I had a thread on the old EasternUsWx forum titled "The October Snowfall Myth" and it went through the statistics of it.
The biggest reason that myth gets perpetuated is because BOS (and HFD/BDL is similar too) has like a sample size of 4 Octobers in the last 50 years with measurable snow and they mostly sucked (2009, 2005, and 2011). You could include 1979 in there even though BOS got a trace, but much of the city had measurable. Anyone with a rudimentary understanding of statistics would tell you a sample size of 3 or 4 is worthless in this context. But when we expand the sample to interior sites like ORH where we have double digit October snowfalls, the mean snowfall following years with measurable October snow increases to right around the long term mean of all years. (I.E., it has no correlation) Blockbuster years that had measurable October snowfall at ORH include 2002-2003, 2000-2001, 1960-1961, and 1961-1962.