hello, here is a system i developed, the basic strategy is: at end of each day, use updated historical data and some algorithms to rank stocks, the next day select the top ranked stocks, of these stocks those which also pass certain criteria hold for 3 days and sell at the close of the third day. Its possible on certain days no stocks are selected to trade. If there are selected stocks allocate 1/3 of funds into these stocks evenlyso here are the results trading from 2006-2011:total return on equity: 38.64xreturn(annualised) 107.69%sharpe ratio(annualised) 29.58total trades 674total profitable trades 441profitable trades% 65.43%(Chinese index over the 5 years grew a little under 100%)

Is that a back tested system or one you developed in 2006 and walked forward ?

whats the difference

The difference is that anyone can come up with a trading rule that 'would have' generated a profit.But if you do this with historical data you are just data mining, it is meaningless unless you have your rule a priori and trade with it.

Thanks TinMan. That was an odd question from Richyiee. But then again I've come across so many people who have developed systems based on backdated studies that were over fitted. In fact I'm sure someone brighter than I could make a convincing case for doing the exact opposite of what a found system tells you. Particularly when small sample sizes are involved.

i'll let you know in a few months if im broke ;p

QuoteOriginally posted by: Richyieei'll let you know in a few months if im broke ;por rich !

QuoteOriginally posted by: daveangelQuoteOriginally posted by: Richyieei'll let you know in a few months if im broke ;por rich !do you really think Richyiee will tell if he gets rich thank to his trading system ??

QuoteOriginally posted by: edouardQuoteOriginally posted by: daveangelQuoteOriginally posted by: Richyieei'll let you know in a few months if im broke ;por rich !do you really think Richyiee will tell if he gets rich thank to his trading system ?? I would if I got rich trading his system

If I may ask, what is the total number of parameters that you calibrate. This could probably hint if there is any data mining.

Is it a long-short strategy or just long-only?

its a long only strategy, i monitor and calculate likelihood of large set of "events", high probability events are traded. The "likelihood" of events will change through time, e.g some high probable events from say i = 1:400 may be less probable from i = 400 onwards, these changes are accounted for in the system.

you sell at the close, but where do you buy? the open? once you have the signal, do you assume you get into the position the next day? if yes, how? i mean, you have to make an assumption on where you would get filled.

yes, next days open. Or most orders are able to get filled the next morning at todays close, although on average next days open is 0.002 lower than todays close so id be loosing a little bit of profits that way. btw are you making immature inuendos

QuoteOriginally posted by: Richyieeyes, next days open. Or most orders are able to get filled the next morning at todays close, although on average next days open is 0.002 lower than todays close so id be loosing a little bit of profits that way. btw are you making immature inuendosNothing immature. Finance is not about averages...