Print

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

[Surma]

Are the values scaled to the total number? e.g. if you have all but 1 of them at the default value, and that 1 at 10000, is it (default * x) + 10000, or does it range from the 25600 * (x + 1)?

What I'm trying to abstract there is, say you have OWN_RACE:25600, FANCIFUL:256, EVIL:256, GOOD:256. would it scale from OWN_RACE/26368? Or would it scale from OWN_RACE/102400?

Having it scale from A/(w+x+y+z) makes more sense to me but, I'm just trying to be accurate.

[Ghostpaw]

O well I guess I'll have to refer to my notes on propability and return with my own guestimations.  We wouldn't be having this problem if we had the source code but we don't and there is nothing we can do about.

[ May 24, 2008: Message edited by: Ghostpaw ]

[Ghostpaw]

I just realized I didn't completely understand something.  For ART_FACET_MODIFIER, what OWN_RACE refers to is pretty obvious.  But what I don't know is what FANCIFUL refers to and, to a lesser extent, what EVIL and GOOD refer to.

[Kagus]

They are all creature tags.  Trolls are [EVIL], unicorns are [GOOD], griffons are [FANCIFUL].

[Ghostpaw]

Well I have to say I am not satisfied with the previous poster’s explanation to my question.  When he stated that if OWN_RACE was assigned the value 25,600, that 99 out of 100 images will have that race doing something I accepted it.  However it did not give me much for a comparison to figure out how drastically the numbers affected the outcome so I asked for an example of what will happen if OWN_RACE was assigned the value 10,000 instead?  Instead of out right stating that he was pulling numbers out of his ass and that his previous example was full of shit, his attitude changed to that of an arrogant person and threatened to reference the file again and pulled out another number 97 out of 100 with an extremely questionable piece of evidence.  Is he saying that he is getting his numbers by how large of a difference the inputted value is to 256?  Now alarm bells are ringing through my head but I give him the benefit of the doubt so I ask for an example for the default value of 256 so I would have something to use as a reference point.  I did not get a direct answer even after asking for specifics and instead Sean Mirrsen just gets hostile.  Why did he get hostile?  Sean Mirrsen’s motives become only speculation at this point but I would say it has something to do with his inability to directly and confidently answer my question.  Afterwards he admits to have been playing me for a fool the entire time and really had no real information pertaining to the subject.  Not even any justifiable logical guesstimation.

However, I think I've finally figured it out.  At first I was thinking about how 25,600 was 100 times higher than 256.  To make things easier to follow lets assume that OWN_RACE was changed ffrom its default setting of 256 to [ART_FACET_MODIFIER:OWN_RACE:25600] with the other objects staying at 256.  A person would assume that OWN_RACE would also be 100 times more likely to be chosen over the other objects.  But what if all the objects numbers were modified to 25600?  The objects (OWN_RACE, FANCIFUL, EVIL, GOOD) can't ALL be 100 times more likely to occur.  That would be impossible.

So that got me thinking to how I would use these values.  I would assign each object to a section of a range of values (now to be referred to as a set) and have a randomly generated number using the computers own internal clock as a seed along with a long complicated formula to create unpredictability and have the resulting value compared to the whole or established range of values (now to be referred to as the range of values).  I would then have any value equaling or rounded to 0 to be rejected while any other values will be compared to the range of values.  If the value is not within the range of values it will be rejected.  All accepted values will be compared to each object’s set and if the resulting value is also found in that object’s set that object will be selected.

“*” will represent the multiplication sign without the quotes.

So how would this work with the ranges 0-25,600? Well quite easily once I figured out how I would do it.  Simply assign the first object to a set from the range of values from 0 + (1*inputted value).  The other object’s set will be established from the range of values starting from the [end value of the previous object’s set] + 1 to [previous object’s end set value] + (1 * inputted value).  For clarities sake I will use an example.  Let’s assume OWN_RACE was assigned [ART_FACET_MODIFIER:OWN_RACE:25600] while all other objects (FANCIFUL, EVIL, GOOD) will be using their default values of 256.  Let’s also assume that OWN_RACE will be the first object whose set is defined.  It will start out as 0 + 25600.  The answer is 25600 so OWN_RACE’s set will be;

[OWN_RACE set: 0 – 25,600]

Let’s now assume FANCIFUL will be the next object whose set is defined.  It would work like this.  25,600 + 1 to 25,600 + (1* 256).  This would give a set range of;

[FANCIFUL set: 25601 – 25856]

Now let’s assume EVIL will be the next object set to be defined.  This would be how it would be calculated from my theory.  25,856 + 1 to 25,856 + (1* 256).  Again this would give a set range of;

[EVIL set: 25857 – 26112]

By now you will have figured out the pattern but I’ll complete the last object’s set to be able to go on to my next point.  Let’s assume GOOD will be the last object whose set is defined.  Like the others it could work like this.  26,112 + 1 to 26,112 + (1* 256).  The set range for GOOD is now;

[GOOD set: 26113 – 26368]

Now the whole range of values is 0 – 26,368.  Now to discover the real  probability of an objects likelihood of appearing simply take its inputted value and divide it by the end value of the range of values.  To use our previous example I will find out the probability of OWN_RACE being selected.  25,600 / 26,368 = 0.97087.  This gives OWN_RACE a 97% chance of being selected!  

If anyone finds anything wrong with what I’ve thought up of feel free to give me some feedback or even to just ask for clarification.

[Sean Mirrsen]

The only solid fact is that we don't know anything about how the system operates. From the comments we can see that the values are multiplicative, which seems to imply that there's a certain preset chance for each, and the values we edit are modifiers for those, so depending on the initial values (which may not even be equal) the percentage correlation can vary wildly. We can only edit the variables in realtion to whatever's set in the game by default. In this sense, yes, EVERY variant can be 100 times as frequent.

And I wasn't offensive, I was merely annoyed. The "Vogon-after-destroying-a-planet-and-discovering-a-hitchhiker" kind of annoyed.   Without knowing the inner workings of the system, there's no way to accurately specify any kind of value and measure it against another. Especially when there isn't so much as a hint at an incentive to actually have that kind of accuracy. What's the point of knowing three billion digits after the floating point in PI, when a hundred yields millimeter accuracy in planetary orbit calculations?

Also, you might have taken a simpler approach at your goddam calculations. 25600 is 100 times as frequent as 256. With other values at 0, the chance for this one to be selected is 100%. Add three 256 values to the three other variants, each 1% of the 25600. Now deduct 3% from 100. The net result is roughly 97%.   I won't be surprised if the method works with sufficient accuracy for other value combos.,,

[ May 25, 2008: Message edited by: Sean Mirrsen ]

[Deon]

Sean, obviously you shouldn't be annoyed about questions. If you don't like them, don't answer them =).
At first I thought they were stupid ones too, but after reading questors' posts I realised that they had a sense.

As for me, I use relative numbers to make sure my percent (i.e. 512 for plants and leave all other things default 256 - thus I suppose that my plants will be 1/3 chance while types are 1/6).

[ May 25, 2008: Message edited by: Deon ]

[Sean Mirrsen]

When I offer help, I usually follow through, unless I get bored and lose interest. This wasn't the case, and the questions seemed normal until I got mistaken for a supercomputer with direct access to DF source code.

Speaking of which, I wonder what SkyNet would do if it found DF...