Random Number Generator Questions

[Draconi]

The next dev essay is on our random number generator. A lot of people have been asking questions about it, how things in the game work, etc.

The RNG is essentially our game's Dungeon Master, capable of casting the die of fate in favor of friend and foe alike. It's... less than perfect, and I'll be addressing that (and the corrective measures we're taking) when you next hear from me.

I won't be done with the essay for a while (just finished an SA milestone, moving into another), but this thread is for any and all questions about Ultima Online and our Random Number Generator. I'll actually answer questions where appropriate, so feel free to be specific!

Please don't hijack the discussion, and ask away!

 

[UOKaiser]

The next dev essay is on our random number generator. A lot of people have been asking questions about it, how things in the game work, etc.

The RNG is essentially our game's Dungeon Master, capable of casting the die of fate in favor of friend and foe alike. It's... less than perfect, and I'll be addressing that (and the corrective measures we're taking) when you next hear from me.

I won't be done with the essay for a while (just finished an SA milestone, moving into another), but this thread is for any and all questions about Ultima Online and our Random Number Generator. I'll actually answer questions where appropriate, so feel free to be specific!

Please don't hijack the discussion, and ask away!

Improvements in the RNG will be nice though first you will need to know where the faults are. Maybe it seems as the RNG is off balanced a bit moving towards the negative instead of staying balanced? Could be wrong and can be only perception just seems that way.
As long as you don't change the code to what Poxnora did to there game in order to allow money tournaments. By making numbers absolute and removing the RNG. I enjoy the randomness I don't enjoy mathematical calculations and chess like patterns.

 

[Foolio the Bard]

I always thought it would be sweet if the RNG's result was displayed (like watching a dice roll at your DnD game) when the RNG was used.

The screen would say something like: "Numbers 1 thru 100. You need a 90+ to get the artifact. You roll: 89"

Any chance of this ever happening in certain situations? Or am I looking at this too simplistically?

(Edit: you could make this something that one could toggle off or on, for those that like it or dislike it. Also, congrats on the SA milestone!)

 

[Maplestone]

It feels like RNG gets "stuck" when doing something repetitive and can be "unstuck" when you do something like moving around or changing up what you're doing for a moment.

Is this real? If so, is it RNG or have my reflexes become so robotic that I'm hitting anti-macro code?

Are there seeds tied to the tile a character stands on and/or skill rolled against?

Are the above two questions related to how the legendary 8x8 method worked?

 

[Uriah Heep]

It seems that at 110 magery, your chance to cast 8th circle should be good.
But I have fizzled often, quite often, 5-6 times in a row. So I was told, when you start fizzling something you think should work, move, or eat. This makes the odds roll start over. Any truth to this?
Will your fix to the RNG end the streakiness we experience?
For instance, going to spawn and soloing it up to third level, and getting only one SoT?
going back to the next spawn, and getting 3 scrolls in the first level?

It's all kinda odd to me

 

[Vexxed]

How far from Truely Random is the current generator? From what I've observed it seems to go on "streaks" of similar #'s more frequently than it should...

aka.. Failing a 50% roll 7x in a row or suceeding etc.

 

[Basara]

I've already piped in on the directly-relating-to-RNG issues on Facebook, so I have a few more things that are still tied to the RNG but not the RNG specifically. I'd really like the info, for inclusion in the Stratics FAQs and guides too, since those are my projects here.

While something that uses the RNG (as opposed to the RNG), questions many crafters would like answered are:

We know there's a 50%/50% Weapon / Armor division when rolling for smith BOD type, but what are the other checks, and what is the order of the checks?
What is the info for the Tailoring BODs, of this type? Are tailor BODs REALLy supposed to be 2/1 normal/Ex at 120 skill?

Is the 10% bonus to quality for GM Smith (announced when BODs were introduced - makes the "roll" 60% exceptional, instead of 50/50) still in effect, or was it removed with the 2008 (or 2003, even) BOD changes, resulting in apparent RNG anomalies. Was this ever applied to Tailor BODs, as inferred in the patch notes and Dev system descriptions from that era?

