Opponent's all time W/L ratio, the original rating system ratio, or the season W/L ratio? Either way, it sounds like it would have a similar outcome to the system I made, but that the overall points have potential to become a lot more exaggerated.

Let N be total number of people who have sparred during the season.
Let R be the current rank position of the defeated player compared to all the other people.
Let S be (N+1)/R; this provides the relative measure of skill for the defeated player.
Let P be 100*ln(S); the natural logarithm provides diminishing returns on the points gained due to the opponents' measure of skill; the multiplier of 100 (arbitrary) increases the spread of possible point gains.

The people you'd expect to win spars were ranking high. Why I consider it flawed: people with a higher rank can fight one another almost exclusively and boost off each other. Instead, if S were a measure of the comparative level of skill between the defeated opponent and the spar winner, it would make a lot more sense.

Aside from that there could be scalability issues in relying on overall rank, which essentially means maintaining an ordered list of all of the sparrers, as opposed to only the top sparrers. Maintaining a full list can become computationally expensive as it grows.

The system you described can be gamed in the following way: Spar a bunch of newbs to get your seasonal ratio high, then exclusively spar with other higher-ranked players. You'll get diminishing returns the more you fight each other, but if each players' ratio is boosted then each one will get a higher number of points by sparring amongst themselves as compared to the general population.