Conditional sampling methods are generalized by applying a matched-filter on the reference signal. This generalized method is illustrated by data obtained in a magnetized toroidal plasma without rotational transform. By proper choice of the filter characteristics we can observe a rapid ‘crash’ in the electrostatic plasma potential in the entire plasma column. The crash is then followed by a slower recovery phase. The results are noticeably different from those found by using a more traditional conditional analysis of the same data. The significant differences between the two results indicate that generalizations of the conditional sampling method can give new insights. Conditionally obtained local power spectra demonstrate an enhancement in the amplitude of the fluctuations preceding the potential crash. The analysis is also illustrated by using a synthetic dataset.