Lastly, any chance of getting the female plate, shield & helmet BODs moved over to the "weapon" side of the Smith decision tree? Doing so would reduce the glut of iron BODs, make the "colored/iron" check meaningful for that half of the decision tree, and substantially improve the Smith BOD selections by removing the nearly half (11/24) of the current "Armor" decision tree that are non-combinable with larges, in effect reducing this disparity between the Smith and Tailor high-end runics (though tailor barbed runics would STILL drop 10-100 times more often than a Valorite or verite hammer)

 

[Storm]

I have often wondered is there just one RNG generator or for example are there ones on each server (dungeon/malas/britain etc) the reason I ask is that it seems that I can be having terrible time gaining and I cross a server boundry and boom the gains come back?
also I can cast in one spot and fail after fail and move 1 space and boom the spells go off!

 

[Cear Dallben Dragon]

Questions:

What do you use to seed the RNG? Server time, complex algorithm, client side system time, player ID number of some kind, anything? ect.

What do you use in certian situations to set the min,max for the RNG to pick from?

Do dice and darts have anything to do with the RNG? If so whats the relation.

Do you plan to change anything about it? What and why.

 

[Lord Gareth]

For Fishing when it comes to getting an SoS, Fishing Net, Big fish or something else the chance is rolled but, when you are moving Slow north or Slow South you fish up 10x as more Sos, Nets etc then you do fishing out one spot.

Does the generator work better if you are moving or if you are standing still?

 

[Littleblue]

Does Axis Location have anything to do with the generation of the seed?

 

[Whinemaker]

Ok here's my question:

Does the RNG, as it is currently and from what you've found out, "favour" any one?

 

[GalenKnighthawke]

Are the streaks intended? If now, how do they occur?

Is the RNG "weighted" to reward those who already have what they are looking for? (By way of example: Since Age of Shadows, many have pointed out that once you get one Doom artifact, others can follow in short order. But it takes "forever" to get the first one.)

One whiny question: Why has the RNG always sucked for me?

*chuckles*

-Galen's player

 

[Olahorand]

Well, I got once a Doom artifact after 6 month, and after that never again.
Not with that char (archer), not with my tamer.
I gave up on that.
*Salute*
Olahorand

 

[Connor_Graham]

Will there be an anti-RNG (something that tells the RNG that something happening x number of times in a row is almost statistically impossible) considered to prevent unbelievable streaks?

 

[phantus]

I can't help but feel any RNG is going to be similar to another. They are always going to seem to be streaking. Short of creating something that is not random to accompany the current algorithm creating the numbers I have a hard time feeling that players are actually going to be able to tell the difference. A fixed random would be the only thing that would actually create the sense of random accomplishment that people are going to benefit from in a game setting.

Best example is the tokuno drops(I'd say gauntlet but it's just not as good in practice as it is in concept.) The chance for a drop was random but based on an accumulating percentage. While not a random it seems to be more fair for lack of a better term.

The other example is you are fighting something(be it player or monster) and you miss 11 times in a row. While entirely possible in a random setting this is something that would be perceived as a streak.

Do you intend to create some type of fixed random generation for UO? Short of that I'm not sure your efforts are going to bear fruit if you goal is to make it seem as if the pendulum is swinging both ways for purposes of gaming success/failure.

 

[uoBuoY]

Is there one RNG module on every server (dungeon/Malas/Britain etc) and every char/beast/etc. on that server uses it sequentially? Because if this is the case streaks couldn't technically occur unless you were the only person on that server. Also the seed would be very unimportant because of the totally random way in which the RNG would be accessed.

-or-

Does each char have their own RNG module? This seems impractical but it would account for streekiness. Seed selection would be critical.

 

[Draconi]

I always thought it would be sweet if the RNG's result was displayed (like watching a dice roll at your DnD game) when the RNG was used.

The screen would say something like: "Numbers 1 thru 100. You need a 90+ to get the artifact. You roll: 89"

Any chance of this ever happening in certain situations? Or am I looking at this too simplistically?

(Edit: you could make this something that one could toggle off or on, for those that like it or dislike it. Also, congrats on the SA milestone!)

