I should clarify that the NDX "on average" during those years (since 1986) tops out within 8 days plus or minus 6 days.
That's equivalent to saying the NDX "on average" has a local top at least once a month. Which, of course, is a true statement for *every* month, not just January. By that logic there is never a reason to go long.
More data...
1993: NDX higher following the 8-day Jan window 1994: ditto 1995: ditto 1996: ditto 1997: ditto 1998: ditto 1999: ditto 2000: ditto 2001: ditto 2002: Ah, here's one, actually topped in early Jan this time 2003: And a second one... 2004: A wash... 2005: Nice decline on that one
From this its clear the odds of hitting new highs late in January is highly correlated to being (or not) in a bull market.