Title: A novel energy reconstruction method for the MAGIC stereoscopic observation

Abstract: We present a new gamma ray energy reconstruction method based on Random Forest to be commonly used for the data analysis of the MAGIC Telescopes, a system of two Imaging Atmospheric Cherenkov Telescopes.
The energy resolution with the new energy reconstruction improves compared to the one obtained with the LUTs method. For standard observations i.e. dark conditions with pointing zenith (Zd) less than 35 deg for a point-like source, the energy resolution goes from $\sim 20\%$ at 100 GeV to $\sim 10\%$ at a few TeV.
In addition, the new method suppresses the outlier population in the energy error distribution, which is thus better described by a Gaussian distribution. The new energy reconstruction method enhances the reliability especially for the sources with steep spectra, in higher energies and/or in observations at higher Zd pointings.
We validate the new method in different ways and demonstrate some cases of its remarkable benefit in spectral analysis with simulated observation data.