Yes that's correct. A negative correlation between X and Y means that the greater X is the less Y is, and vice versa of course. So that graph is showing that on average the offense affects the defense, and vice versa, in a positive way: the more points you score the (marginally) fewer points your defense allows and vice versa. Now, correlations on their own don't have any directionality to them so you can't a priori determine causality, but one way to see which is more likely to be the cause and which the effect is to find the best-fitting line when you assume one of the variables is "independent" (the cause) and the other is "dependent" (the effect). You look at the "noise" in the fit, or how good the fit is in either direction. The smaller it is the more likely you're looking at a tighter relationship and at least statistically more likely to be a cause-effect relation. In the case of points scored and points allowed, the fit is about the same so this seems to be bi-directional as far as one can see based only on this data.