We systematically evaluated the binding energies of d(10), d(8), and d(6) transition-metal complexes with various pi-conjugate systems such as Pt(PH3)(2){C2H4-n(CH=CH2)(n)}, Pd(PH3)(2){C2H4-n(CH=CH2)(n)}, [PtCl3{C2H4-n-(CH=CH2)(n)}](-), [PdCl3{C2H4-n(CH=CH2)(n)}](-), and [PtCl5{C2H4-n(CH=CH2)(n)}](-) (n = 0-4) using the MP2 to MP4, CCSD(T), and density functional theory (DFT) methods. The MP4(SDQ) and CCSD(T) methods present a reliable binding energy, whereas the DFT method significantly underestimates the binding energy when the size of the pi-conjugate system is large. The underestimation occurs independently of the coordinate bonding nature; the pi-back-donation is stronger than the sigma-donation in the Pt(0) complexes, as expected, they are comparable in the Pt(II) complexes, and only the sigma-donation participates in the coordinate bond of the Pt(IV) complexes. The DFT method provides moderately stronger charge-transfer (CT) interaction than the MP4(SDQ) method, suggesting that the underestimation of the binding energy by the DFT method does not arise from the insufficient description of the CT interaction. From theoretical investigation of several model systems, it is concluded that the underestimation arises from the insufficient description of electron correlation effects.