Now this is an interesting idea. The main issue is really, the RNG is not only our dice-rolling DM, but also handles the "physics" of UO. Everything we take for granted like timing delays, regen rates, falling damage, or fatigue checks, essentially all the little things, would be pure information overload if exposed to the playerbase. Besides that, we really wouldn't *want* everyone to know everything about how the world worked, lest the Great Mystery be rent asunder

Now, that being said, I'm tempted to still add an item, much like the Skill Tutor Statuette, that could tell you your "rolls" in *certain* situations where it doesn't complete destroy the sense of immersion. 11th year vet reward?

 

[Draconi]

It feels like RNG gets "stuck" when doing something repetitive and can be "unstuck" when you do something like moving around or changing up what you're doing for a moment.

Is this real? If so, is it RNG or have my reflexes become so robotic that I'm hitting anti-macro code?

Yes, this is real and a reproduceable issue. The current implementation for the RNG hasn't changed in years, and is no different from what a Comp Sci. student learns about the rand() function in beginning C++ courses.

Now, given, you're no longer dealing with the anti-macro code. MrTact removed that quite some time ago. Unfortunately, the way we generate random numbers is still prone to cycling numbers (repetitive series), and we're planning on addressing it.

 

[Draconi]

It seems that at 110 magery, your chance to cast 8th circle should be good.
But I have fizzled often, quite often, 5-6 times in a row. So I was told, when you start fizzling something you think should work, move, or eat. This makes the odds roll start over. Any truth to this?

Yes, this can really happen, and multiple players doing different things can break it out of its funk. This seems counterintuitive, but it's a consequence of the naive way the original RNG was being used.

 

[Draconi]

How far from Truely Random is the current generator? From what I've observed it seems to go on "streaks" of similar #'s more frequently than it should...

aka.. Failing a 50% roll 7x in a row or suceeding etc.

Well, it's a pretty standard implementation, so I'd imagine "not very." The problem isn't the function itself though, just the way the code to use it was written.

See: Diehard Tests

 

[Storm]

Now this is an interesting idea. The main issue is really, the RNG is not only our dice-rolling DM, but also handles the "physics" of UO. Everything we take for granted like timing delays, regen rates, falling damage, or fatigue checks, essentially all the little things, would be pure information overload if exposed to the playerbase. Besides that, we really wouldn't *want* everyone to know everything about how the world worked, lest the Great Mystery be rent asunder

Now, that being said, I'm tempted to still add an item, much like the Skill Tutor Statuette, that could tell you your "rolls" in *certain* situations where it doesn't complete destroy the sense of immersion. 11th year vet reward?

That would be freakin cool except dont make it like the quest to get skill tutor quest !(its broken does not work right)

 

[Jhym]

1) is it doing as I suspect -- pulling a random 32 bit integer, then restricting it down to the roll parameters, or is it doing a bitwise pull around the die that is to be rolled? If the former, then I can see why there'd be major streaks: in a d100 roll, 323959235 is the same as 323957748 -- but the original rng value is diverse.
2) would it make sense to have an external random 32 or 64 bit generator, attached to the server, that a process would then constantly poll for a RNG queue?
3) Hanse specifically had the Bod routine pull something like 100K bods to determine whether the color distribution was working, and he stated that he was pretty close to expected distributions. In light of that, and what you have found, was his testing method flawed or has something changed since that time?
4) How would you have QA do the final testing of this -- since it is only useful in the context of actual activity, they cannot build a test plan that would ever actually show you succeeded in "fixing" it. Whether it is random or not, you will not see exact, reproducible results on even testing servers.
5) Have you actually pulled live comparison numbers from a live server, matched to the "check" they are against, and THEN determined how random they are? Or are you just pulling the numbers from the generator directly, before you actually process them for decision making?
6) Is each PROCESS using its own seed and own rng procedure, or are the processes pulling the rng from a single process? If the former, isn't that taxing on the server? If the latter, why not set up the rng call so that it actually polls a group of rng generators, pulling from a "random" one?
7) We know we have been told that doing "things" doesn't actually modify your rolls (like eating, skipping servers, etc) -- if so, why WASN'T this implemented. Eating food gives a 50/50 chance it will reset the RNG with a new seed. Or skipping to a new subserver switches which bits are pulled to do the rolls. Or emoting *dances for luck* adds 1 to the next random rolls.
8) You DO realize that no matter whether you "fix" it or not, the exercise will be moot -- players will always feel like it's not random enough or way too random.

