Notice that to get a "step" plot it is necessary to assume that probability of getting an A is strongly correlated with the SAT score. The paper says that for h

Notice that to get a "step" plot it is necessary to assume that probability of getting an A is strongly correlated with the SAT score. The paper says that for humanities there is no strong correlation. On the other hand, the "result" for physics is an elementary consequence of repeated trials (binomial distribution)