Handle data gathering and training more efficiently

[SereMark]

Give grandmaster games higher value, only use games above x ELO, etc...

[SereMark]

Also Incorporate game results in the training (how a move lead to the games outcome), and also reward moves that lead to victory, penilise the opposite, etc...