Mizhael recently hijacked a thread on backtesting and portfolio simulation technologies, anybody knows more details?
HFT is very <b>de</b>l·e·te·ri·ous/ˌdeliˈti(ərēəs/ Adjective: Causing harm or damage
I certainly think have high frequency data is helpful to detect market shifts, I, however, don't really believe an individual investor can find anything profitable in HFT if that is the goal and if that is why you gathered tick data. The idea is not hard to understand. To play HFT, you need a co -located servers, fast execution engine that is written by the best programmers and maybe a data pipe between new york and chicago (cost 1.2 million setup + 200K+ per year) What are the general directions for the type of strategies you worked on if you don't mind talk about them just a little bit?
Found this site a couple of weeks ago: http://www.tradesignalonline.com/default.aspx Quite a few free strategies to back test. You can join and create your own.
And if they don't create data mining bias themselves (by not using genetic algos and some such) and discover something useful during the first program run (as they should), then the problem will be in the pattern database summarizing the collective data mining efforts of not only the vendor, but of many other researchers and practitioners that came here before... controlling for which is equally impossible as fiding out and correcting for the number of significance tests performed by AI black boxes. Guys, before buying such software (for OPM), check if individual creative effort is no longer required in academic research and new drugs get FDA-approved every day of the week...
No TSL demo trial available and the price is in the 60K range. There is another product called Adaptrade Builder but a user is complaining it is a scam: http://strategytraders.org/index.php?topic=117.0 and this user complains that strategies quickly fade: http://groups.google.com/group/adaptrade-builder/browse_thread/thread/3f6e7dcbf68c5ea2?hl=en I have no way of checking their conclusions. PAL is OK if you are interested only in OHLC patterns. It offerr true OSS and additional capability for checking for randomness based on a test of profitability across several markets. Code is fully disclosed.
I actually started to check more systematically what the PAL previsions would come to this year following the results on DJIA stocks on the blog - for deltas over 10 and Significance at 8 or more. It doesn't look good so far and it might be one of the reasons Michael Harris stopped posting those. The main issue is the patterns found through PAL might be completely random. Can you find a distinctive predictive edge to the PAL patterns ?
I think he is still posting but only for PAL users. Regardless, I think you are talking about the indicator part because I have seen the posts. I have also seen posts with patterns. The last one about TLT was right on the post with perfect timing: http://www.priceactionlab.com/Blog/...rding-to-technical-and-price-action-analysis/ The long SPY pattern earlier this month had also perfect timing. SPY opened lower the next day and just went straight up to hit the target: http://www.priceactionlab.com/Blog/2012/03/a-short-term-bullish-pattern-in-spy/ But keep in mind that these examples are from the indicator and the scan. I do not think I have seen any system examples in the blog but there are some in the product website. The indicator value changes each day after the close. I do not like indicators anyway. I think the pattern scan is much more interesting.