A cool lemma in Ramsey theory is the following:
$$R(m, n) ≤ R(m-1, n) + R(m, n-1).$$
It is not hard to prove it. However, a friend of mine said that, if $R(m-1, n)$ and $R(m, n-1)$ are both even, then
$$R(m, n) ≤ R(m-1, n) + R(m, n-1)-1.$$
How can I prove it?
Consider that you have a graph of $R(m-1,n)+R(m,n-1)-1$ points. And you want a clique or an independent set.
Now take any point if it has an edge with $\geq R(m-1,n)$ points then by induction you are done. If it has an edge with $\leq R(m-1,n)-2$ points then it will have a non edge with $\geq R(m,n-1)$ points and again you are done. So one is left with the case where the chosen point has exactly $R(m-1,n)-1$ edges.
In the language of graph theory it has degree $R(m-1,n)-1$. Now you are able to chose this beginning point at will, thus if at least one point does not have degree $R(m-1,n)-1$, you are done.
Finally the case where all points have degree $R(m-1,n)-1$ is impossible since $$2(\text{# of edges})=(\text{# of vertices}) \cdot \text{degree}$$ $$=(R(m-1,n)+R(m,n-1)-1)(R(m-1,n)-1)$$ and under your assumptions this last number is odd so it cannot be $2(\text{# of edges})$ thus at leas one point does not have degree $R(m-1,n)-1$.
An example of this is $R(3,4)=9$.
But $R(2,4)+R(3,3)=4+6=10$.