Shape Detection Using Hough Transform in Matlab

Question

I'm using a Properly working Matlab code (The original code is from here) that uses Hough trnsform to detect basic shapes like round, square and triangle.Here below is the important code segment.

[H, theta,rho]=hough(S);

Above H is the Hough transform matrix and S is the Black and White image of the shape.

for cnt = 1:max(max(H))
    data(cnt) = sum(sum(H == cnt));
end

Here is the shape detection part.

[maxval,maxind] = max(data);
medval = median(data);

[p]=polyfit(1:maxind-5,data(1:maxind-5),2);

if maxval<3*medval
    set(handles.txtResult,'string','Triangle');
elseif  p(3)>100
    set(handles.txtResult,'string','Square');
else
    set(handles.txtResult,'string','Round'); 
end

I can understand the "data"(which stores Hough Matrix intensity frequencies). I just can't understand the logic it uses to detect the shape. maxval<3*medval and p(3)>100