10-20thousand pirates.
Devs made guaranteed drops to give us what we want and to cause less anger. Finishing Yukio quest 30-40 times getting no earrings pissed some people off. While Medusa is still farmed by some players.

Some drops are counted more valuable and have less chance to be generated by system. Like chances to get expensive Slither (never got it) and useless Petrified Snake or Stone tooth dagger (got 10+ for now).
For Axem quest I believe it is not true random. Nobody ever said it should be. Quest with pottery fragments appear more often. And I never bring em, never collect em. Quest changes when you accept and finish one. So coming with pile of tattered scrolls and Tomes makes it fast. Coming with just untranslated Tomes bores me.


Happy New Year, wish you do only things that make you happy in this new nice year!

 

[gwen]

What is TOT content, and is it still on, and how do you do it - any information on that?

Some ToT explanation was given earlier by other users. Currently it is Blackthorn dungeon where you get minor artifacts killing Captains and can turn those Minax artifacts in to pick some reward of your choice. Removing random rewarding to some extent. Of course people who kill faster or have more luck or prepared better will get more. Good is you will not end with useless thing. (I ended up once picking wrong one, but it was my fault)
Last ones were Wildfire and Demonic Forces events in Fire and Hythloth dungeons. Hope we will get more of them in the future.

 

[Pawain]

I agree the basic roll to get a 1 in X chance with no weight works fine. Since we do not know the weighted chances then we can not tell. But we can tell the weights work on simple encounters. that we can test.

The last examples are the Anniv gifts and holiday gifts.

I have never opened a rare anniversary gift before. I usually stop at around 50 tries and keep the tokens.
This time I wanted a Harp with my char name on it. I opened hundreds and didn't get one. Then I decided to quickly open a bag of 15 and not look at the results. I got 2 with my name. Yay!
But I do not know what the odds for the rare ones are. So, that is not the point. To occupy my mind while clicking the tokens I decided to sort them by Jeweler. By doing that I found that no jeweler name was dominant or submissive.

So the 1 in X for a jeweler name worked fine.

The holiday gift paintings are another in in X. I collected avg 24 of them on each shard I chose to visit.

Again the city names were very spread. I saw no city that got more paintings than the rest.

For the weighted rare fire paintings. I did do enough to get a guess at the rarety. All the shards I did, I received two fire paintings except my home shard. I got 3 on one shard. So that seemed to be a 1 in 12 change for me.

Also I disagree with the Axem quest not being a 1 in X chance. I did that quest sooo many times years ago. I have all the first editions and most of the non plains and a ton of plain books.

I also remember the times it took me 4 tries to get the one I had on me. But I tend to forget the times I got the one I wanted first. That's is an easy one to prove right now. Just go click him and write down what quest he gives.

Percolem also seems to have bad RNG, if I want to kill raptors I get something else.

The earring quest was also fine as a 1 in x chance. Some got earrings on their first try some didnt until 30 tries. But, you have to do that one hundreds of times to see it works, and players on LS did just that. Not me, I did it 17 times got 2nd earring and quit. Bought 6 more from the ones that kept going.

If you see a Ter Mer book in this order, it was originally made by me.
I have seen it on many shards. I always make runes in different spots so more than 1 person can get to the same place.

 

[gwen]

@Pawain, nice book, I'll make same.

Hunter quests are some kind bad - you'll never get what you want to kill. There should be some logic there, not just random.
Alchemist quests are different: you finish 2 with slug and ant going and only get tasty treats quests for some time.

For heartwood quest (tinker) is better. I do plugs for talismans and recipe. 1/3 chance but sometimes it takes long streak of chimes and helms. Good way to do is get Chime quest from one weaver and another is left with only 2 quest choices. If there be 2 Axem quest givers, museum quest will be less boring. Or if Axem will just switch 1-2-3 quest after one is cancelled.

Rare deco should be rare. Like Arch demon heads are.

 

[skett]

What’s the earring quest ?
and rare deco quest ?

