Can we say if MAPE is greater than X the forecast is not accurate? Such cut-off points exist for example for the percent relative standard error (PRSE). If a parameter has PRSE>30% it is usually declared unreliable.
No, there is no such cut-off value for MAPE. The reason is the fact that MAPE is leveled (translating the time series, to be forecast, up on the y axis will diminish the value of MAPE but not your forecast accuracy). One can use MASE (Mean Absolute Scaled Error) introduced by Rob J Hyndman, which has such a value, 1. That mean, if the value of MASE