I just looked to verify my memory. The first few top quark papers were base on low two-digit numbers of events and error bars were on the order of +/-10% ... though their “certainty that this was not random” was near one-in-a-million by the time of the “solid” dual announcements by CDF and D0, the two FermiLab collaborations in 1995.
Still looking for the paper here.
Houses of cards in our locality.