Ultima real bad roll sheeting and possible Fixes.

[PenNPaper]

I have been playing for about a couple of weeks now and so far Ultima seems to have rally bad roll sheeting. Roll sheeting is when the random number generator generates a low number for many rolls. So you can end up with a 1,1,1,1,1 roll witch would possibly be 5 misses and that is a problem. This stems from the fact that computer random generator does not seem very random when you look at the rolls generated.

Back in 2004 I came out with a random number generator that addresses this. It still uses the rand function but has added features. The way this works is you can rand 2 rolls and compare them so if it comes out 1 1 it drops this and generates another set of numbers so say you 2 , 1 and the last roll was a 1,1 it would choose 2 as the roll. if it rolls 1,1 - 1,1 then it can pass a 1. The function works on the principle that even if it rands a bunch of 1s a lot less 1s will get passed to roll value.

Another possibility is to use two rand roll values so say 1.1 -1,2 so the 1.1 would be 1 and the 1.2 would be a 2 so you have the values of 1 and 2 then rand would either choose 1 or 2.

Generally speaking you do not want the same number to come out for short periods of time so it would still be possible to roll a few 1's but not a hole bunch of them

 

[petemage]

Feels like getting 10 times red in a row on roulette. What are the odds? Greater zero unfortunately

 

[Uriah Heep]

Yeah. I love having a 48% chance to imbue something and failing 15 times in a row...would have never thought that possible if I didn't play UO

 

[Prawn]

..................

 

[Basara]

It's had that problem for a long time - the problem is that it's not "Roll Sheeting" per se.

The issue is much deeper, in that the RNG is prone to streaking at ALL levels, not the bottom end. As a result, when they ran tests (OVER TWELVE YEARS AGO), with millions of "rolls", no streaking was apparent - "Roll Sheeting" would have shown streaking on the bottom end.

We FINALLY got Draconi to test for low-level streaking, and he was the one to discover that the streaks were happening at all points of the curve, not one end or the other. He started working on a replacement RNG in his spare time, but right before he was going to bring the RNG work in-house to start alpha-testing it (between KR and SA), EA went on one of their nonsensical "size-cutting" sprees (It always seems to be the branches least needing cuts that get cuts) and laid him off.

So, for almost 10 years, we've been stuck with an RNG that is KNOWN to have streaking issues. (It had streaking issues for over 20 years, now, but the actual proof has only been known 10 years)

 

[grimiz]

My favorite is trying to open a gate or res and fizzling 10 times in a row... then you move one space and it works the first time. It. Is. Infuriating.

 

[Uriah Heep]

My favorite is trying to open a gate or res and fizzling 10 times in a row... then you move one space and it works the first time. It. Is. Infuriating.

I had a counselor tell me one time long long ago, (yes I pages about that problem) and asked him why if we stepped a step adn tried it it would work. He said it made the RNG start all over? makes no sense to me, but whatever lol.
And then he added that is why eating would fix it, it makes it all reset

 

[Riyana]

I had a counselor tell me one time long long ago, (yes I pages about that problem) and asked him why if we stepped a step adn tried it it would work. He said it made the RNG start all over? makes no sense to me, but whatever lol.
And then he added that is why eating would fix it, it makes it all reset

That's interesting. While I know it's been hammered in that it doesn't matter, I've often found eating to help when I have a bad casting streak.

I actually created two identical characters on test center once to compare fullness and casting--I had one character that was full and one that I killed over and over again to make it very hungry, then I cast the same spell with each multiple times.

I didn't take my experiment far enough to be meaningful, but it didn't seem to be yielding any difference. This counselor's explanation, however, sounds like a plausible reason that there are so many anecdotal accounts and confusion about eating in game.

 

[The Craftsman]

While I know it's been hammered in that it doesn't matter, I've often found eating to help when I have a bad casting streak.

Theres a difference between eating being programmed to help with something (which we've been repeatedly and quite rightly told isnt the case) and it having a side effect for which it was not designed. Thats what we are seeing here.

