I was puzzling over this a bit at Tyler Dellow's blog, and I don't think I've seen the answer anywhere. Feel free to chime in if you have a different calculation. Here's the process I followed: for every RFA and UFA who signed in the 2009 off-season, I added up their 2009-10 salaries and their projected win values from Tom Awad's VUKOTA projections. I then subtracted the league minimum salary from each player's total salary (VUKOTA is already with respect to replacement value) to determine the marginal cost per marginal win.
The results:
| # | Salary | Marg. Sal. | Wins | $M/Win | |
| RFA | 70 | 89.9 | 54.9 | 41.3 | 1.33 |
| UFA | 106 | 216.9 | 163.9 | 73.4 | 2.23 |
| Total | 176 | 306.8 | 218.8 | 114.7 | 1.91 |
So an RFA takes a 40% discount relative to a UFA. I think that's actually very close to the discount that baseball players take in their arbitration years.
At any rate, I'm interested in hearing differing takes on this question.