From taking a quick look at page 8 above (numbered 104 in the document), if gRR used X instead of OLC(X), it would result in something like the above article. The point is in having consistent probability distribution over held beliefs, not over some infinite closure of beliefs.
From taking a quick look at page 8 above (numbered 104 in the document), if gRR used X instead of OLC(X), it would result in something like the above article. The point is in having consistent probability distribution over held beliefs, not over some infinite closure of beliefs.