 

[MalagAste]

I was told the other day that when I start missing all the time as an archer I should eat... but not till I'm full full....

I am fairly sure my character is always starving...

As for RNG streakyness it's not streaky so much with me as it always and consistently rolls low for me... There is no streaks of "luck" for me... it's always and consistently BAD. As in normal people might get a 1:100 chance at something but I or any of my accounts have something closer to a 1:1,000 chance.

 

[The Craftsman]

As for RNG streakyness it's not streaky so much with me as it always and consistently rolls low for me... There is no streaks of "luck" for me... it's always and consistently BAD. As in normal people might get a 1:100 chance at something but I or any of my accounts have something closer to a 1:1,000 chance.

Do you have a single shred of evidence to back that up? I suspect its most probably just your own perception.

 

[PenNPaper]

Sweet

That's interesting. While I know it's been hammered in that it doesn't matter, I've often found eating to help when I have a bad casting streak.

I actually created two identical characters on test center once to compare fullness and casting--I had one character that was full and one that I killed over and over again to make it very hungry, then I cast the same spell with each multiple times.

I didn't take my experiment far enough to be meaningful, but it didn't seem to be yielding any difference. This counselor's explanation, however, sounds like a plausible reason that there are so many anecdotal accounts and confusion about eating in game.

Hi I have read in various places that eating till you are full helps. so there might be something to that. Also I have noticed with archery when full I tend to get more hits

 

[PenNPaper]

I find my archer has long streaks of missing mobiles that are my level or a little bit higher. I will land some nice hits and then it would go on a steak of any ware from 5, 10, and the odd time 15 misses. Hell just the other day my 70 archery missed an eagle twice. Even with accuracy gear I still missed a lot.

 

[Basara]

I've seen streaks of failing 5 times in a row on a 95% or higher ore smelting chance - I've also seen 6 exceptional items in a row on an item with less than 10% exceptional chance. I've also had 6 crafters claiming 3 BODs each, end up with 4-5 BODs for the same item (some with the same quality and count) out of the 18 (and smith BODs should be a 1 in 50 chance for two bods being for the same item, 1 in 55 for tailor). The odds of getting 4 out of 18 for the same Smith item type ( saw, getting 4 ringmail sleeves or 4 broadsword smalls) should be 0.000012%. (2%^4 * 98%^14)

The higher numbers streaking are almost impossible to spot by a player using combat testing - a hit is a hit.

It's only in crafting-related tests, where you can see the base chances for high effects and low effects actually displayed as numbers before you start crafting, that you can see both low-chances being enhanced or high-chances being suppressed by streaking.

 

[Victim of Siege]

I've seen streaks too . . .

 

[Parnoc]

It's had that problem for a long time - the problem is that it's not "Roll Sheeting" per se.

The issue is much deeper, in that the RNG is prone to streaking at ALL levels, not the bottom end. As a result, when they ran tests (OVER TWELVE YEARS AGO), with millions of "rolls", no streaking was apparent - "Roll Sheeting" would have shown streaking on the bottom end.

We FINALLY got Draconi to test for low-level streaking, and he was the one to discover that the streaks were happening at all points of the curve, not one end or the other. He started working on a replacement RNG in his spare time, but right before he was going to bring the RNG work in-house to start alpha-testing it (between KR and SA), EA went on one of their nonsensical "size-cutting" sprees (It always seems to be the branches least needing cuts that get cuts) and laid him off.

So, for almost 10 years, we've been stuck with an RNG that is KNOWN to have streaking issues. (It had streaking issues for over 20 years, now, but the actual proof has only been known 10 years)

I can personally verify this, I got in a conversation with Draconi and he said the exact same thing, "It's broken, it's been broken for years, I'm fixing it". Then boom, EA fires what I think was one of the best devs or maybe the best in UO's life and since then the devs have neglected this glaring, much talked about, error in the files of our beloved game.

