Try gainforex for data. There is also Olsen & Associates. They are supposed to have high quality stuff, but You have to pay, and I mean to PAY...

There is also the Wiki Wilmott website with several free data links .

This is perhaps an obvious question and I didn't notice if anyone else asked it yet, but did your backtesting of the algorithm include transactions costs? I ask this because some apparently profitable trading rules end up not covering transactions costs.

QuoteOriginally posted by: DavidJNThis is perhaps an obvious question and I didn't notice if anyone else asked it yet, but did your backtesting of the algorithm include transactions costs? I ask this because some apparently profitable trading rules end up not covering transactions costs.You're right. That's one reason why I'm not willing to implement the algorithm myself as I don't have an account with cheap transaction fee.

QuoteOriginally posted by: DavidJNThis is perhaps an obvious question and I didn't notice if anyone else asked it yet, but did your backtesting of the algorithm include transactions costs? I ask this because some apparently profitable trading rules end up not covering transactions costs.Think about your total transaction costs too. Things I look at (trading equities) include:Execution cost - usually a fixed cost per shareMarket Impact - if you are assuming an intraday execution, market impact can be huge even for small ordersSlippage - again, can be large especially if shortingStock Borrow - can be anywhere from 50 bps to 10% per annumDividends - are you a net payer or receiverFinancing - what are your net financing costsTicket Charges - if you're generating a large number of small orders, these can be substantial

Given your results, I wouldn't worry that much about transaction costs except for market impact and slippage. And I'd only worry about those if you are trading illiquid stocks or stocks under $10/share. Your backtest suggests you will average 3% per roundtrip (thats $300 profit on a $10k position) which is more than enough to cover transaction costs at a retail online broker.Transaction costs are really only a major issue for day traders that are trying to pull pennies per share in profit from small intraday moves and/or for undercapitalized traders for which a 3% move is only $30 on a $1000 account. Decently capitalized long-term traders that generally have much higher average profits per trade aren't going to be killed by transaction costs.You really do need to try trading this -- you'll learn a lot in the first few round-trips. I'd recommend actual trading AND collecting the market data for simulated trading -- You may discover that your executions don't occur when (or at the price) you think they will occur and that the strategy has different risk return characteristics that the backtest analysis shows. No amount of simulation or backtesting will reveal that fact.Good Luck!

Thanks for all comment!One last question, if I really would like to write things up and try to submit to journal, which journals should I submit to?

for publication, you might think to translate your algorithm into easylanguage/tradestation and send it to futurestruth.com, they will do a performance ranking w/o publishing the algorithm itself. nevertheless, i have only seen that, but i don't have any experiences about how trustful this service is (the performances published seems to be a little bit too nice )

QuoteOriginally posted by: actuaryalfredThanks for all comment!One last question, if I really would like to write things up and try to submit to journal, which journals should I submit to?Practitioner journals would be more interested in this stuff: The Journal of Behavioral Finance, Financial Analyst Journal, Journal of Portfolio Management, etc.

I always thought the most obvious thing to do if you created a successful trading algorithm was to trade it and keep it for yourself, ior if it doesnt really work, sell it to someone who would buy it... is there a benefit to publishing it that i'm unaware of?

It seems to me that any trading algorithm has a finite useful life.The more important issue is what you've found.If you assemble together enough combinations of functions, it is certain that some of them will appear to do better than a simple portfolio.But a more profitable process is where you've found some mechanism in the way your market works that forecasts certain types of change under conditions that occur often enough to be useful.That means you have a family of trading processes, and can use a small subset of them to prove yourself to a backer.

Total number of trade is 141Mean return is 0.0313794SD is 0.0306327Maximum is 0.188679Minimum is -0.01162790.01 quantile is -0.007810730.05 quantile is 0.0020.1 quantile is 0.00557491Median is 0.0230263Geometric mean is 0.0309463Looking at the trage logs, if your mean return is expressed in percents, then it's 3.1% and it sounds reasonable. Though then you have a little problem that standard deviation has the same magnitude as the return. I hope you didn't optimize it yet and actual business is better.And by a 3% return per transaction you really don't have to worry about transaction costs, though this business requires being thorough and accurate, so i do advise including those as well. Also, do the bid/ask spread to estimate worst case. Even if these don't crash your strategy, believe me you'll find it useful when talking to people.Another thing is the backtesting period. You need it long, and well, you might find it stops working somewhere. Best thing is if your model tells you in what market conditions it works and in what it doesn't. In any case, this is normal, since markets change. You might find you need to monitor trading and slightly adjust its parameters to keep it going. Best of luck, let us know when you start a hedge fund!