Sure just buy Monday morning, and will compare in a couple of years. Your buying after a 10 year bull market and SPY up almost 20% YTD. I guess if you have a decade or two your buy and hold may work.
I think the point they are making is what I have been saying all along. Buying at cash close and selling at cash open the next day works great in a 10 year bull market. Then you get a blip like, I dunno', Feb 2018, Oct 2018, and May 2019. That kinda' messes up yer mojo. And what happens when we get a 5 year stair-step down 30% or 40% or whatever? And you never answered the question...Where'd you vanish to after Oct 2018 reared it's ugly head? It is not risk free, man.
This is really interesting. Your backtesting seems to show the exact opposite results of what the article claims. So in basic theory, the inverse operation should be successful. You should be able to see large gains by buying at the open and selling at the close everyday. Did you use regular market hours in your backtesting? (Premarket & after hours may skew the results.)
From 2005 until today (12th of July 2019) it looks like this (3453 full trading sessions): The Open is simply the difference between the cash close at 16:00 and the open the next day at 09:30. ES futures. So, yes, unsurprisingly there is a skew towards a positive open price, but less than you might expect. 45 % of days did in fact open negatively. I also did a check on the other claim from the article and posted my findings in another thread a little while back. I found that my results did not equal the claims from the article. Adding in a filter, i.e., buying the cash close after a strong down day might yield better returns. Just buying the cash close and selling the open does not seem like a robust strategy to me. I certainly wouldn't trade it.
Your backtesting seems to show the exact opposite results of what the article claims. Using yahoo data adjusted for dividends and splits but ignoring transaction costs, buying one share of SPY at the open (regular trading hours) and selling at the close (regular trading hours) from January 29, 1993 through July 11, 2019 results in: numValues 6660 sum -0.9536529999998 min -10.092755 max 10.116016 mean -0.000143191141141111 sampleStdDev 1.02499980228868 median 0.0298219999999993 medianAbsDev 0.417475499999998 skewness -0.538117835395494 excessKurtosis 11.2369003505817 >Thresh_0_Pct 51.79 This more-or-less agrees with the original article's "not counting dividends ... buying the E.T.F. at the first second of regular trading every morning at 9:30 a.m. and selling at the 4 p.m. close ... down 4.4 percent since 1993" because the data I used was adjusted for dividends. So I think the original article's premise is correct if there are no transaction costs.
The buy the close sell the open does not work everyday. I know of no system that does. And yes you should use stops.