Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Holy Mittens]

Nah, I've had the exact same results happen over and over with the same seed.

[The13thRonin]

Nah, I've had the exact same results happen over and over with the same seed.

So you can reproduce the elf king exactly as he was?

[Vester]

Yeah, would the parent-eating and the duels happen all as they did before?

I'm still confused as to how creatures affect the path of events in worldgen.

[DeathOfRats]

I believe during worldgen battles and duels are still decided by %'s and RNG's. So you'll never have the same world, even after genning from the same seed. You will have a similar one, but never the same.

That doesn't make much sense. After all, if the seeds are the same, and nothing else has changed, the RNGs generated at each step will always be the same, as well as the percentages. Therefore, you should get the exact same result each time.

Either that, or the seed system is bugged somehow.

[Neruz]

I'd say what's happening is the game is using the % of megabeasts dead AND the year to start checking as a part of some of it's random seeds, so when you change those values you change the seeds, and thus change the world.

[Random832]

Maybe Dwarf Fortress really works that way - maybe a given seed won't generate the same history over and over again because world generation itself has a random factor. (Who gets born, who dies, who kills who, who wins the war).

Um, the whole point of a seed is to control the random factor. The history process (along with the worldgen process) can have millions of places where it asks for a random number and makes a decision based on it - what the seed does is determines what the list of random numbers will be. The seeds control the RNG. That's what they're there for.

@Neruz, yes, any factor will affect it, not just the seeds. But it doesn't change the random number generation. How it works is since any other change might make it ask for a random number in a different place for a different decision, and every time after that will get a random number from a different point in the sequence, that changes how the world turns out.

[Neruz]

The key to remember is that the better the random number generator is, the more factors it draws upon. The best random number generators grab unrelated data from all over your system to use in their calculations; the seeds are only the starting point for the calculations, nothing more.

[Vester]

Maybe Dwarf Fortress really works that way - maybe a given seed won't generate the same history over and over again because world generation itself has a random factor. (Who gets born, who dies, who kills who, who wins the war).

Um, the whole point of a seed is to control the random factor. The history process (along with the worldgen process) can have millions of places where it asks for a random number and makes a decision based on it - what the seed does is determines what the list of random numbers will be. The seeds control the RNG. That's what they're there for.

@Neruz, yes, any factor will affect it, not just the seeds. But it doesn't change the random number generation. How it works is since any other change might make it ask for a random number in a different place for a different decision, and every time after that will get a random number from a different point in the sequence, that changes how the world turns out.

Well, you have the word Random in your name, so you must be correct.

[Random832]

The key to remember is that the better the random number generator is, the more factors it draws upon. The best random number generators grab unrelated data from all over your system to use in their calculations; the seeds are only the starting point for the calculations, nothing more.

Um, except when what you want is to be able to produce the same thing every time.

The random number generator only uses the seed. The same seed will produce the same sequence of 'random' numbers every time. This is extremely useful for some purposes, and DF world generation is one of them. If this was not the case, there would be absolutely no point in providing the seed at all. The worldgen process is influenced in a chaotic way by other factors in the way I described, but that is not the same thing as saying those factors affect the RNG.

You are talking about random number generators for e.g. cryptographic applications, which is an entirely different usage for which that quality is not desirable, and for which using stuff from all over your system (I/O timings, line noise from the sound card, etc) is done. And those don't have anything that can be described as a "seed", because it would be pointless since the only purpose of a seed is to be able to get the same sequence of random numbers.

[DeathOfRats]

The key to remember is that the better the random number generator is, the more factors it draws upon. The best random number generators grab unrelated data from all over your system to use in their calculations; the seeds are only the starting point for the calculations, nothing more.

Um, except when what you want is to be able to produce the same thing every time.

The random number generator only uses the seed. The same seed will produce the same sequence of 'random' numbers every time. This is extremely useful for some purposes, and DF world generation is one of them. If this was not the case, there would be absolutely no point in providing the seed at all. The worldgen process is influenced in a chaotic way by other factors in the way I described, but that is not the same thing as saying those factors affect the RNG.

You are talking about random number generators for e.g. cryptographic applications, which is an entirely different usage for which that quality is not desirable, and for which using stuff from all over your system (I/O timings, line noise from the sound card, etc) is done. And those don't have anything that can be described as a "seed", because it would be pointless since the only purpose of a seed is to be able to get the same sequence of random numbers.

