More on APS and out_of_sample testing

Quote from ronblack:

Thanks for the correction of the pseudocode.

I have implemented this approach to results I get from APS and in certain cases I get significant improvement in performance. I appears that some patterns are more suitable as entry signals and others as confirmation signals. This is part of the idea behind this approach.
More...

Do you get better results with both daily and intraday patterns?

By the way, I figured out a cool way of doing what you suggested using a genetic program. The algorithm will take the APS patterns and let evolution determine the best system.
 
Quote from ronblack:

Thanks for the correction of the pseudocode.

I have implemented this approach to results I get from APS and in certain cases I get significant improvement in performance. I appears that some patterns are more suitable as entry signals and others as confirmation signals. This is part of the idea behind this approach.
More...

No problem. This is an excellent idea. I am working on it and I will let you know of the results. I want to test it on 15-minute intraday patterns. I am currently running an APS search on 100,000 bars long history for high profitability QQQQ patterns and then I will test the results in tradestation.
 
Quote from intradaybill:

No problem. This is an excellent idea. I am working on it and I will let you know of the results. I want to test it on 15-minute intraday patterns. I am currently running an APS search on 100,000 bars long history for high profitability QQQQ patterns and then I will test the results in tradestation.
More...

Amibroker seems to be a better platform for APS pattern testing since the program can do the AFL for the conditions. It is also a more accurate backtest I believe and Ami is inexpensive. You can also do portfolio level testing with different patterns for different markets, something that cannot be done easily in TS.
 
I like to keep up with the work I do and I continued forward testing the APS patterns since the time I posted the first results here:

http://www.elitetrader.com/vb/showthread.php?s=&postid=2834308#post2834308

From beginning of this year and up to 07/30/2010, the QQQQ APS patterns with 1% target and 2% stop have generated 20 trades, out of which 16 were were profitable. Win rate = 80% and target/stop = 1/2. This is roughly a profit factor of 2.0 (losses = 2 x wins). These are the APS out_of_sample testing results from 01/01/2010 - 07/30/2010

http://www.4shared.com/photo/qN6ye1gm/QQQQ_T1S2_20100730.html

Next I have results for target 2% and stop 4%. Now the win rate is 73% and the target/stop again 1/2. This is roughly a profit factor of 1.35. These are the APS out_of_sample testing results from 01/01/2010 - 07/30/2010

http://www.4shared.com/photo/OasR7E7P/QQQQ_T2S4_20100730.html

I find the results exceptional for this kind of market. Results for SPY turn out similar. I will post new results after year end. This is interesting stuff. By the way, the APS product was terminated by its developer but there is something based on it
here.
 
Quote from goodgoing:

From beginning of this year and up to 07/30/2010, the QQQQ APS patterns with 1% target and 2% stop have generated 20 trades, out of which 16 were were profitable. Win rate = 80%... for target 2% and stop 4%. Now the win rate is 73%
More...
Very interesting. Thanks for posting.

Have you tested 1% target/1% stop? How about 2% target/1% stop? 4% target/2% stop?
 
Quote from Kevin Schmit:

Very interesting. Thanks for posting.

Have you tested 1% target/1% stop? How about 2% target/1% stop? 4% target/2% stop?
More...

Interesting question. I have done extensive testing with APS and I have found out that it is more capable of finding robust patterns with a high win rate and target/stop < 1. This may have to do with the markets, not with the program I suspect.
 
GoodGoing, your work is interesting. I'm just wondering if you have back tested the results using no stop and no target, just the price on close as your exist. I'm curious to know if the system is still profitable under these conditions.

cheers,

Runningbear
 
Quote from Runningbear:

GoodGoing, your work is interesting. I'm just wondering if you have back tested the results using no stop and no target, just the price on close as your exist. I'm curious to know if the system is still profitable under these conditions.

cheers,

Runningbear
More...

Thanks Runningbear. I admit I do a lot of testing, close to 6 -8 hours a day, and during the weekends too. Now, this is what I have done. I took the QQQQ patterns and I applied them to SPY data just to see whether they were particular to that series.

The following is the Amibroker bactesting report for the QQQQ patterns from 01/2010 to 08/03/2010, with 1% target and 2% stop on SPY data. They generated 30 trades, out of which 24 were were profitable. Win rate = 80%. The profit factor is 2.12 slightly higher than the value of 2.0 I calculated from the APS results for the QQQQ patterns. This is the Ami report:

http://www.4shared.com/photo/Xi0NS04f/T1S2.html

Next, I removed the target and stop and I made each trade to exit at the close of the bar. The profit factor was 1.08, showing a slight winning bias.

http://www.4shared.com/photo/JWqJ2Vx4/T1S2_onclose.html

This study I found even more interesting. You can get these results from Price Action Lab, a program similar to APS using just the demo, you do not have to purchase the software. The only thing is you will not be able to generate code for Amibroker but I have already found the results to agree well. Actually, I think Ami is the most accurate backtesting program I have ever used.
 
You must log in or register to reply here.
Back
Top