AV, the course stats [driving distance, driving accuracy, greens in regulation & putting average] are published after each event on the PGA Tour site on a season-to-date basis. You can calculate the per tournament stats by comparing one week's stats with those for the previous week, but instead I like to take a 12-month rolling average as it eliminates short-term variability ['noise']. If the current tournament is 24th of 45 in current year, then I calculate 12-month stats as 23/45 * this year's stats + 22/45 * last year's stats. Close enough to 12-month average as a rule of thumb.
Re: 3-year course form. I think three years' data is sufficient to generate a good view of who plays well on a particular course without statistically discriminating against the newer pros who have only joined the Tour in the previous year or two. I think lengthening the time span would make the 'has-beens' look better and the 'gonna-bes' look worse. What is more, the course will be modified every year, or the time of year that the event is played may be changed, so such cumulative changes can become very large over time.
I also use the course form stats in conjunction with the course compatibility stats as they measure how well a player should perform given the relative strengths of his game on this course. Three years' data on what they did score and/or what they should have scored should be plenty.
