Erdős introduced the quantity \(S=T\sum ^T_{i=1}|X_i|\) , where \(X_1,\dots , X_T\) are arithmetic progressions that cover the squares up to N. He conjectured that S is close to N, i.e. the square numbers cannot be covered “economically” by arithmetic progressions. Sárközy confirmed this conjecture and proved that \(S\ge cN/\log ^2N\) . In this paper we extend this to shrinking polynomials and so-called \(\{X_i\}\) quasi progressions.