 

[Vyal]

DRACONI, I can not for the life of me figure out why you are saying that making a random number is hard.

This is very very very basic math. Let me give you a link,
www.pscode.com search for random number and you will have a list of thousands of programs to do just what you want to do.

Jhym 4 da luv of god man! This is not as complicated as you are making it out to be ROFL ROFL ROFL.
This line cracked me up the most out of any thread ever on Stratics.

"1) is it doing as I suspect -- pulling a random 32 bit integer, then restricting it down to the roll parameters, or is it doing a bitwise pull around the die that is to be rolled? If the former, then I can see why there'd be major streaks: in a d100 roll, 323959235 is the same as 323957748 -- but the original rng value is diverse."

You understand that EA has made ALOT of 3d games and 3d collision detection is about 700 thousand times more complicated then a freaking rng ROFL give me a break man. I can't wait to see your essay Draconi ROFL.

 

[Cloak&Dagger]

DRACONI, I can not for the life of me figure out why you are saying that making a random number is hard.

This is very very very basic math. Let me give you a link,
www.pscode.com search for random number and you will have a list of thousands of programs to do just what you want to do.

Jhym 4 da luv of god man! This is not as complicated as you are making it out to be ROFL ROFL ROFL.
This line cracked me up the most out of any thread ever on Stratics.

"1) is it doing as I suspect -- pulling a random 32 bit integer, then restricting it down to the roll parameters, or is it doing a bitwise pull around the die that is to be rolled? If the former, then I can see why there'd be major streaks: in a d100 roll, 323959235 is the same as 323957748 -- but the original rng value is diverse."

You understand that EA has made ALOT of 3d games and 3d collision detection is about 700 thousand times more complicated then a freaking rng ROFL give me a break man. I can't wait to see your essay Draconi ROFL.

Where does he say that its hard to do? That is my first question, I think he explains what is wrong with the current RNG as it is a basic random int that any basic c++ student would learn making it a VERY SIMPLE mathematical problem as far as math and coding go. The RNG is more than just making a random number, you obviously have very little idea how to properly implement a RNG in a game as vast as UO, a simple program that can make a random number is not what it is about, the current RNG is OUT DATED as he stated, thus it is less "random" than it should be altho you can never have true random. The guy you quoted with the question if the RNG is pulling a 32 bit int has nothing to do with EA's ability to creat 3d games, he asked if that is how the current RNG works, if you read it it is a question and makes perfect sense. If you can only hit a number from 1-100 and those numbers are represented by a 9 digit number that means all numbers with in a certain range will be = to 1 and all numbers in another range will be = 50 so why is the question so funny? he wants to know what method they are using for the RNG and simply it is most likely that method since Draconi said it was the simple rand function in c++. And PLEASE if you are going to criticize the knowledge of other people, learn how to write.

 

[Ferrut]

I finally get a 120 magery scroll to avoid those damned 10 in a row EV fizzles, and they are gonna fix it! Hehe.

Anyway, joking aside, I was wondering about the Weather effects in the ('2D'/Legacy) client...

I'm assuming they are a client-side RNG, but I was just wondering if the 'streaky-ness' (all the rain/snow spawns from the same spot) in that system had any bearing on this discussion!? (I think it has been put forward as an example in the past)

 

[popps]

Now, that being said, I'm tempted to still add an item, much like the Skill Tutor Statuette, that could tell you your "rolls" in *certain* situations where it doesn't complete destroy the sense of immersion. 11th year vet reward?

I think this is a very very good idea and a significative appreciation of a customer loyal to the game for so long.

I think it should be redeemable AND only usable by 11 years accounts, though.

 

[Vyal]

vast?
http://www.youtube.com/watch?v=xbFCI_ZWLCk
Is a 2d engine in vb that has more to it then the engine UO uses. Do not say UO is vast rofl.

