I do not think there's anything particularly theoretically special about claiming that a racket with MgR/I = 21 is 'the best'.

If I follow travlerajm posts properly he just observed statistical correlation between player's rank and player's rackets MgR/I value, and from that he therefore deducted that 21 is the optimal value.

Now, it is interesting but in my opinion unfortunately a bit far fetched.

Per his own data the accuracy of MgR/I needs to be calculated with 0.1 precision. that means to back up his claims with empirical data (that is player's rackets specs) that data must be pretty darn accurate. However per his own admission he just estimated the mass and swing weight of the player's rackets based on available specs for mass and balance for unstrung rackets (see

http://tt.tennis-warehouse.com/showthread.php?t=387620). Is that possible - well, estimation is, but you can't really use the result of the approximation in cases where the very exact number is needed. Or in other words - the formula used to calculate swing weight based on mass and balance of the racket, and assumption about racket's weight distribution, developed by Cross (I think), while indeed gives pretty decent result as far as swingweight is concerned, is by no means an absolute theoretical truth in the sense for example the parallel axim theorem is.

Now if you add to this the fact that you are supposed to adjust the value if you wear a sweatband - than I think we are leaving the realm of physics....