1 indeed, statistically speaking, that score would be poorly distributed (massed around 0 and some few values above).
2 what would happen if a player wins more than half of his games vs borg ? Feldfeld, Val, Aristol .. ? 1 game is enough is to be 100% He would be the new benchmark ?
3 this score is calibrated vs 1 player and not all of them. So the gaming style matters and borg can decide to improve your score or even refuse to play vs you
To stick to your idea, better take a well rounded player like ffm as benchmark and make this
ln (Number Wins / Number Fails)
So it will gravitate around 0 for those with same level than ffm ; positive for those who won most games ; negative for those who lost most and no DIV/0 or infinite problems. The score distribution will look better