Castle Paradox

mjkrzak · Joined: 24 Nov 2009 Posts: 43

I'm looking for the correct syntax to set global variables via their ID number rather than their name.

Now, I know this doesn't work, but what it's supposed to do is systematically compare globals # 128-135, and make sure none of them are equal. If they are, randomly add -5 to 5 to the first compared value, set the test again flag, then do everything over again until no two values are the same.

So, what do I use in place of "global variable (count)" to get the job done? 'Cause I'm not seeing anything in the dictionary that fits what I need.

TMC · Posted: Sun Oct 21, 2012 4:58 am Post subject:

The commands you want are 'read global' and 'write global'.

By the way, you could write your for loops more simply like so:

mjkrzak · Joined: 24 Nov 2009 Posts: 43

I tend to be happy enough that my code is usually functional. I don't practice it nearly often enough to make the evolution to elegant.

Thanks for the assist.

Oh, and before anyone says so, I already spotted my misspellings of variable.

sotrain515 · Joined: 17 May 2010 Posts: 39 Location: Connecticut

mswguest · Guest

Doesn't this approach create a lot of wasted calculation? In fact, I wasn't even aware that Hspeak allowed a script to call itself (has this always been possible?).

It seems to me that this approach could very easily lend itself to potentially infinite loops. For one thing, every time you add "random(-5,5)", there's a chance you just added 0 so you haven't gained anything.

More importantly, with this approach, there's a chance that you keep changing things but never actually gain any ground in terms of unmatching variables. For example, imagine if all of the variables start out the same:

Bob the Hamster · Posted: Tue Oct 23, 2012 6:27 pm Post subject:

TMC · Posted: Thu Oct 25, 2012 4:16 am Post subject:

I would like to do an in-depth analysis, but probabilistic algorithm analysis is not my speciality, and this one is quite complex to analyse. Each element of the array is distributed differently and dependently: the first element can be changed up to 5 times through the function, while the last never is.

I would hope to estimate the running time by modelling the 5 array elements as random variables normally distributed around 0 (assuming the worst case where the initial values are all equal). Thanks to the Central Limit Theorem, this should be a decent approximation.

I did not complain about the use of recursion because intuitively the probability of a long running time is extremely small. The script interpreter allows up to 128 scripts running at once, so this is no problem. So msw's alternative is far too complex to justify.

But I figured I could do a quick simulation (starting from all-zero). Here's what I found:

Mean number of loops = 2.71 max over 5000 trials = 11

Increasing the array length to 20:

Mean number of loops = 8.20 max over 5000 trials = 43

Also, I looked at the distribution of values. Length 5 array:

Bob the Hamster · Posted: Thu Oct 25, 2012 6:26 am Post subject:

That is interesting, TMC.

Actually measuring speed is always better than just guessing at it :)

I am really curious how these numbers are actually being used in the game. What is their purpose?

Guest · Posted: Sun Oct 28, 2012 9:49 pm Post subject:

This strikes me as far too much analysis for what I intend to use it for, but we do find our fun in the oddest places so...

The ultimate goal of this is to make a simplified battle system in a semi-plug in form. Like a multi-layered rock/paper/scissors dealie. I make it work, release it into the wild, then modify it again to my own giant golem castles slugfest minigame for my likely never to be finished RPG.

As it stands, the script is a simply a double check on the initial initiative scores. Each of the up to 8 targets gets a randomized value of 100-1000 + a speed score. (Don't want it to go negative, as I use -1 for targets not in play.) Once given, I run the checker to make sure no two values are the same. If they are, bump one and double check again. Repeat as needed, and use the numbers to determine who goes first-next-last.

As I see it, 8 randomized values in a field of 900 are unlikely to duplicate. If they do, there's a 10 in 11 chance that it'll be corrected first run and better odds in subsequent runs. While there's a slight chance for things to go off the rails, I see the odds as insignificant for the amount of times I intend to call the script. What, once a fight for maybe a half dozen times in a playthrough?

mswguest · Guest

Yeah, I noticed later that I misread your code. Now I notice it's a lot like what I was doing but backwards (you 'fix' numbers starting from the back in some sense, since your final number is never changed and your 2nd to last is 'fixed' mostly by changing the preceding numbers). I also imagined all of the numbers fairly close together, not in a field of 900! I don't know why; I guess because the 'bumping' was in a range of 10, I figured you were trying to keep the numbers fairly close to each other...? I guess this still confuses me a bit. If the initiatives are random from the start, why 'bump'? Why not just reassign altogether?

mjkrzak · Joined: 24 Nov 2009 Posts: 43

My first response would probably be: "That's what I thought to do first and there's no real reason why I couldn't reassign."

The rationalizing answer would be: "They came up close in the first randomization, so they should be next to each other in sequence rather than potentially throwing off the rest of the shuffle."

If pressed, the fallback position from the rationalizing answer would be a grudging first response.

In the end the numbers don't mean much, as they all get translated into a simple 0-8 (0 for unused) order.