The subjectivity for me comes because not all players are equal, lumping multiple players into one value is very slap-dash method of determining who's better than who. For instance take the last EHL tournament on Xbox. In that season Russian Rockets finished in 6th place, so each of their players would sit at +1 based off your model. Which is fine, shows the team as a whole is atleast 'decent', on the other hand it doesn't show any variance within the team, instead it suggests that all players are at the same skill level. A player who got 70pts in 26 games would have the same rating as one who 18pts in the same time span, how is that not a flawed model if you're trying to rank how good players are? Say you were to adjust it to take into account points, goals and assists. How would you then differentiate play styles (For instance, one player who got 50pts could be a blue liner who gets a lot of breakaways, while one player plays the same position and plays a much more rounded defensively orientated game)? With the two contrasting play styles mentioned I personally would go with the latter player being more 'skilled', however then we get into who he's playing with amongst a whole host of other confounding variables. There's a reason advanced stats are used by actual NHL franchises to judge a players worth, without that independent eye you're just seeing if a player passes an 'eye test' whether they are good are not, which is extremely unreliable. (Due to many many biases) Now don't get me wrong, it is certainly possible, that said it would have to be so complex to take everything into account and hence would be very subjective unless fully planned with every variable accounted for. I'm talking time with puck average, average number of puck touches, average number of dekes, average shots for, average shots against... et cetera et cetera It is possible I'm thinking way too much into it but unless you really go in depth, any form of a player ranking system would be at the base of it, unreliable and subjective.