|
ChaosHunter can do sports prediction as long as you set things up correctly. Let's say you want to predict the winner of soccer games. Your output should be Home team score - visiting team score. If it is positive, you bet on the Home team, and negative bet on the Visiting team. Your optimization goal then should be Maximize % same sign.
Now your inputs should be constructed the same way as your output. For example one input might be Home Team Goals Last Game - Visiting Team Goals Last Game. Actually we'd probably use some sort of weighted average of goals over the last X games, or goals last time the two teams played each other, etc. Another input could be Number Home Team Injured Players - Number Visiting Team Injured Players. Your knowledge of the sport will help you decide what to use, and ChaosHunter will further decide, but the point is if you represent BOTH teams in the same variable, then there will only be half as many variables for ChaosHunter to sort through, plus the subtraction preprocesses the variable in a powerful way. Now you build a historical data based of all games played in the league for several years.
If you are into horse racing, it will be much harder because there are so many "teams". All we can suggest here is to build a general model that rates horses independent of the others in the race. Let’s say it ranks a horse from 1 to 10 based on previous performances. Then you run all horses in the race through the model and compare rankings. .
|