Filip W.
Sweden Linköping
Euros are better with dice!

I'm using absorbing Markov chains to predict the outcome of sequential battles but the software I'm using only allows for a matrix as input. So I need to manually calculate all the possible states in which a battle may be.
For example, If A is fighting B and b then after A's attack there are three possibilities: A (alive), B (alive), b (alive); AB (with b dead) or Ab (with B dead). Then I need to calculate the probabilities for each, feed them into a matrix and so on for each state in the battle (this was a single state; ABb, another may be AB, Ab, BB etc.)).
Does anyone know if there is a software where I could setup the rules for the battle (i.e. probability of a successful attack by A, B and b) and the software would calculate all the possible states automatically  or even better feed them into the Markov analyzer and present me with a set of finished outcomes with probabilities.


Matt Davis
United States New Concord Ohio

Well, I'm not sure exactly what calculations you need to do, but MATLAB is a fairly standard tool for any kind of matrix work. You could probably do what you need in Mathematica or Maple, and they might even have packages designed for Markov processes specifically. I think Mathematica has a free trial thing where you can do most computations, but not save any work, though I'm not sure about that.


John Lopez
United States Tucson Arizona
Get off my lawn!
The explanation: Impossible Triangle + TW (my company initials) = my logo.

filwi wrote: I'm using absorbing Markov chains to predict the outcome of sequential battles but the software I'm using only allows for a matrix as input. So I need to manually calculate all the possible states in which a battle may be.
For example, If A is fighting B and b then after A's attack there are three possibilities: A (alive), B (alive), b (alive); AB (with b dead) or Ab (with B dead). Then I need to calculate the probabilities for each, feed them into a matrix and so on for each state in the battle (this was a single state; ABb, another may be AB, Ab, BB etc.)).
Does anyone know if there is a software where I could setup the rules for the battle (i.e. probability of a successful attack by A, B and b) and the software would calculate all the possible states automatically  or even better feed them into the Markov analyzer and present me with a set of finished outcomes with probabilities.
I do this kind of computation frequently, and I simply toss together a quick Monte Carlo simulation and run a few hundred thousand battles or even a few million if there are a lot of possible outcomes. The advantages:
a) You can model complex systems with the same framework simply by plugging in a new CRT resolver. b) You don't have to worry about combinatorial explosion. c) You can have selected individual battles create a report so you can see if the results of the quick, medium and long battles are what you wanted (or other criteria you use to trigger reporting). d) You can accumulate interesting stats along the processing at various stages.
EDIT: typos


Darlene Lopez
United States Tucson Arizona

Most stats geeks end up writing their own when necessary, but I know of a couple that you can get. There's some out of Canada that I've seen used a couple of years ago http://www.cs.toronto.edu/~radford/fbm.software.html . This used to run just on unix systems, but I think it'll run on linux now as well. I think this is still a freebie, but you'll need to check.
Dr. Stewart out of North Carolina also has written a small program that works very nicely. I've seen this more recently and the output is similar to other stats packages. I'm not sure what if anything he charges for a copy of the software though. Here's his homepage http://www.csc.ncsu.edu/faculty/stewart/welcome.html
The most user friendly (but definately not the cheapest) is AMOS from SPSS. If you can find someone with a copy of it floating around (my last version is 4.0 and they're at 7.0 now) then you can fiddle around with it. AMOS does all sorts of structural equation models with a nice Markov Chain Monte Carlo sim.
Good Luck.



