Random dungeon finder - apparently not so random

[Fraza]

If you don't know the seed algorithm, they are for all intents and purposes, random. A hardware random number generator would not yield any better results noticeable in loot rolls or other wow randomizations. In fact they are only really necessary for cryptography. Of course I'm probably being pedantic and you were probably being facetious anyway.

Actually, it's not just about security. Software random number generators are known to be not that random. True RNG generators use some input known to be random like cosmic white noise. Of course I'd doubt Blizzard would use a hardware based solution just to generate rolls for epic lewt, but the problem remains that software based algorhythmically generated numbers are not random.

Just to give an example: remember any time when 2 players rolled 100 on an item? I can recall this happening on multiple occasions. The chance of this happening statistically is 1 out of 10 000. I don't think I have even seen that many rolls (in raids, that is, I don't even follow the need-greed system).

[Simca]

Actually, it's not just about security. Software random number generators are known to be not that random. True RNG generators use some input known to be random like cosmic white noise. Of course I'd doubt Blizzard would use a hardware based solution just to generate rolls for epic lewt, but the problem remains that software based algorhythmically generated numbers are not random.

Just to give an example: remember any time when 2 players rolled 100 on an item? I can recall this happening on multiple occasions. The chance of this happening statistically is 1 out of 10 000. I don't think I have even seen that many rolls (in raids, that is, I don't even follow the need-greed system).

You do realize that you're attacking random numbers with an anecdote, right? That is a logical fallacy, hence the Dilbert comic on the first page. I could set up a hardware random number generator to roll out of 100 five times, and if I got 100 all five times, I still would be unable to logically make the conclusion that it wasn't random.

Proving that random numbers are not random is nigh impossible.

On topic: Anyway, it is possible there is a bug considering they fixed another bug a few days ago along the same lines... randoms not working right while leveling.

[Fraza]

You do realize that you're attacking random numbers with an anecdote, right? That is a logical fallacy, hence the Dilbert comic on the first page. I could set up a hardware random number generator to roll out of 100 five times, and if I got 100 all five times, I still would be unable to logically make the conclusion that it wasn't random.

Proving that random numbers are not random is nigh impossible.

Yes, but if 10 million players set up a hardware random number generator and rolled 100 five times, would you still make the assumption that the RNG is working correctly? By that logic I could make an "RNG" that outputs 1 every single time and say you can't prove it doesn't work.

At some point you need to take the data you have and say "the chances of this happening are just too slim, the problem must be somewhere else".

[Simca]

Yes, but if 10 million players set up a hardware random number generator and rolled 100 five times, would you still make the assumption that the RNG is working correctly? By that logic I could make an "RNG" that outputs 1 every single time and say you can't prove it doesn't work.

At some point you need to take the data you have and say "the chances of this happening are just too slim, the problem must be somewhere else".

I guess I agree with the general idea of that, but I still think that one user's experiences are not a enough to draw a conclusion on anything. You would need to run an extremely large test, something on the scale that you purposed, to get an answer.

Maybe if you had an addon set up to roll once every second for 72 hours across 3 days exactly starting at midnight with a slight time delay at certain points to cover every millisecond and then compiled the data, you could make a conclusion about the long-term effectiveness of the RNG. You'd have to have spread tests to test the short-term effectiveness, of course.

[Monsterbaby]

Ha ha ha ha, when I queue for a dungeon I always get Halls Of Origination. Only once I've been so lucky getting stonecore.
Now I queue for a specific dungeon.

Indeed, it is retarded when You've done the same instance over and over again, and I'm about to vommit just thinking about Halls Of Origination.

[Fraza]

Maybe if you had an addon set up to roll once every second for 72 hours across 3 days exactly starting at midnight with a slight time delay at certain points to cover every millisecond and then compiled the data, you could make a conclusion about the long-term effectiveness of the RNG. You'd have to have spread tests to test the short-term effectiveness, of course.

My initial testing suggests that no one number is preferred. I wrote a script that rolls 1-100 exactly 1 billion times (1 000 000 000) and displays how many times each number occurred. Ideally each number should occur very close to 10 million (10 000 000) times, and they did. The largest variation was 0.05% (9 995 375) and even that was gone on the second testing. This is using the default software algorhythm Blizzard probably uses as well (or maybe they have some variables for hunters in there).

If I have more time, I'll test for stuff like rolling the same number in succession etc.

edit:

Tested succession. Statistically, the chance of getting the same roll twice is 1 out of 100. Again running the rolls 1 billion times I got 10 001 018 such cases. A 0.01% variarion to the statistic ideal.

I guess the software RNG is good enough. Must be the hunter variables then.

edit2:

Just to make this clear, this wasn't done in WoW. Blizz would probably ban my arse to the moon if I rolled 1 billion times in the game.

[Hresna]

I've actually noticed this as well with random heroics... By now, I've got at least 20-25 random runs done and I've only gotten Vortex Pinacle, Stonecore and Lost City of Tol'Vir. If i want to do any others than those three, I find myself needing to queue specific. Design flaw or just terrible luck? (And yes I have all the dungeons discovered, not to mention you get deadmines and sfk on heroic)