Okay, I will just put it here it is probably easier that way, but I will warn you that I am a little long winded on this kind of stuff.
What I did was load the rosters of each D3 team into a database and then take the top 3, top 6, and top 9 players from each team for each core attribute (I did all this last season). My thinking in doing that was to find attributes that are important to everybody (like athleticism) and those that are important to have a couple of players high in (like perimeter).
Once I did that, I ran a script that would apply a weight to each of those values and come up with a total team rating. It would then rank each team by rating and compare it with the projection report rankings and return a total variance. I had it continue to apply different weights to each group to get the lowest variance per team that I could. It ended up so that about 70% of the teams were within about 40 (e.g., a team I had at 150 was within the range 110-190 on the projection report 70% of the time). Since there was a lot of bunching in the rankings except for the very top end teams I thought this was about at as good as I was going to get it.
I then proceeded to work on determining my own method of calculating the spread between two teams using their rating and the site of the game. I did this by analyzing about 300 games between SIM teams only (I think human controlled teams have a pretty significant effect on the outcome through game planning and I was concerned that it would poison the results so I skipped any game with a human involved). I found that the most accurate prediction I could make was multiplying the rating the difference between the two teams by 2.4 and then giving the home team 1.1 points. I could then determine the percentage that a team would win by applying the percentage that the results of the actual games that I analyzed were within the spread. For example, 50% of the games that I analyzed were within about 7.5 points of my spread, so I would calculate a team that is supposed to win by 7.5 points will win about 75% of the time. That is how I came up with the predictions for the tournament results.
There are a few things that this does not take into account, most notably game-planning. It also does not give credit to coaches who have recruited specifically for the offense and defense that their teams run. In order to be a more accurate picture it would probably need to have separate formulas for each offense-defense combination. It also does not take into account IQ for those teams that may be switching offenses or defenses.
I did learn a few things in regards to what attributes are most important, at least if my analysis can be trusted. I was overvaluing LP in my post players and not giving rebounding enough weight. I am going to recruit posts heavily weighted toward athleticism and rebounding from now on. I also found that I need to value passing more in all positions, but ball handling just in the guards and small forward. I think the conventional wisdom is that ball handling is important for the bigs, too, so we will see how that works out for me. Athleticism is king as everyone knows, but can be sacrificed a little bit in the guard positions in order to get more speed, passing, ball handling/perimeter.
That’s about it. If anyone has any specific questions let me know.