If you rate like this, how do you account for exceptional seasons? There would be no difference between Ovechkin's 65 goal season in 07-08 and Crosby's 52 goal season in 09-10.
Good question. My initial instinct was to range the spread between 50 and 95, and then keep 96-99 in reserve for exceptional performances of the sort you describe.
In aggregate, though, it doubt it would matter all that much either way, particularly if we continue to use a three-year sliding scale in our calculations (50/25/25 or whatever). So long as the resulting ratings give us an appropriate representation of where the player ranks versus other players in the pool, I'm not convinced it particularly matters what the ratings themselves are, so long as the guy who probably should score the most actually does more often than not -- within the limitations of the sim, of course.
For example, we've discovered in this version of the sim that players with SK below certain levels appear to be far less effective in a number of respects than more fleet-footed specimens, so this would all have to be subject to testing, just as blueliners' PA and SC ratings were.
Also, Matthew's repeated simming of individual game days recently shows a very, VERY high degree of randomness in outcomes from one run to the next. As it should be, I suppose - if we could write the script for a given game in advance, there'd be no point playing it! But I think it's important to recognize that whatever we do to try and normalize or quantify the ratings process, the actual outcomes of any given season will most likely vary significantly from expectations, purely because of the random factor.