 

[Bombastic Fail]

My biggest RNG pet peeve (maybe some others can back me up on this). When you have 99% lrc (or 95%+) and you seem to get the "MORE REAGANTS NEEDED" way more often then 1% of the time. (Try about 5-15% of the time)

 

[Old Man of UO]

... Now, that being said, I'm tempted to still add an item, much like the Skill Tutor Statuette, that could tell you your "rolls" in *certain* situations where it doesn't complete destroy the sense of immersion. 11th year vet reward?

THIS is the part that I worry about, loss of the sense of immersion. A bland adoption of this where everything you do gets a message of the die roll is too much information, distracting, and looses the feel of being part of the game. Don't take all the mystery out of UO. Limited implementation would work, just not on every battle or corpse loot. YECH!

 

[Tina Small]

Why do you often get a skill gain in conjunction with the last use of a tool?

Why does it often feel like the server is ignoring your existence for the first 20-30 minutes of log-in for purposes of doing things like spawning a rare gem, bumping up loot values for luck on your suit, or giving gains in skills that are relatively high (80+)? And then when you do get a gain or a gem, why do you get two or sometimes even three in quick succession? Or you mine for an hour without getting any gems on one day and then the next time you log in, what makes you get two-four gems within the first five minutes of starting to mine again? It almost feels like there's a subroutine keeping track of what you should have gotten but didn't get because of the "streaky" RNG and it takes effect every time you log in but only "rewards" you if your "luck" was exceptionally bad on the last log-in.

Does the RNG factor in the server's current population for things like luck or skill gain?

Why do some old characters seem to have no luck getting arties in comparison to freshly made characters? Have you added a switch or a value or total to the character profile that is difficult to reset on an older character?

Since last summer's events in Moonglow, it seems that a lot of the more challenging monsters (e.g., rune beetles) have been dropping some types of loot in triplicate. It isn't unusual to open a corpse now and find three rings, three bracelets, three of the same type of weapon, or three of the same type of weapon. Is this change a result of a change in loot tables or something weird going on with the RNG?

 

[hawkeye_pike]

I would reconsider disclosing the RNG status to players. I think UO is more exciting if there still remains some random mystery that cannot be broken down to mathematic formulas. This may sound idealistic, but people should go hunting for the fun of it, and not in order to reach the last 10% probability they need to achieve an artifact.

In fact, the whole gaming world, including monster spawn, rewards, ressources, secrets, should even be more random and dynamic. UO has become a pretty predictable game. You can exactly tell where you have to go to get a certain item, you can predict every monster's AI. This takes adventure and challenge out of the game.

 

[DevilsOwn]

I would reconsider disclosing the RNG status to players. I think UO is more exciting if there still remains some random mystery that cannot be broken down to mathematic formulas. This may sound idealistic, but people should go hunting for the fun of it, and not in order to reach the last 10% probability they need to achieve an artifact.

In fact, the whole gaming world, including monster spawn, rewards, ressources, secrets, should even be more random and dynamic. UO has become a pretty predictable game. You can exactly tell where you have to go to get a certain item, you can predict every monster's AI. This takes adventure and challenge out of the game.

agree.... knowing how it works would be enough for me (not that I'd understand it, mind ya!) and I really wouldn't need to see it ingame in any form

 

[Barry Gibb]

We know there's a 50%/50% Weapon / Armor division when rolling for smith BOD type, but what are the other checks, and what is the order of the checks?

Basara - The order of the checks ultimately does not matter when determining the type of BOD you receive. The probability of one determination should not influence the others. For example if the probability of weapon or armor type is 50%/50% (P-type = 0.5) and the probability of normal or exceptional quality is 50%/50% (P-qual = 0.5), you will still have a 25% chance to get an exceptional armor deed (P-type x P-qual = 25%; 0.5 x 0.5 = 0.25). This holds true if the equation looks like this P-qual x P-type = 25%. Consider adding quantity (10/15/20; P-qty) and assuming they are all equal (33%/33%/33%). Now to get a 20 exceptional armor you have
P-type x P-qual x P-qty = 8.25% (0.5 x 0.5 x 0.33 = 0.0825). The result is the same if the equation is written like this (to rearange the order of determination): P-qty x P-qual x P-type = 8.25%. If all materials are equally probable (P-mat = 11% = 0.11), then to get an 20 exceptional valorite armor the probability is 0.9075%. I have not included the armor piece type or any of the elder and legendary skill bonuses, so my examples would be clearer. From your FAQs I gather you have a strong mathematical background, so I hope this helps others that don't.