 

[Ruppy2]

I've seen streaks too . . .

TRIVIA... But what song was on the B-side of the 45?

 

[Spartan]

TRIVIA... But what song was on the B-side of the 45?

B-side of "The Streak" was "You've Got The Music Inside" - release date Mar 27, 1974

 

[Victim of Siege]

TRIVIA... But what song was on the B-side of the 45?

It's up in the attic at the moment. i have all mine individually wrapped and sealed to avoid damage. (i recorded them and burned them to cd)

 

[G.v.P]

RNG, however modified, is random; humans have a hard time appreciating things that are truly random. A repeating 9 is random, but we try to make a pattern out of it because we generally try to seek out patterns for understanding. A true RNG, however, should not have a defined pattern; over the short term it is at times "unfair," but over the long haul, things balance out.

I guess I would look at the Gauntlet; I am unsure what other systems we have that include a guaranteed drop to improve upon the expectations of an RNG? Maybe it would be better if failure getting good BODs, for example, would bump the chance to get the best LBOD, etc. I guess you could consider the wisp shop a different form, by which you can purchase pre-generated items with a currency you have earned, which makes it a choice against RNG. Not sure if people still use wisp shop, though? To me, antique items are not worth the limited lifespan.

 

[The Craftsman]

RNG, however modified, is random; humans have a hard time appreciating things that are truly random. A repeating 9 is random, but we try to make a pattern out of it because we generally try to seek out patterns for understanding. A true RNG, however, should not have a defined pattern; over the short term it is at times "unfair," but over the long haul, things balance out.

Very true. The randomness is measured over tens or hundreds of thousands, not a small sample. Within the large sample you will get high number and low number streaks where numbers repeat. That doesnt mean anything is broken or its not random. People wrongly assume that random should give an even pattern. It doesnt in the short term. This is the very reason why you'll see a roulette wheel repeat a number 3 or four times in a row on occassion. Its still random, but randomness will by nature give repetition.

 

[Basara]

The problem is that technically RNGs are really pseudo-RNGs. They operate from a seed, to create the numbers.

When Draconi listened and checked what we were describing, he found that there were far more sequences of similar numbers in a row scattered through larger samples than random chance would allow. At least a magnitude more. So many, and evenly distributed, that pulling a million samples, they'd balance each other out.

However, that doesn't help when a crafter hits one of those streaks, and breaks 10 30% success enchantments in a row, or a miner splits his valorite ore into 8-ore piles to minimize loss (55% chance at 105 skill with the gloves) then completely loses 4 piles out of 10 (that's 4 different streaks of at least 5 fails in a row (if not continuing into the next pile) to take a pile of 8 large to 0, on a chance of 45% per fail, out of about 26-40+ attempts) - and have it happen nearly every time they mine.

It's not happening at the account level - the 6 smiths I mentioned in an earlier post pulling too many BODs for the same base item (before factoring in metal, number and quality) were on 3 different accounts, their BOD runs done over 10-15 minutes. So, it's either something server-side or client-side (most likely the server).

I primarily noticed the nature of the streaks personally because I was a math/CS teaching major in college who had to drop out for health & financial reasons as a senior. If I'd turned in an RNG as an assignment like that, I'd have gotten a lower grade. Matter of fact, Draconi said that, on looking at the RNG for UO, it was essentially one of a group of similar RNGs used by many free games (MUDs, Facebook games, etc.) because the type is public domain, and a type often used for teaching RNGs in computer programming - but not actually randomizing enough for robust use.

 

[The Craftsman]

So many, and evenly distributed, that pulling a million samples, they'd balance each other out.

Which is fine. That sounds like an acceptable number over which the 'rolls' would balance themselves out. It would be pointless balancing over a low number for obvious reasons. A million checks to the RNG probably happens over a relatively short time period.

 

[Basara]

Which is fine. That sounds like an acceptable number over which the 'rolls' would balance themselves out. It would be pointless balancing over a low number for obvious reasons. A million checks to the RNG probably happens over a relatively short time period.

