Looks like the model choose a different strategy using variables for the second try, nice.