What we would like to know is (for both smithing and tailoring):
"Are any of the determinations (type, quality, quantity, material, armor piece) are dependant on another?"
I expect and hope is not true.

Lastly, any chance of getting the female plate, shield & helmet BODs moved over to the "weapon" side of the Smith decision tree? Doing so would reduce the glut of iron BODs, make the "colored/iron" check meaningful for that half of the decision tree, and substantially improve the Smith BOD selections by removing the nearly half (11/24) of the current "Armor" decision tree that are non-combinable with larges, in effect reducing this disparity between the Smith and Tailor high-end runics (though tailor barbed runics would STILL drop 10-100 times more often than a Valorite or verite hammer)

Though this will improve the qaulity of armor BODs obtained, by eliminating the female plate, shield & helmet from the armor side. They will polute the weapon side of the smith BODs and potentially hurt the non-vetran smiths. I would rather see an them eliminated from the BOD list (as long as the leave the plate helm and chanimail coif ). This would bring the total # of smith BODs significantly closer to the total tailoring BODs, and still retain the rarity of the higher end runic hammers.

Are there any specific things we can do as players to limit our exposure to the streaks (other than our current superstitions)? At least in the long run, there are good streaks too. We just remeber the bad ones.

Stayin Alive,

BG

 

[Connor_Graham]

Basara - The order of the checks ultimately does not matter when determining the type of BOD you receive. The probability of one determination should not influence the others. For example if the probability of weapon or armor type is 50%/50% (P-type = 0.5) and the probability of normal or exceptional quality is 50%/50% (P-qual = 0.5), you will still have a 25% chance to get an exceptional armor deed (P-type x P-qual = 25%; 0.5 x 0.5 = 0.25). This holds true if the equation looks like this P-qual x P-type = 25%. Consider adding quantity (10/15/20; P-qty) and assuming they are all equal (33%/33%/33%). Now to get a 20 exceptional armor you have
P-type x P-qual x P-qty = 8.25% (0.5 x 0.5 x 0.33 = 0.0825). The result is the same if the equation is written like this (to rearange the order of determination): P-qty x P-qual x P-type = 8.25%. If all materials are equally probable (P-mat = 11% = 0.11), then to get an 20 exceptional valorite armor the probability is 0.9075%. I have not included the armor piece type or any of the elder and legendary skill bonuses, so my examples would be clearer. From your FAQs I gather you have a strong mathematical background, so I hope this helps others that don't.

Am I the only one who's eyes glaze over when seeing paragraphs like this?

I hate math in UO....

 

[Barry Gibb]

I would reconsider disclosing the RNG status to players. I think UO is more exciting if there still remains some random mystery that cannot be broken down to mathematic formulas. This may sound idealistic, but people should go hunting for the fun of it, and not in order to reach the last 10% probability they need to achieve an artifact.

agree....

I also agree that having this ingame would take the "game" out of the game and make it feel more like a "chore". I would be perfectly content hearing from the Devs:
"based on the feedback of the player base, we investigated their concerns about the RNG and have fixed the flaws we found. You should now enjoy and despise its results in a fair manner".
We (the player base) have a general understanding that there is randomness in UO. Please leave it at that.

Stayin Alive,

BG

 

[Barry Gibb]

Am I the only one who's eyes glaze over when seeing paragraphs like this?

I hate math in UO....

You cannot Kal Ort Por away from math.

Stayin Alive,

BG

 

[Connor_Graham]

You cannot Kal Ort Por away from math.

Stayin Alive,

BG

No, but you can sure scroll the mouse wheel to the next post, or hit the back button pretty damn easy.

 

[Barry Gibb]

No, but you can sure scroll the mouse wheel to the next post, or hit the back button pretty damn easy.

It's your choice to not read the whole discussion.

