tconte
13 posts
msg #107572
- Ignore tconte |
8/13/2012 11:03:44 AM
Also bjfeen, your signals don't seem to match the ones from StockFetcher. For instance on Aug26 2008, 999 days back, which is as far back as the system can go I think, SF shows SSO long with these readings:
SDS %D(5,1,5) 59.72 %D(5,1,10) 58.64
SSO %D(5,1,5) 37.44 %D(5,1,10) 39.21
and on Aug27 2008, we have the signal to switch and go long SDS with these readings:
SDS %D(5,1,5) 52.76 %D(5,1,10) 56.65
SSO %D(5,1,5) 43.34 %D(5,1,10) 40.98
Just wondering why your %D calculations are different
Also, your spreadsheet has the buy SDS signal on Aug26 2008 and the sell SSO signal on Aug27 2008, which means that you have probably calculated holding both positions for 1 day, but the principle of the strategy is to hold only one position at a time. Also your %running gain or loss is from the close to the close, but the principle of the strategy is to start from the open of the next day, and also the way that you are adding percentages at the end is not mathematically correct.
Just giving you a few pointers for accuracy.
|
bjfeen
24 posts
msg #107573
- Ignore bjfeen |
8/13/2012 11:26:06 AM
I'm not sure why it is different, %K(5) is calculated using highest high and lowest low of the last 5 days and the closing value. %D(5,1,5) is the average %K(5) for the last 5 days and %D(5,1,10) is the average %K(5) for the last 10 days. It's a work in progress. I have made the spreadsheet editable for anyone to play with if they want and I'll keep a copy of the master. I'll keep updating the master as people find bugs or issues with it...
|
jackmack
334 posts
msg #107574
- Ignore jackmack |
8/13/2012 11:37:20 AM
Kevin
Not looking to modify this in any way just wondering if I wanted to add an addition filter to this as a smoothing mechanism using the 30 minute SPY chart and an EMA (4) and EMA (24) cross (above and below - for long or short) what then would be the best stochastic settings?
Thank you
|
tconte
13 posts
msg #107575
- Ignore tconte |
8/13/2012 11:37:42 AM
Here is a link to my own manual backtesting:
https://docs.google.com/spreadsheet/ccc?key=0Ah-vGIsKIZandG1rYlE5SWdCWUoyZVNod3ZQSFJXN0E
|
novacane32000
331 posts
msg #107576
- Ignore novacane32000 |
8/13/2012 11:42:21 AM
tconte,cannot access your link w/o a password.
|
tconte
13 posts
msg #107577
- Ignore tconte |
8/13/2012 12:28:33 PM
Hi bjfeen,
I think it's in your "Min Low(5)" column "I" formula. I checked it for SDS on Sep15,2008, and the minimum low should be $66.42, but your spreadsheet shows $69,50.
|
tconte
13 posts
msg #107578
- Ignore tconte |
8/13/2012 12:35:31 PM
novacane, sorry about that, I changed the settings of the file now, and anyone with the link can access it without a password. Please try it and let me know if you have been able to access it. Thanks.
|
novacane32000
331 posts
msg #107581
- Ignore novacane32000 |
8/13/2012 1:40:41 PM
tconte.I can see it now.Thanks for sharing your work.
|
blumberg
27 posts
msg #107582
- Ignore blumberg
modified |
8/13/2012 2:13:04 PM
tconte: It looks like your backtest produced much lower returns than Kevin's as well. Must be bc didn't include stops?
|
tconte
13 posts
msg #107584
- Ignore tconte |
8/13/2012 3:35:35 PM
thanks novacane for your feedback.
blumberg, the capital appreciation is different because of the timeframe. I could not get more than 999 days back from SF, which represents about 4 years of data. Kevin started in Jan 2007 so that's 1 and a half year more for a total of 5 and a half years approximately. Nevertheless, my results look identical on the chart I just finished producing, especially the period from Sep2009 to Sep 2011, which shows a decline and a come back to pre-decline levels. When you think about it, it took 2 years of trading this system to recuperate the money lost during the Sep-Oct 2009 decline. I don't know about you, but I think I would have given up . I don't know what happenned in Sep-Oct 2009, it was just a blimp in the uptrend of the S&P. And in the 2 years afterwards, there were 2 consolidation periods (Jan-Aug 2010, and Mar-Jul 2011) where you think the system would have performed well being an oscillator type. If there was a way to identify and correct the system to prevent it, I think I would be more confortable trading it.
|