The Kelly system works by leveraging the odds...finding spots where there are overlays and betting a portion of your bankroll based on how much of an overlay it is. By taking into the account the expected rate of return and the risk involved Kelly's utility function maximizes the overall growth of your bankroll. Betting more than the suggested Kelly amount is an undue risk; betting less than the Kelly amount will cause your bankroll to grow more slowly ...
And that's what I found back in your graphs ... Regards, Ton. ----- Original Message ----- From: Edward Pottasch To: [email protected] Sent: Wednesday, August 27, 2008 10:28 AM Subject: Re: [amibroker] Re: backtest / equity curve hi Ton, wasn't the Kelly principle when betting with a coin you put in 25% of your funds on every wager within certain rules. Not sure. I just wanted to test an intraday trading stop when a certain profit is reached since sometimes I see a nice profit evaporate in thin air. This evolved from an original intraday system, working superb, but that was hard (if not impossible) to execute in the practice. I made adjustements and now I can execute the system very closely in the practice but ofcourse now the profits are pretty crappy. I noticed however that many times I reach a certain profit during the day and later that day I may loose it all. This is what I am testing now. regards,Ed ----- Original Message ----- From: Ton Sieverding To: [email protected] Sent: Wednesday, August 27, 2008 9:54 AM Subject: Re: [amibroker] Re: backtest / equity curve Ed are you showing us the Kelly Principle ? Regards, Ton. ----- Original Message ----- From: Edward Pottasch To: [email protected] Sent: Wednesday, August 27, 2008 8:59 AM Subject: Re: [amibroker] Re: backtest / equity curve hi Mike, I guess it could but I solved it in a more simple manner. Just used the equity curve from the original backtest. So I can put this in a chart below the system. When I run the backtest the curves taking profit at 500$/day, 1000$/day etc. are the automatically plotted along side the original equity curve, see chart. Clear from the chart is that taking a profit and stop for the day works very well for this typical system. regards, Ed ----- Original Message ----- From: Mike To: [email protected] Sent: Tuesday, August 26, 2008 10:13 PM Subject: [amibroker] Re: backtest / equity curve Couldn't custom backtester code be written to just set positionsize to zero for all subsequent trades after reaching your daily maximum? That way you would not have to alter your actual trade logic at all. Whenever possible, I prefer to keep my trade logic in its original form, then handle constraints in easily swapped custom backtest code. Mike --- In [email protected], "Edward Pottasch" <[EMAIL PROTECTED]> wrote: > > hi Barry, > > thanks for your reply. I am close to a solution myself. > > I tried something similar as you suggest but the problem I am having is that if the daily target is reached it needs to jump to the next day of intraday data and not jump out the loop entirely. I am close to the solution, > > regards, Ed > > > > ----- Original Message ----- > From: Barry Scarborough > To: [email protected] > Sent: Tuesday, August 26, 2008 5:06 PM > Subject: [amibroker] Re: backtest / equity curve > > > Hi Edward, > > You can keep track of your daily gain and when that reaches your > threshold block your buy and short signals. For instance you might > use buyPrice and sell price to calculate gain / trade that keep gain > in its own array or static var, I prefer the latter, and just bump > the gain each time a trade is complete. > > buy = buyCondition and gain < threshold. > > This way the back tester never sees the additional or potential > trades and will give you better results. Also in back testing if you > use buyPrice, etc., if you are using conditions that occur mid bar, > such as MA cross, then the back tester gives more accurate results. I > had to use explore and dump that to excel to finally get the back > tester to give correct results. Using close of the bar give incorrect > results. > > Barry > > --- In [email protected], "Edward Pottasch" <empottasch@> > wrote: > > > > hi, > > > > anyone ever do a backtest on intraday data to exit trading for the > day when a certain profit is achieved? In the chart below the > intraday profit / loss curve is shown. I guess I could put this data > in a composite and use the compositie during a more detailed > backtest. Still the coding is pretty difficult since when the daily > profit is achieved you want the backtest to continue for the next day > of intraday data. > > > > Another way to test it would be to just analyze the data below. For > instance if +1000$ is achieved intraday jump to the next day, if not > take the profit/loss value at the close and just add these value. Not > so easy to program or have I been drinking too much liquor lately? > > > > Wonder if anyone solved this problem or sees the solution? > > > > thanks, Ed > > >