I will summarize my post:
Basara asked about the order of the checks used to determine BODs. I explained, with examples, why that information is unnecessary and that we only need to know if they are truely independant of one another.

I provided my opinion on Basara's suggestion of a BOD change.

I asked a question regarding RNG streakiness.

Stayin Alive,

BG

 

[Rykus]

I don't understand how anyone in this thread, developers included, could possibly be concerned about immersion at this point. We lost a TREMENDOUS amout of immersion the moment AoS was patched onto this once beautiful game. Seriously, having a floating d20 somewhere on the screen will do much less immersion damage than the overly complicated resists, and attributes we currently have. Has noone here even seen DDO? There is a d20 floating on the screen, which is movable and optional, that shows the results of each action. It's actually very cool, and many feel that it is a wonderful homage to the pen and paper origins of the RPG.

I also fail to understand how something like this should be the exclusive privilege of those who own an 11 year account. Not that it matters to me so much, mine is there in a few more months, but that seems like an odd thing to restrict in such a manner.

$.02

 

[Draconi]

I have often wondered is there just one RNG generator or for example are there ones on each server (dungeon/malas/britain etc) the reason I ask is that it seems that I can be having terrible time gaining and I cross a server boundry and boom the gains come back?
also I can cast in one spot and fail after fail and move 1 space and boom the spells go off!

There is one instance of the RNG per subserver. Each UO shard is divided up into many smaller subservers which each handle a portion of the seamless world (much like zones in other games, but with better connection tech). Each subserver is completely independent from the others, and they coordinate with a central server for the shard.

So, yes, it's quite possible to leave one area and have a completely different experience with the RNG.

 

[Draconi]

Questions:

What do you use to seed the RNG? Server time, complex algorithm, client side system time, player ID number of some kind, anything? ect.

What do you use in certian situations to set the min,max for the RNG to pick from?

Do dice and darts have anything to do with the RNG? If so whats the relation.

Do you plan to change anything about it? What and why.

Ah, I can't get too detailed, but it's nothing complex.

 

[Draconi]

For Fishing when it comes to getting an SoS, Fishing Net, Big fish or something else the chance is rolled but, when you are moving Slow north or Slow South you fish up 10x as more Sos, Nets etc then you do fishing out one spot.

Does the generator work better if you are moving or if you are standing still?

While moving, certainly. This is also related to how many resources are available per "chunk" of the world.

 

[Draconi]

Ok here's my question:

Does the RNG, as it is currently and from what you've found out, "favour" any one?

Nope, it pwns all of us equally it seems, even the developers.

But seriously, yes, there are situations that the design team *counts* on being decently random, but aren't getting it. For example:

• Coin Flips: Trends towards Heads more than Tails over the long run
• Dividing portions: Tends to give more portions to even numbers than odds
• 2d6: What should be a nice pyramid like distribution is whacky and malformed

In the greater scheme of things, this affects us all equally, but is still bad.

 

[Draconi]

Are the streaks intended? If now, how do they occur?

Nope.

Is the RNG "weighted" to reward those who already have what they are looking for? (By way of example: Since Age of Shadows, many have pointed out that once you get one Doom artifact, others can follow in short order. But it takes "forever" to get the first one.)

In most places it's not, but let's not confuse the RNG itself with the Reward systems we build around it. Doom is on the same point system now as the Treasures of Tokuno. This is more like a pack of cards where you need to choose the right one, but every time you choose wrong, the bad card is removed from the deck, making your chances better the next time.

There are many such solutions that better reward players than just relying purely on dice rolls, and while we're going to fix the core RNG problems, we're keeping these things in mind.

One whiny question: Why has the RNG always sucked for me?

Wouldn't it be awesome if Luck actually affected your RNG rolls?

 

[Draconi]

That would be freakin cool except dont make it like the quest to get skill tutor quest !(its broken does not work right)

Yeah, I really am liking the idea of an 11th year veteran reward only usable by 11th year vets.

It would be a powerful artifact that only the wisest and eldest could use to pierce the very fabric of reality! But not too much of the fabric, no, just a peek.

