I am reading Zee's Quantum Field Theory in a Nutshell. In the section about Grassmann numbers, there is an identity:$$\int dx\int dy\,e^{yAx}=\det A\tag{II.5.13}$$ where $x=(x_1,x_2,\dots,x_N),y=(y_1,y_2,\dots,y_N)$ are $N$ Grassmann numbers and $A$ is an antisymmetric $N$ by $N$ matrix. If $A$ is an arbitary matrix (not necessarily antisymmetric), does $(II.5.13)$ still hold? When I try to prove $(II.5.13)$ for an arbitrary $A$, I encounter a problem. To save writing, I only show the case for $N=2$.
For $N=2$ and arbitrary $A$, $$\int dx\int dy\,e^{yAx}=\int dx\int dy\,(y_1A_{11}x_1y_2A_{22}x_2+y_1A_{12}x_2y_2A_{21}x_1)=\int dx\int dy\,(-y_1A_{11}y_2x_1A_{22}x_2-y_1A_{12}y_2x_2A_{21}x_1)=-A_{11}A_{22}+A_{12}A_{21}=-\det A.\tag{2}$$ The equation $(2)$ seems wrong since there is an additonal minus sign even for an antisymmetric $A$. Where is the problem in my proof?