Please fix the Random Number 7 Generator

[GarthGrey]

Edited, it was probably just a troll post anyway.

 

[Lorddog]

Isaac Asimov - Foundation series

 

[Winker]

The RNG has never really been Random. It tends to have streaks and runs on certain numbers

 

[MalagAste]

One of my friends who does a lot of the pirate ship hunting said that it seems different on different shards... some shards #4 might be "rare" but on another shard #5 seems rare etc...

But IMO the RNG is a steaming pile... and hates most people.

 

[PaulCH]

I call it this, because I have 31 Part7 ship plans, and 0 part 3s. That's not coincidental numbers, that's garbage. The ship parts , while not terribly difficult to get, are also not terribly easy. No piece should be any more rare or common than the other. So please explain to me how I have this breakdown.

#1-7
#2-1
#3-0
#4-5
#5-5
#6-1
#7-31
#8-15

That doesn't look random to me.

As I understand it, the RNG is random but flawed. Over a large sample size, results will come out random but over smaller sample size, results will tend to group.
Scroll down in this thread to see the post by Basara... RNG revamped

 

[Basara]

Well, i dont consider that a small sample, as I've crafted and sold close to 50 ships, still have 6 completed sets, and those are my current ones. And anyone who frequents Dread Pirates knows that when you capture them you either get, gold, gold and special ammo, or gold and a ship part. I also have over 500 rounds of special ammo, so I've sunk enough for it to be a fair sampling of the pitiful RNG.

By Computing standards, samples that aren't in the thousands (if not hundreds of thousands) of instances are considered "Small". That's why it took Draconi until 2007 or 2008 to find the small-sample streaking issues - then was laid off by EA before he could finish a fix for the RNG.

With the large samples (typically, they were testing it with 500k-2 million instance samples), you had so many small-level streaks across all parts of the range, that they averaged out, giving the false impression of no streaks (at least, until you actually went into the data and looked at small regions within the data).

 

[Lorddog]

for those who didn't get my "Isaac Asimov - Foundation series" post
- the series is set in the future when technology knowledge diminishes enough so that people cannot create it anymore. they can only repair and service.

UO code is so old and build Frankenstein patchwork so that no one can any longer create new enhancements for it. all they can do is add a patch here and there.

 

[Lord Nabin]

Isaac Asimov - Foundation series

Lord that brings back memories

 

[Val-Tur]

I call it this, because I have 31 Part7 ship plans, and 0 part 3s. That's not coincidental numbers, that's garbage. The ship parts , while not terribly difficult to get, are also not terribly easy. No piece should be any more rare or common than the other. So please explain to me how I have this breakdown.

#1-7
#2-1
#3-0
#4-5
#5-5
#6-1
#7-31
#8-15

That doesn't look random to me.

For 3-4 weeks I hunted pirate ships daily for the plans. I had the same experience as you. I did get 3 full sets in the end. But it is very very frustrating.

 

[Spartan]

Lord that brings back memories

Scary ... I thought the same thing. And let's face it ... UO is in the maintain phase - has been for many years.

 

[Uvtha]

Well, i dont consider that a small sample, as I've crafted and sold close to 50 ships, still have 6 completed sets, and those are my current ones. And anyone who frequents Dread Pirates knows that when you capture them you either get, gold, gold and special ammo, or gold and a ship part. I also have over 500 rounds of special ammo, so I've sunk enough for it to be a fair sampling of the pitiful RNG.

But, of course, with a rng even a perfect one, this kind of thing WILL still happen. Otherwise there would be some check in place to prevent streaks which would make the system no longer random. It's perfectly possible to roll a dice 50 times and never hit 1, or flip a coin heads 50 times. Unlikely, sure, but still possible.

While I personally would prefer that all RNG systems would be replaced with systems with specific results based on set variables, that's unlikely to happen, so sometimes some people will just get big runs of bad luck. Good luck too, but those aren't as memorable.

 

[petemage]

But, of course, with a rng even a perfect one, this kind of thing WILL still happen. Otherwise there would be some check in place to prevent streaks which would make the system no longer random. It's perfectly possible to roll a dice 50 times and never hit 1, or flip a coin heads 50 times. Unlikely, sure, but still possible.

While I personally would prefer that all RNG systems would be replaced with systems with specific results based on set variables, that's unlikely to happen, so sometimes some people will just get big runs of bad luck. Good luck too, but those aren't as memorable.

Dilbert Comic Strip on 2001-10-25 | Dilbert by Scott Adams

 

[Basara]

examples of streaking on the small vs. large level.

I did a large sample of BODs:

Over thousands of BODs the spread of smith BODs came within tenth's, if not hundredth's, of a percent of the expected number.

BUT

with the 7 smiths I was pulling Smith BODs with, in the days before you could pull 3 BODs at once, I would sometimes get 3-5 smalls of the same general type.

The counts and qualities might differ, but those are 1 in 3 & 1 in 2 decisions respectively. On the other hand, the chance of getting, say, 4 scimitar BODs from 7 pulls is a bit extreme (1 chance in 49 for each BOD after the first to match (48/49 for misses)), so getting 4 BODs for 1 item type in 7 pulls is 0.0000083264, or 0.00083264%. (1 * 1/49^3 * 48^3/49^3)

And this was happening several times a month. Something was not right, and since it was happening across 3 accounts, it meant that it was probable that either the time/date, or something from the user's computer (which of course would be the same for all 7 BOD runs), was the part of the RNG seed, causing the streaking issue. However, the streaks don't occur from multiple pulls on one account (such as with the 3 successive pulls from one character), but does still make the initial pulls with each character a bit streaky in relation to each other.

 

[apple][]

chi-square test on the data from the original post is 2.4747E-15%. Assuming each of the 8 items are dropped with same probability as in fair dice case, the randomness of the RNG is questionable.