| Raw data | KOF |
| Version | vsoc 0.04-vsoc 0.05a |
| Class | vsoc.camps.goalgetter.GG1TestKickOutFactor |
| Project homepage | http://vsoc.sourceforge.net |
Watching previous simulations showed the effect, that hitting the ball out of the field was a common task although it is not advatntageous for soccer playing clients. This behaviour was forced by the fact that hitting the ball increases the fittnes of players. Specially if the ball lays near the out line players start to increase their fittness by hitting the ball out again and again.
To perevent players to kick out the ball this behaviour was punished. The fittness function was adopted that every plyer who kicks out the ball gets a penalty. The hight of this penalty relative to other penalties or rewards is called the kick out factor. This test attempts to find an optimal value for that kick out factor.
The fittness was calculated by the following function
public double fitness(Member m){
return
kickPerMatchFactor * m.kickPerMatch() +
ownGoalFactor * m.otherGoalsPerMatch() +
otherGoalFactor * m.ownGoalsPerMatch() +
kickOutFactor * m.kickOutPerMatch();
}
Attention: This function contains an error. OwnGoalFactor and otherGoalFactor are multiplied with the wrong values. The result of this is that players of this simulation try to score own goalrather than goals. This is a little bit confusing but does not really influence the results. Whenever you see ownGoals it means otherGoals and vice versa.
The default values for the factors where:
private double kickPerMatchFactor = 1.0;
private double ownGoalFactor = -100.0;
private double otherGoalFactor = 100.0;
The valuse for kickOutFactor where: -1, -10, -20, -50, -100
The amount of observations seems to be sufficient.
This diagramm shows that punishment of kicking out results in less goals. This would mean that it is better to have low absolute values of KOF. Eventually this result was produced by too short simulation runs (max. was 25000 matches).
It is interesting that very high (-1) and very low (-100) kick out factors result in more kicks. I would have expected, a lower kick rate for low kick out factors. If players are punished for kicking out the ball it seems to be logic that they develope the strategy of less kicking, as it should result in less kick outs.
Everything as expected.
To see if a low value of of KOF could ever lead to a better goal rate than high values the time series of selectedt simulation runs where analyzed.
It was expected, that the goal rate for low KOFs should increase at a certain number of matches that it might possible outreach the goal factor for high KOFs. The analyse does not show such a point. Only the increasing variance at 25,000 matches might indicate increasing goal rates. But the variance increases for every value of KOF at 25,000 matches.
All results seem to indicate that punishing of kicking out leads to less goals. Although the results seem to be obvious, further tests with higher numbers of matches should be executed.
For further tests the kick out values of 0 and 1 should be included.