Very true. For that reason I would like to see this done with highest resolution data as possible. Also the ability to tweak the entry and exit times to see how it affects the results. One can trade shortly after the open or shortly before the close and get tightest possible spreads.
Lol, it's all good. I know how to compartmentalize things. We can have a tiff here and we are still cool as cucumbers in my book. I seem to be about as popular as a root canal with some here. That's good, I must be doing something right.
I hope that is where this leads. We shall see. Arbitrarily entering and exiting the market everyday won't work, I already know that, and have never said it would work. But, this is at least a starting point where a strategy with a good edge that can exploit this behavior perhaps can be developed.
I've a mind to try it again, but after the last three peaks I cannot stomach another go now. I just can't and it is burning me up because of the potential loss of not being in it. But I gotta' covet prudence and stay low now. I have been getting bitten on each new peak. Good luck there, Krug. I was supposed to end this with appropriate or soothing music. I clicked enter prematurely. Mmm
Alright folks, here is the first pass at my own analysis. The data set I have to work with is 15 minute OHLC bars for the past 6 years and 4 months (1/1/2013 to present). The study enters 15 minutes after the open and closes 15 minutes before the close. Presented is the PL curve, trade metrics and trade execution sheet. Market Hours Only Overnight Only
Thanks for the spreadsheet. I think column M should be simply column K minus column J. Otherwise you aren't accounting for dividends, which distort overnight returns on ex-dates. After making that change I get the following numbers: Sharpe 1.02 Daily Vol 0.619% Median 0.058% Mean 0.039% Min -8.688% Max 5.952% % Up 56.366% Total 2.59667 Worst DD -39.733% T-stat 5.17 That 2.6 total equates to 1200%+ total return. Last 5 years looks better, with a Sharpe of 1.25. Also I think Sig's perception of greater overnight vol might be due to the fatter tails in the overnight (vs. day) distribution. Excel's kurt function returns 19.46 for overnight against only 9.25 for day. Not a perfect measure of tailweight, but still...
One thing I find fascinating is how during the 20% selloff last December, it was barely a blip with the overnight only test.
I literally spent 10 min on it did it because I could not fall asleep until 3 AM. Gonna take some ambien tonight once Asia closes. Just a thought on market correlation and isolating this alpha. Given that it's a global phenomenon, I think the "right" theoretical strategy would be to do c2o longs vs o2c shorts by timezone (e.g. go long spooz at the US close, short nikkei at Tokio open, flip to long, short stoxx at Europe open, etc). That still would not get around the low pnl/tradeval (in fact, it would lower it since it would trade double now), but it would make it market neutral.
That is of interest. I would suggest to try and isolate when/where profits are coming from, try opening the trade at 3:00am NY time (European Open) and compare it to opening the trade at US market close. You might find that the US close to the EU open is similar to the US RTH. At least you will find out if profits are derived from US close to EU open or EU open to US open or equally dispersed across the original trade time.