I base the analysis off of the overall team performance which has some flaws to it. Putting out 3rd best lineup against a strong team while on the road, the AI tends to sub in lesser players with a big lead, the losing team tends to fire up 3pt shots very frequently in the 4th...these things may lead away from a team's strong point and lend itself to skewing the data.
With as many data points as we have, it still tends to only skew the data mildly. Plus, lets be honest, most of the time the way that enthusiasm affects games and how we respond to good teams/home court advantage...the overall data bias is towards the ends. In other words, the best teams are probably not quite as good as the stats show and the worst teams probably aren't as bad, but the middle couple or 3 teams are what the data says they are.
I have the means...its sloppy so I don't do it...to get the specific data for every game. I suspect that I'd be able to give an even better analysis if I gathered that data. I broke down my team last season by game and one interesting thing came out of it. My defense was topnotch in d3 last season, but the best offense still shot 48%. My takeaway was: Individual matchups can significantly impact a game. The average shooting percentage against me last season from partway through the season was 32.9% ovr, 16.35% 3pt. Most games were around the 30% range.
This type of information broken out across the league with some adjustments for "expected performance" would probably make for an interesting read. A little more complicated and a good bit more time unless I could get the data into a spreadsheet format more quickly.
Maybe someday on that analysis...as for player performances...without a pbp or a better breakout of sub information...its unlikely I'll ever even consider working backwards from there.