In this talk, we will discuss an ENO type polynomial reconstruction for high order finite volume methods solving hyperbolic conservation laws. Adaptive stencil choosing process was adopted in the traditional ENO method based on calculation and comparison of local divided differences of cell average values for high order finite volume methods. The computed divided differences measure the relative smoothness of each candidate stencil. One can then decide which is the smoothest stencil to choose for polynomial reconstruction. The new stencil selection strategy proposed here will rely on an approximate measurement of local variation computed from the local cell average values and the corresponding reconstructed polynomial values at interfaces. The main motivation is that the reliance on the divided difference of traditional ENO methods makes it very difficult to generalize to multidimensional problems on unstructured meshes. We also show that when the selection is biased toward a central stencil, the loss of accuracy around extrema by the traditional ENO method can be avoided