# How To Make Money From The Close Until Tomorrow’s Open in SPY/S&P 500

This post have some simple suggestions on potential strategies from the close until tomorrows open in SPY (the test period is from October 2005 until present). This post is a summary of all several tests I have done on strategies with entry on close and exit on next day’s open (so far). The best strategy is at the end of this post and one I’ve been trading live for some months.

Here is the first strategy:

- SPY closes at new 20 day low
- Close is above 200 day moving average
- (c-l)/(h-l)<0.5
- Go long at the close, exit at tomorrow’s open

P/L in % | #trades | #wins | Avg. |

8.4 | 35 | 29 | 0.24 |

Thanks to Rob Hanna for this strategy.

The criteria about the close must be above 200 day moving average is important. The equity curve gets a lot worse without it.

My favourite interval is 5 days (in SPY). Here is a new twist:

- SPY closes at new 5 day low
- Close is above 200 day moving average
- (c-l)/(h-l) gives a ratio lower than 0.4
- Go long at the close, exit at tomorrow’s open

P/L in % | #trades | #wins | Avg. |

12.6 | 90 | 64 | 0.14 |

Most trades are below 0.4 (criteria 3) anyway. Here is the equity curve for this strategy:

This is a pretty nice equity curve! Profit factor is 2.6.

What happens if we turn it upside down and go short? It gets a lot more erratic. There is a small negative edge, but not tradeable. However, nice to know, though (not to go long these times).

I also tested this using RSI(5):

- RSI(5) is below 25
- Close is above 200 day moving average
- (c-l)/(h-l) gives a ratio lower than 0.5
- Go long at the close, exit at tomorrow’s open

P/L in % | #trades | #wins | Avg. |

13.57 | 59 | 46 | 0.23 |

The equity curve is just as upward sloping as the one above and has profit factor about 3.8. Here is the annual return trading 490 000 USD per position:

Here is another twist which I belive is the best, considering it has more fills:

- (c-l)/(h-l)<0.5
- RSI(5)<30
- Close is lower than the 50 day moving average

P/L in % | #trades | #wins | Avg. |

36.55 | 168 | 109 | 0.23 |

This is quite impressive. Why? Probably because the market is short term oversold and get a quick bounce up from a bad day. Buying if close is above 50 day average is no good at all using these criterias. Quite a difference from the above strategies that went long if above 200 day moving average. If holding to the close instead of selling at the open, the total P/L increases to 80% but the equity curve is a lot more erratic. So buying on a 5 day low is a lot better for holding several days.

From previous research I know there is a good chance of filling the gap if SPY opens down and is short term oversold. So I put in a new twist: if SPY opens lower (the next day) than -0.1%, simply hold until the close (but exit at the open if opens higher than -0.1%). This boosts the strategy:

P/L in % | #trades | #wins | Avg. |

55.81 | 168 | 129 | 0.33 |

Here is the profit curve of this strategy:

The last years it has paid off to hold this position until next days close even on days when SPY opens up. But this was not the case earlier.

I think the last strategy is pretty good! Combined with others it can boost performance. Sometimes there is overlap with the 5 day low strategy, but not that often.

Hi. Can you kindly send me your email address again? I have misplaced it. Wanted to share a strategy with you mocked up in excel

Hi,

My address is oddmund@quantifiedstrategies.com.

Do you ever utilize any statistical tools for significance of findings? I’m increasingly afraid this way of testing (trying more or less semi-random combinations of indicators w/parameters, until the equity curve “appears good”) is near meaningless. As there are unlimited combinations, any statistical test would be invalid on the strategy results themselves (because you cannot adjust for the (underlying) degrees of freedom), even including testing out-of-sample, CANNOT avoid over fitting randomness, as the underlying possibilites are endless.

No, I don’t do that. There is always a danger for just hitting a random strategy, but to me this make a little sense. Otherwise I wouldn’t have traded it. The returns increases when oversold because of its mean reversion.

This is the way I have traded since 2001 and based basically all of reseach on. I trade many strategies and that (hopefully) should minimize the optimastion effect.

Hi Oddmund,

Nice post. Glad to see some others are embracing the idea of overnight trading!

Rob Hanna

Hey thanks for sharing this pattern. A few questions:

1) Is there economic logic you apply to ensure a high likelihood of this theoretical edge persisting into the future?

2) How would someone trading this know if the theoretical edge erodes, at low cost?

Hi!

1. Economic? I don’t use any fundamental research at all in my trading. I’ve been trading this one for some months and has worked well so far.

2. Looking back, it seems S&P is more mean reversion now than it was in the past. By past I mean prior to 2000. I simply use a cumulative graph of the return. As a rule of thumb I usually start questioning after 6 months with flat or negative returns… I have no other answer/solution as regards this problem. It has worked for me in the past. But I always trade double digit number of strategies.

Great job Oddmund! Have you tried to extend the timeframe back to the issuing of SPY? And maybe would be great seeing them joyned all together in only one backtest. I think results will be interesting. I’m doing a similar job at my blog and with my trading although I’m focusing on much more specific bar patterns.

Hi, thanks for the nice comment. I’ll have a look at your blog later. You’re mainly trading similar patterns with seasonality?