simardcasanova’s Twitter Archive—№ 19,942

@andreamatranga I do have a lot of zeros on some datasets (stored as checks that the simulation runs as expected to), I guess it explains why the gains are so huge (they are more limited with another dataset that has less zeros)