Look at this graph, which shows the returns of a strategy that had simply bought SP500 after the index rose the previous day (dividends not included, nor transaction costs), and stayed in cash otherwise (red line; blue line did the opposite): As you can see, it has been historically quite profitable. However, since around year 2000, the strategy not only stopped making a lot, but it even became unprofitable. Is this just a (very long) bad run and we should expect returns similar to 1950 to 2019 to repeat? Are actually 2000 to 2019 the kind of returns we can expect in the future? I'm amazed at how big is the difference in returns between one period of time and the other, it really makes you tremendiously skeptical of backtests over small sample sizes. Even 69 years (17371 market days) doesn't seem enough to know what can we expect of a strategy in the future...
Bought the "S&P 500" - LOL. No - you didn't! (or do you have specific stock weightings?) Next thing, you're going to tell us you bought on the open of an index. Which open is that? One you could actually participate in? You also fail to tell us your exit strategy. PS - You're appear to be using Yahoo data. Not a trustworthy source.
This uses Yahoo Finance's GSPC. What is wrong with that? You could implement this strategy just using futures, there's no need to do anything involving stocks. This uses close prices, not open ones. The strategy is very simple: If today's close is higher or equal than yesterday's close, buy/hold SP500. Otherwise, sell/stay in cash.
Its odd that the system wins and then fails. Certainly worth a little study to understand what changed. On the other hand, this is a system based on a fallacy anyway. While its a fact that stock indices (not stock index members) have an upward buoyancy due to membership revisions, that doesn't extend downwards to the individual daily performance level. Even in consistent and long-running uptrends, the ratio of up days to down days is not much different from 50-50.
I wonder if something really changed (if so, why?) or if it's just a bad run (although quite long). It's still a great system if you just check the start and end point, and spectacular from 1950 to 2000. I'm adding buy and hold too to compare:
That would make sense if investors in more recent times have shorter holding targets. I suspect this is true, given the expansion of share trading, the increase of share trading for price appreciation as opposed to share-holding for dividends, lower transaction costs, greater and faster information flow etc. But none of these factors should negate the long-term upward buoyancy of the stock index value.
My guess would be the public/dumb money used to trade from day old newspaper headlines or from TV news after work when the markets had closed. Then the Web gave every one free and easy access to instant stock news. Starting from around 1999 people with a PC based desk job got instant stock quotes & news at work, and you could place trades instantly as well. This would mean there is now much less follow through buying on the next day.
It's surprising that not only returns from 2000 stopped being way better than buy and hold, but they actually become lower (even negative). If the previous situation was somehow arbitraged away because of fundamental changes like the ones you commented, I'd expect this system to be about as good as buy and hold, not worse (that may be variance, though, although 19 years is already a considerable sample size).
I think it is surprising that it ever worked so well in the first place. It looks too good. So my first thought would be to be very suspicious of the early data. eg of it not taking into account any opening gaps properly on the next day.