 

[Draconi]

1) is it doing as I suspect -- pulling a random 32 bit integer, then restricting it down to the roll parameters, or is it doing a bitwise pull around the die that is to be rolled? If the former, then I can see why there'd be major streaks: in a d100 roll, 323959235 is the same as 323957748 -- but the original rng value is diverse.

Yessiree.

2) would it make sense to have an external random 32 or 64 bit generator, attached to the server, that a process would then constantly poll for a RNG queue?

It's an interesting idea, but doesn't fit the current UO architecture - We can solve many problems right now by leaving it where it is and fixing the wrapper code around it.

3) Hanse specifically had the Bod routine pull something like 100K bods to determine whether the color distribution was working, and he stated that he was pretty close to expected distributions. In light of that, and what you have found, was his testing method flawed or has something changed since that time?

I couldn't know whether it was flawed or not, but we also started running a lot of tests like that. Especially to find out "how bad it was..." The BOD systems' chances were astronomically *not* in the players favor for rewards, so I can't imagine color distributions are much better.

4) How would you have QA do the final testing of this -- since it is only useful in the context of actual activity, they cannot build a test plan that would ever actually show you succeeded in "fixing" it. Whether it is random or not, you will not see exact, reproducible results on even testing servers.

Good question. We can do empirical tests ourselves and *show* the difference in distributions, streakiness, etc. But the true test must come from the players themselves.

5) Have you actually pulled live comparison numbers from a live server, matched to the "check" they are against, and THEN determined how random they are? Or are you just pulling the numbers from the generator directly, before you actually process them for decision making?

We've done both, and both ways show the same problems.

6) Is each PROCESS using its own seed and own rng procedure, or are the processes pulling the rng from a single process? If the former, isn't that taxing on the server? If the latter, why not set up the rng call so that it actually polls a group of rng generators, pulling from a "random" one?

Former, and it's not taxing at all.The RNG takes up very little CPU time compared to everything else going on. Although, I will mention that we recently found a single line of bad code that was doing something very very bad in its attempt to use the RNG, and I'll give more details later.

7) We know we have been told that doing "things" doesn't actually modify your rolls (like eating, skipping servers, etc) -- if so, why WASN'T this implemented. Eating food gives a 50/50 chance it will reset the RNG with a new seed. Or skipping to a new subserver switches which bits are pulled to do the rolls. Or emoting *dances for luck* adds 1 to the next random rolls.

Hmm, I disagree with that previous assertion. Anything you do that calls the RNG can affect future rolls.

8) You DO realize that no matter whether you "fix" it or not, the exercise will be moot -- players will always feel like it's not random enough or way too random.

We'll see!

 

[Draconi]

DRACONI, I can not for the life of me figure out why you are saying that making a random number is hard.

Oh, I never did.

Although I *am* saying that 12 years ago some poor soul decided to use a pretty naive method that doesn't work very well.

 

[Maplestone]

Wouldn't it be awesome if Luck actually affected your RNG rolls?

um ... not really, no.

If I have X% chance of success, I'd sort of like to have X% chance of success.

Mind you, you could simply take away or everything that tells the chance of success (after all, even though a lot of craftables tell you the chance of success, they don't don't actually tell you what it is you're about to make) and invalidate all the carefully accumulated tables players have made over the years.

As geeky about math as I can get sometimes, I too have a limit on the amount of statistical analysis that I find enjoyable to figure out if I'm wasting my time attempting something or not. I remember back when I started and I was picking up items left and right, I simply could not figure out if one weapon was better than another. I'd craft stuff, I'd loot stuff off ettins, I'd pick stuff off the ground ... I'd take things out and hunt with them for a while and come back feeling frustrated because without sitting down and writing out the numbers and success/fail, I simply could not tell the difference between good weapons and a string of good luck by just using them.

 

[Uriah Heep]

Okay for the computer-literacy impaired of us (or dummies like me)

Go back to fishing, and give an example.
Moving slow north, gives more reward than sitting still. Is this in fact because each time the boat hits a new tile, the RNG resets?
Or is it because there are only at any given time so many rewards spawned per subserver, and the mere act of moving takes you to them? Like finding the T-chest...its there , you just have to get to it.
Not sure if that's clear, but best way I can think of at the moment to word it...