All aspects of the MAGIC performance (including tabularized energy-wise performance numbers) are described in detail in Aleksić et al., Astroparticle Physics Volume 72, 1 January 2016, Pages 76–94. The performance shown here corresponds to data taken in 2013.
We define the sensitivity as the integrated flux of a source above a given energy for which the number of excess events (N_excess) divided by the square root of the number of background events (N_bkg) is equal to 5 after 50 hours of effective observation time. In addition, we require that N_excess > 10 and N_excess > 0.05 N_bkg. The corresponding energy threshold is computed as the peak of the true energy distribution of a Monte Carlo sample with a Crab-like spectrum. Integral sensitivity computed according to different definitions and tabularized numbers are available in Aleksić et al. (2016). All the curves are computed using samples of Crab Nebula data. For example, the sensitivity computed according to Li&Ma significance with 5 background regions is (0.67+/-0.04)% C.U. above 290 GeV and (1.45+/-0.02)% C.U. above 100 GeV for 50 hours of observations.
Even while 50 GeV is usually below the analysis threshold of the MAGIC telescopes (for a standard stereo trigger) for a Crab-like spectrum, it is still possible to perform scientific observations with the MAGIC telescopes at those energies. Due to unfolding corrections for the energy resolution and bias it is also possible to reconstruct reliable spectra at energies not so much below the energy threshold.
Moreover, for steep spectra sources (far AGNs, GRBs, pulsars) the peak energy shifts to lower values. For example, a strong 25 sigma detection of the z=0.939 flat-spectrum radio quasar PKS 1441+25 corresponds to 5% Crab flux sensitivity above 55 GeV for 50 hours of observations.