Hi i just developed my first mechanical strategy... after lots of testing i'm thinking about starting to trade it with a small number of contracts to see how it performs in real time I've attached a spreadsheet with the backtested results (using 60 min data from 1/1/07 to 12/30/07) I've also tested with different periods and there isn't much difference from the results... I would like your opinion on the results mainly regarding what type of "pitfall" i should be aware of I didn't use a position sizing algorithm so all trades were tested with U$100.000 and $150 slippage. Thanks!
I did test it with 2008 data from 1/1/08 to 5/30/08 and it got about the same results. The one thing i'm worried about and trying to improve is the avg win to avg loss ratio... however i think that the % of profitable trades is offsetting it...
test on another 100 stocks going back 5 years and if those results look the same then id say it looks pretty good don't know why you only chose 9. i hope you didn't cherry pick them based on results
Actually this is a pairs trading strategy, and these 9 stocks are the ones that i usually trade, however i didn't trade them mechanically, i had my system but i also used discretion while trading... These stocks were hand picked based on cointegration tests and my trading portfolio wouldn't change much from these nine... (these aren't stocks but spreads between them), but i'll try adding a few more spreads and testing with them I didn't cherry pick them, as you can see there are a couple of stocks that don't perform very well... i was very concerned to avoid curve fitting, the only optimization i've done was on some parameters length
I am thinking about starting spread trading myselfe! So I think this sounds very interesting. Maybe you could desbribe what strategy you are using? I am curently trading 2 different strategies, which only applies to intraday trading. It looks like you are planing to hold your positions several days? Veggen
It is a mean reversion type of strategy using stop loss and a volatily based objective. Even though most trades turn out to be daytrades some last more time... Now that i am trading the strategy for real i can say that what differs most from the results is the time in trade, while in backtest i had an average of 2.5 ~ 3 days for time in trade, and now running the strategy it seems that the average time in trade is closer to 1.5 days.
Hi, I have a few questions: How does this perform if you were trading all 9 stocks at the same time? As a portfolio, what are the stats? Are all of the stocks related, ie same industry? If so then I think you should test on a longer time period or a period when that industry group performed differently then in 2007.
I don't know how to backtest a portfolio in tradestation, however using this strategy last month gave me aprox. 1% for the month (using some discretion to position sizing). The spreads are between stocks in the same industry however from most spreads are from different industries and they all got bullish and bearish periods in the backtested data.
The portfolio bit of it will show you what your capital usage is at any given day and drawdowns on all positions. What if you had signals to get long all 9 stocks and then all 9 drewdown? What's your pain threshhold. Other point is that even intra-industry can get highly correlated and then not so much through different time periods. I think these are fairly basic questions that need to be answered before applying any real money. Hope this helps