With 1. cxd4 losing to Nxb5, the first move has to be 1. b6, and the theme must be connected with the threat of promoting the pawn either on a8 or b8. Since 1 . . . Nc8 loses to 2. b7, Black must get his rook into position to stop the pawn, so it has to be 1 . . . Rd1+ 2. Kf2 (we reserve the right to change this to Kg2 if it improves matters later) Ra1]
Only with this cumbersome move can Black stop the pawn: instead Rb1 would lose to bxa7, while Bb3 or Bb1 lose to b7. (We now see 2. Kf2 was right: on g2 the king would have been vulnerable to a bishop check after Bb1.)
Now if it were Black's turn to play, he would be very short of moves. Any of Bb3, Bb1 or Nb5 would block the b-file and lose to b7. So White plays 3. h4] to begin a phase of play in which each side tries to run the other out of moves: 3 . . . e4 4. Kg2 (4. Ke3 Rc1] 5. Kd2 Ra1 surprisingly gets nowhere) e3 5. Kg3] and, with Rc1 no longer threatening a check on c3, Black can only resort to a series of rook checks.
5 . . . Rg1+ 6. Kf4 Rf1+ 7. Kg5 Rg1+ 8. Kh6 (safe at last) Ra1 and now comes the point: 9. h3]] and White wins. Any rook move loses to bxa7, bishop moves lose to b7, Nb5 and Nc8 lose to b7 and Nc6 loses to dxc6.
Bravo] Well, almost. There is something wrong, and to discover what it is we must return to the starting position. White still has all eight of his pawns, but how did they get to their present squares? Everything has drifted eastwards.
The original g-pawn must have captured on h3, so the pawns currently on g6 and g7 must have begun life on f2 and d2, if not even further away. Adding it all up, we realise that to reach their current formation, White must have made at least eight captures, which is unfortunate since Black has nine men still on the board, making a total of only seven men captured. So the position is illegal and the whole study must be abandoned.
Sadly, Herbstman never did manage to set his elegant idea in a position that could not be convicted of breaking the law.