Anyone have any experience with genetic algorithms?

Anyone have any experience parallelizing genetic algorithms?

I am doing a project for school and it has to do with speeding up a genetic algorithm using parallelism.

Yes I do, but I can't really help you with no information provided.

Making a program that is trying to make the best possible fantasy football team using a genetic algorithm. Then trying to speed it up using parallelism.

This is a project for school that I haven't started yet, just got the idea for the project yesterday.

What are you going to use to determine success? Let me give a more concrete example:

Let's say I want to figure out the quickest way though a maze.

So you have the program:

1: pick a random path through the maze.
2: calculate the distance of this path.
3: make random changes to the path.
4: calculate the distance of the new path.
5: repeat steps 3 and 4, a certain number of times, then spit out the path that it found that had the shortest distance.

So to rephrase my question: how will you do step 2 and 4 with a fantasy football team? You don't have next year's stats to compare each trial team with to see how well they did.

I am using last years stats. I have the "perfect" teams for the first 4 weeks of last year that I can compare too. All of those teams have roughly 230 points total without exceeding the salary cap.

Well then you shouldn't have any trouble creating the perfect team for last year :P

eventually I want to alter it to use projections so that it can predict next week's best team.

do you think next week's performance is a function of last week's (or for that matter, last year's)?

no, I just have last years and can test with it using the data I have and being able to come up with a fitness score for each team.

Converting it to predict future teams will just use the projections instead of last years data to do the same thing. The functionality will be the same, I will just have to change how the fitness score is being calculated after I get it working with past data.

The problem I am trying to figure out is how do I evolve my teams? If I create 1 million random teams as the starting population how do I "kill" the bottom 90% and keep the top 10% and evolve these into the elite team?

I don't know exactly how fantasy football works, but I image you are only allowed to choose X number of quarterbacks, Y wide receivers, and so on. There is about 30 teams, maybe 2 or quarterbacks per team? And so on for each position. Might be possible to use brute force instead of random evolution to check em.