It's a while since I've seen it tested but if memory serves it at least used to be profitable to sell the (regular session) open S&P500 futs and buy the close every day - probably not very much so after transactions and slippage but the point stands: all the gains come over night, the day session is net negative. This may no longer hold but it was true maybe 10-ish years ago?
%% True; but as the Stock Trader Almanac[Hirsch]notes, its still an upward bias, like JAN. BUT DOW has a tendency to go down more in JAN; dont know what its doing now, SDOW was up nicely this week- not likely it will be up like in 2018, due to DOW downtrends. Even something like SPY+ related =up 70%of the time past 10 years, [JAN]was nicely down 30%[3 Jans out of 10 JANS] YHOO data is not a big problem; just be sure to have 1 or 2 other reliable sources, especially charts!!,,50 years is plenty of data; but look @ all data, because you[any ] can't predict. They call it weather forecasts, not weather prediction.
You can start by looking at an index, but then you must test on something tradeable, such as SPY or S&P futures.
I think your analysis fails to account for the full concept here. This appears to calculate short-term momentum. It's an interesting concept - an up day leads to a greater likelihood of a following up day. Market psychology would definitely allow this possibility. Do short-term uptrends tend to last several days essentially? Which would mean short-term downtrends last for several days, I suppose, if the 50-50 distribution is indeed accurate. It seems quite possible that, especially after a reversal, the movement will statistically continue for longer than a single day, which should apply to either direction theoretically. I like it!
"Globex opens on Sunday at 5 and closes on Friday at 3:15." Can't you just trade futures after the market closes?