I wonder how they caught this guy in the first place—perhaps giving the help exclusively to black kids and ignoring the rest?
No. The full methodology that led to targeting specific Georgia school districts, specific schools, and specific classrooms (i.e., teachers) was in the report prepared for the Georgia governor.
In brief, the investigators looked at the number of erasures on the standardized test sheets to see what percentage of the erasures changed wrong answers to right answers, then compared that to the expected number based on a large number of criteria.
Schools with a large number of classrooms that fell a large number of standard deviations from the norm (classrooms with results that statistically just wouldn't have happened randomly except once in hundreds of trillions of tries) were identified.
In some Atlanta Public School System schools, up to 80% of the classrooms experienced wrong-to-right erasures that FOR ONE CLASSROOM would happen by chance less than one time in 700 trillion times, yet 80% of the classrooms experienced those wrong-to-right erasure rates, or higher.
It wasn't by race. You just looked the answer sheets for a teacher's studnets and noticed that everything had been erased and changed to the right answer.