I have to admit, "prediction" and "confirmation" among modelers is a pretty dodgy business. To normal people, the "prediction" has to precede the data that "confirm" it. But among modelers, it's standard practice to "predict" data that already exist and could be peeked at. Modelers should be required to include a supplementary document showing all the "predictions" they made (even in their heads), BEFORE looking at any data, that do NOT check out.