Well said! I'm with Vester, Your name was well chosen 

[Neruz]

The key to remember is that the better the random number generator is, the more factors it draws upon. The best random number generators grab unrelated data from all over your system to use in their calculations; the seeds are only the starting point for the calculations, nothing more.

Um, except when what you want is to be able to produce the same thing every time.

The random number generator only uses the seed. The same seed will produce the same sequence of 'random' numbers every time. This is extremely useful for some purposes, and DF world generation is one of them. If this was not the case, there would be absolutely no point in providing the seed at all. The worldgen process is influenced in a chaotic way by other factors in the way I described, but that is not the same thing as saying those factors affect the RNG.

You are talking about random number generators for e.g. cryptographic applications, which is an entirely different usage for which that quality is not desirable, and for which using stuff from all over your system (I/O timings, line noise from the sound card, etc) is done. And those don't have anything that can be described as a "seed", because it would be pointless since the only purpose of a seed is to be able to get the same sequence of random numbers.

Oh yeah, if you want a reproducable random number generator then you make it only use a set of input values; DF can accurately reproduce the same worlds over and over, so it clearly only uses the values given in the world gen, but judging by the fact that changing the year to check megabeasts and the megabeast % changes the world it seems pretty clear that it does not only just use the seeds for world gen, but other values in the world gen as well.



The fact that the exact same world, complete with same fights and battle outcomes, occurs with the same world gens tells me that everything in world gen, including the movement of creatures and the outcome of battles, is based on the values in the world gen file, although apparantly not always on the seed values.

[DeathOfRats]

The fact that the exact same world, complete with same fights and battle outcomes, occurs with the same world gens tells me that everything in world gen, including the movement of creatures and the outcome of battles, is based on the values in the world gen file, although apparantly not always on the seed values.

That's logical, though. If you change, say, the year the world starts checking for megabeast percentage, you *are* going to get a different world, simply by dint of having to do different calculations, even if you're using the same random number sequence.

Likewise, if you change, say, mesh percentages, you'll get a different terrain gen even with the same numbers.

In either case, the RNG sequence is still the same, only you're not using the numbers the same way. Logically, you're going to get different results in that case.

[Neruz]

The fact that the exact same world, complete with same fights and battle outcomes, occurs with the same world gens tells me that everything in world gen, including the movement of creatures and the outcome of battles, is based on the values in the world gen file, although apparantly not always on the seed values.

That's logical, though. If you change, say, the year the world starts checking for megabeast percentage, you *are* going to get a different world, simply by dint of having to do different calculations, even if you're using the same random number sequence.

Likewise, if you change, say, mesh percentages, you'll get a different terrain gen even with the same numbers.

In either case, the RNG sequence is still the same, only you're not using the numbers the same way. Logically, you're going to get different results in that case.

Logically you'd expect the end year to have no impact on the actual makeup of the world; i certainly expected the checking stuff to be seperate from the actual generation parameters.

This is clearly not the case, what appears to be happening is the 'watcher' is contaminating the 'experiment' by dint of existing, which is interesting and somewhat unexpected. Such contamination is expected in real life, but generally not in computer programs.

[Random832]

The fact that the exact same world, complete with same fights and battle outcomes, occurs with the same world gens tells me that everything in world gen, including the movement of creatures and the outcome of battles, is based on the values in the world gen file, although apparantly not always on the seed values.

Yeah, that's because the random numbers aren't the only influence.

How about a concrete example: Say you have World A and World B. World A has more megabeasts than World B.

So in the history generating process for World A, some hero gets into a battle with some megabeast that doesn't exist in World B. Now, meanwhile on the other side of the world, there is some battle between two armies. The outcome of the battle may be different between the two worlds because some of the random numbers from the battle between the hero and the megabeast are going to be used in the other battle in World B, and will be different from the ones that got used for that battle in World A. And down the line, different events will happen due to the different outcomes of that battle, etc, and anyway the RNGs of the two worlds are now permanently "out of sync" simply because World A happened to pull a few more numbers from it that one time than World B.

The factors don't directly influence the RNG - it still spits out the exact same sequence of numbers, 16838 5758 10113 17515 31051 etc. But those numbers quickly end up being used in different ways if anything else is different.

[Neruz]

I'm not sure what you're getting at; as we can see from testing, if you use the same world gen then World A does not have more megabeasts as World B, it has exactly the same number, with the same names, of the same types, in the same locations, who fight the same battles and die the same deaths.

If World A has more megabeasts then World B, then you've changed the equation somewhere.