 

[Anon McDougle]

What’s the earring quest ?
and rare deco quest ?

Earrings was part of a treasures of story arc

 

[Pawain]

What’s the earring quest ?
and rare deco quest ?

UO has been very busy without you. We had many events this year that are gone now.

 

[Herman]

I see there is a discussion about probabilities and stochastics in general going on. I enjoy following that but I would like to talk about a different angle for once:


So let's ask ouselfs: Does Ultima Online need high quality randomness and can afford indeterminate delays over it? Do you want the server to block in the middle of the night needing poor @Bleak to get out of bed for it? Because god knows why this poor AWS VM ran low on entropy?

IMO no uo does not need it

All UO need is something that provide an element of luck (or illusion of luck) make the game less predictable
I think one could achieve that without a RNG at all
Unless I was actively trying to find a pattern I would probably not even notice the difference between random and 50 numbers in a loop


What i do think is that there is enough evidence in this thread that alot of UO players do not think that the way UO distribute rewards is FUN
If the goal was only to make content more fun
Making rewards more common probably would have much bigger impact than a perfect RNG

 

[The Zog historian]

RNG works properly. Otherwise Krampush or UW boss will have same loot. It doesn't.

What utter nonsense. You will never know what you're trying to talk about. I would explain things like streakiness, like improper coding leads to claimed percentages not working properly, and that you're confusing the bad randomness of different items for true randomness. But it's like good food in a communist society: you wouldn't get it.

In addition to the examples I've mentioned of imbuing, Axem's quest, and refinements, orc ships are another great example.

But we are people, not computers. While it is ok for me to generate thousands of values in Excel, killing Navery 10 thousand times will be boring. Or sinking and killing 10-20thousand pirates.
Devs made guaranteed drops to give us what we want and to cause less anger. Finishing Yukio quest 30-40 times getting no earrings pissed some people off. While Medusa is still farmed by some players.

I never said you should go kill thousands of monsters and collect the data. I, however, have offered my time to the UO team to show them certain examples.

Some drops are counted more valuable and have less chance to be generated by system. Like chances to get expensive Slither (never got it) and useless Petrified Snake or Stone tooth dagger (got 10+ for now).

Now you're muddying the issue by bringing up oranges when we're talking about apples. Those are intentionally weighted differently. However, when an imbuing or refinement menu states a certain percentage, what's your apologism for the false randomness?

For Axem quest I believe it is not true random. Nobody ever said it should be. Quest with pottery fragments appear more often. And I never bring em, never collect em. Quest changes when you accept and finish one. So coming with pile of tattered scrolls and Tomes makes it fast. Coming with just untranslated Tomes bores me.

That first part in bold is just your bad assumption for something that isn't working properly. That second part is just blabbering nonsense.

Axem's quest is the easy way to show "streakiness" in the pseudo RNG, but if you can't see that much, well, you'll probably never understand.

Happy New Year, wish you do only things that make you happy in this new nice year!

You mean like not playing UO? I did log in today to look for something in my houses, to give to a friend. I forgot to cancel two accounts.

Let me tell you what. To show us again how little you really know about statistics, here's a small quiz.

Take a standard lotto with, say, odds of winning as 1 in 200 million. What are the odds a lottery ticket of 1-2-3-4-5-6 will win, and what are the odds 2-3-4-5-6-7 will win? If someone wins with 1-2-3-4-5-6, what are the odds he will win with the same numbers the next week, and what are the odds he will win with the same numbers one year later? What are the odds he will win two weeks in a row with 1-2-3-4-5-6 both times, and 1-2-3-4-5-6 one week but 2-3-4-5-6-7 the next week?

Now to relate to UO: if I get pottery fragments from Axem, what are the odds I'll get the pottery fragments quest the next time? What are the odds I'll get pottery fragments three times in a row? What are the odds I won't get the tome quest 10 times in a row? If I get pottery fragments the first time, what are the odds of pottery fragments the next time versus getting the tome?

These are very simple statistical questions barely worthy of a first quiz in Probability 101.