Predicting a players future stats/performance

[vzografos]

Hi,
so I want to do some prediction on a given player's future statistics. Say, predict what the average FG% (field goal pcnt) is going to be for the next season.

I have some historical data of that player's past statistics (FG% averages from a number of past seasons).
What would be the best way of going about it?

I was thinking along the lines:

1) Fitting a regression curve to the past data and using this to extrapolate next season (assuming some smoothness etc)
or
2) Coming up with a simple model of stat evolution based on a players age (and perhaps other physical factors e.g. weight, height)
or
3) Assuming a Markov chain of order 1, i.e. next season's stats depend only on previous season stats. So next season prediction is around the previous season stat +- some deviation. Then trying to reduce that deviation as more games are being played in the current season.

or
4) Do something similar to 538's CARMELO prediction. That is, once we have a fitted curve on the stat evolution, try to match it with other (past and current) players with similar curves and interpolate from them.


Those would be my starting points.
What do you guys suggest? Anything else I could look into?

[Crow]

If you haven't already, I'd read what is available recently from @kmedved and try to find old stuff from Ryan Parker about confidence intervals (here, on net or by internet wayback machine). And maybe the skill grade stuff at bball-index.com. Among other possibilities.