So you're stating "it's fine" for the FUNDAMENTAL reason why it is absolutely, positively NOT FINE? If Draconi thought the level of streaking wasn't bad enough to try to replace it, he would have said so. But, apparently, what he found was shocking enough that he budgeted his own PERSONAL time (NOT work time) to try to come up with a fix, as all his time on the clock at work was trying to fix the festering pile of offal that was Kingdom Reborn.

A million checks are NOT happening over a short time period, except when running an RNG check. Hell, UO's programmed (being an artifact of the mid 90s) to where areas of the game go static if a PC isn't in them.

Near as I can tell (without looking at the code - and even then UO code that old by all reports isn't even commented, so I probably couldn't find it even if I knew the language after 20 years),
...all the RNG calls are being done separately, for each mobile (or mobile type), to where each mobile and process has its own seed. So, that the various animals have their own RNG instance, each skill check has its own RNG instance, each system has its own RNG instance, etc. This results in each having their own subset of streaking that can't be broken by mass calls from other systems. Streaking by someone pulling Smith BODs in Luna isn't sharing the RNG output going toward someone else pulling tailor BODs in Tram Skara, or that mass raid going on in Fel T2A - but they are sharing it with others pulling smith BODs on their shard - which usually AREN'T done by hundreds of multiple accounts at once on one shard (that would be required for the numbers you bring up).

If true, this in part explains why animals can often be found up against barriers (even pet ghosts, that wander off in straight lines after the owner dies at well), even though the "drunkard's walk" one expects from a fair RNG should get them around that barrier (only PCs, it seem, have pathfinding - and it sucks even for them). Why every profession sees streaks that don't seem to be affected by the rest of the in-game world, etc.

You know, it would crack me up if, when researched further, it turned out that the RNG seeds for PCs and animals were their hunger levels (which supposedly do nothing for PCs, and animals stopped feeding with the removal of the interactive biosphere in testing, due to the murder-hobos killing everything).

 

[PenNPaper]

The problem is that technically RNGs are really pseudo-RNGs. They operate from a seed, to create the numbers.

When Draconi listened and checked what we were describing, he found that there were far more sequences of similar numbers in a row scattered through larger samples than random chance would allow. At least a magnitude more. So many, and evenly distributed, that pulling a million samples, they'd balance each other out.

However, that doesn't help when a crafter hits one of those streaks, and breaks 10 30% success enchantments in a row, or a miner splits his valorite ore into 8-ore piles to minimize loss (55% chance at 105 skill with the gloves) then completely loses 4 piles out of 10 (that's 4 different streaks of at least 5 fails in a row (if not continuing into the next pile) to take a pile of 8 large to 0, on a chance of 45% per fail, out of about 26-40+ attempts) - and have it happen nearly every time they mine.

It's not happening at the account level - the 6 smiths I mentioned in an earlier post pulling too many BODs for the same base item (before factoring in metal, number and quality) were on 3 different accounts, their BOD runs done over 10-15 minutes. So, it's either something server-side or client-side (most likely the server).

I primarily noticed the nature of the streaks personally because I was a math/CS teaching major in college who had to drop out for health & financial reasons as a senior. If I'd turned in an RNG as an assignment like that, I'd have gotten a lower grade. Matter of fact, Draconi said that, on looking at the RNG for UO, it was essentially one of a group of similar RNGs used by many free games (MUDs, Facebook games, etc.) because the type is public domain, and a type often used for teaching RNGs in computer programming - but not actually randomizing enough for robust use.

This missing for 26-40+ attempts is very noticeable with Archery The amounts of misses are frequent and I go through huge amounts of arrows over a very short period of time "which is not fine"

 

[Pawain]

This fits here.

This must be the most common leg art shield build.

I got three Legs of a single roof boss corpse they were all this build! 2 shield types. The two same shield type had reversed names.



I was wearing this shield!



So, I had 4 on me.