1) My earlier regressions were full of artefacts. Much fewer position/skill-pairs are used in each team rating than I used to think. The idea that there is a fully spelled out position/skill matrix for each offensive and defensive tactic, using many different skills for each of the ratings, was almost definitely wrong. BB-Forrest invested only half an hour after all ;-)
2) Usually there'll be a set of position/skill-coefficients for each team rating. That set will be relatively simple. I've calculated some new approximations, but I still don't disclose them. There are now 192 games in the database, a minimum of 14 for each tactic, and at least 2 games with each of the 40 possible combinations of tactics.
3) The contribution of each position to each rating will be modified by a tactic-dependent multiplier. I'm currently not sure whether the relative weight of each skill used at a certain position as compared to the other skills used at that position is really also tactic-dependent. The new regressions are not conclusive on this.
4) A larger role than I thought earlier is played by the energy-characteristics of each tactic. For example the energy-characteristics of a defensive tactic (fast-slow etc.) will have an impact on offensive team ratings and vice versa. These energy effects cannot be derived fully from the boxscores, since it will make a difference whether a player plays 36 minutes in one go or whether he plays 18 minutes, rests 12, and then plays another 18. Building a better model of the energy effects of the stamina skill is now the main challenge for future work.
5) Team ratings are not used by the game engine code, and they really only tell a very incomplete story. For instance some skills seem to be absent from the computation of all six team ratings, which doesn't mean that they won't have an impact in actual games.
Point 2 ,3 are known as principle,but a more deep research about it should be great,so we can batter explain some evaluations in strange matches
Point 4 should be analyzed compared with the stamina level of the players,but these confirms some theories about the importance of have a good bench(when one player is tired a good bench consent to rest the starter)or to have a great level of stamina for starters
Point 1 e 5 iIagree,the Free throws skill for example can give many points,but the team reatings doens't count it in the game