Go back
I have learned..

I have learned..

Posers and Puzzles

c

Joined
10 Aug 04
Moves
1544
Clock
23 Oct 05
Vote Up
Vote Down

... something new today here that might have an impact on one of my pet projects at home. I discovered in the "birthdays problem" thread that I have been thinking very backwards when considering probability problems.
I have been trying to solve one of these for ten years (off and on).

The problem concerns predicting the outcome of a game of chance that can be described (haltingly) like this:

Two contestants A and B are defined by 4 variables each.
"Life" (Any value >0),"Attack" (0-100),"Defence" (0-100) and
"Damage" (any value >0). ("life" is most commonly much larger than "damage"😉
Chance decides who begins.
They then take turns.

One players Defence is deducted from the others Attack (P1).
A random number (R1) between 0-100 is chosen. If P1 > R1 or R1=0 a second random number is chosen [0-Damage] and that number is deducted from the Defending players Life.
Then roles reverse and the attacking player defends.

The above is repeated until one player has no life left.

I am currently trying to evolve a neural net using a genetic algorithm to
predict the probable outcome given only the eight (normalized) variables.

The net is placing the outcome in four categories 0-25% 25-50% 50-75% and 75-100%. It seems however I have some trouble getting the net to evolve past 90.xx% fitness.

Is there a way (simpler than what i am doing) to calculate the correct outcome?

Cookies help us deliver our Services. By using our Services or clicking I agree, you agree to our use of cookies. Learn More.