Quote:
Originally Posted by GeeTwo
DPR can be meaningless in three ways. First, every team can get the same number (with a bit of noise, of course). Then, it's meaningless, but that would not invalidate OPR. Second, if DPR is systematically incorrect, I have confidence that OPR shall also be systematically incorrect as well. FInally, if DPR has such wild scatter as to appear to mean something, I have confidence that OPR shall be wildly inaccurate as well.
|
Are you saying that one of these three cases happened this year, thus invalidating OPR? If so, which one?
Quote:
Non-linearity and limited game pieces inherently confuse linear models. For example:- Q. How many offensive points is a can burglar (by itself) going to be worth?
- A. If the alliance can't build more than three decent stacks, nothing. If the alliance can build and cap seven tall stacks, perhaps 100 points.
- Q. How many offensive points are making inverted totes and totes on the step scorable?
- A. If the alliance is not starved for totes, nothing. If totes become the limiting factor, up to 44.
|
This is true, these things will cause poorer results in linear models. At the highest-caliber events, many teams were performing better than their OPRs might imply, for reasons such as limited game pieces.
However, at weaker events, OPR seemed to be a very good description of a team's contribution, and at all events, the relative rank of OPR also seemed to be a good descriptor of a team's contribution relative to the rest of the field.